Predistortion processing method and device

Abstract

Embodiments of the present application provide a predistortion processing method and apparatus, comprising: obtaining a first continuous time-series signal containing a time-series signal to be processed; normalizing the first continuous time-series signal to obtain a corresponding second continuous time-series signal; and inputting the second continuous time-series signal into a pretrained back-propagation BP neural network model, thereby outputting a predistorted time-series signal corresponding to the time-series signal to be processed. The BP neural network model is an inverse behavioral model of the corresponding target power amplifier (PA). This solution utilizes an inverse behavioral model based on a BP neural network and normalizes the time-series signal input to the model, thereby improving the nonlinearity of the overall system and the accuracy of the predistortion processing.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a pre-distortion processing method and device. Background Art

[0002] A power amplifier (PA) amplifies modulated frequency signals to the desired power level. As the largest energy consumer in a communications base station, improving its efficiency is crucial. PAs exhibit nonlinear characteristics and memory effects, which lead to in-band distortion and out-of-band spectrum spread, directly impacting base station efficiency and transmitted signal quality.

[0003] Digital pre-distortion (DPD) technology builds an inverse behavioral model of the PA and inserts it before the PA in the link. This pre-processes the signal before it enters the PA, ensuring that the power amplifier maintains linearity even in the high-power range, improving the quality of the transmitted signal. Traditionally, inverse behavioral models for DPD have been derived using generalized memory polynomials (GMPs), which are simplified versions of the Volterra series model.

[0004] However, with the surge in the number of wireless communication users in recent years, especially the gradual commercialization of 5G (5th Generation mobile networks), bandwidth and frequency have also increased accordingly, which requires a larger memory depth and nonlinear order, and has potential numerical instability. In this context, the pre-distortion processing accuracy of the traditional GMP model is low. Summary of the Invention

[0005] The purpose of this application is to solve at least one of the above technical deficiencies. The technical solutions provided by the embodiments of this application are as follows:

[0006] In a first aspect, an embodiment of the present application provides a predistortion processing method, including:

[0007] Acquire a first continuous time series signal including a time series signal to be processed;

[0008] Normalizing the first continuous time series signal to obtain a corresponding second continuous time series signal;

[0009] The second continuous time series signal is input into the pre-trained back-propagation BP neural network model, and a pre-distorted time series signal corresponding to the time series signal to be processed is output, wherein the BP neural network model is an inverse behavior model of the target power amplifier PA used to power amplify the time series signal to be processed.

[0010] In an optional embodiment of the present application, the first continuous timing signals include a first preset number of timing signals before the timing signal to be processed, and a second preset number of timing signals after the timing signal to be processed.

[0011] In an optional embodiment of the present application, normalizing the first continuous time series signal to obtain a corresponding second continuous time series signal includes:

[0012] Obtaining a training data set, and obtaining a mean and a standard deviation of time series data samples in the training data set, wherein the training data set includes a third preset number of PA input time series signal samples and corresponding PA output time series signal samples of the target PA;

[0013] Based on the mean and the standard deviation, obtaining a normalized time series signal corresponding to each time series signal in the first continuous time series signal;

[0014] A second continuous time series signal is acquired based on each normalized time series signal.

[0015] In an optional embodiment of the present application, the method includes:

[0016] receiving, through the input layer, a second continuous time series signal, and converting the second continuous time series signal into an initial feature vector;

[0017] Through the hidden layer, based on the initial feature vector, the corresponding abstract feature vector is obtained;

[0018] Through the output layer, the pre-distorted time series signal is output based on the abstract feature vector.

[0019] In an optional embodiment of the present application, the training process of the BP neural network model is as follows:

[0020] The BP neural network model is trained using the training data set to obtain a pre-trained BP neural network model.

[0021] In an optional embodiment of the present application, the BP neural network model is trained using a training data set to obtain a pre-trained BP neural network model, including:

[0022] Input each PA output timing signal sample into the BP neural network model and output the corresponding predicted PA input timing signal;

[0023] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding PA input timing signal sample;

[0024] Based on the mean square errors, the network parameters of the BP neural network model are updated using the back propagation algorithm to obtain a pre-trained BP neural network model.

[0025] In an optional embodiment of the present application, the method further includes:

[0026] After obtaining pre-distorted timing signals corresponding to a fourth preset number of timing signals to be processed, inputting each pre-distorted timing signal into a target PA to obtain at least one corresponding PA output timing signal;

[0027] Based on each predistorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network model are updated again.

[0028] In an optional embodiment of the present application, based on each predistorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network are updated again, including:

[0029] Input each PA output signal into the BP neural network model and output the corresponding predicted PA input timing signal;

[0030] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding predistorted timing signal;

[0031] The network parameters of the BP neural network model are updated based on the mean square errors.

[0032] In a second aspect, an embodiment of the present application provides a predistortion processing device, comprising:

[0033] A timing signal acquisition module, configured to acquire a first continuous timing signal including a timing signal to be processed;

[0034] a normalization module, configured to perform normalization processing on the first continuous time series signal to obtain a corresponding second continuous time series signal;

[0035] The predistortion processing module is used to input the second continuous time series signal into a pre-trained back-propagation BP neural network model and output a predistorted time series signal corresponding to the time series signal to be processed, wherein the BP neural network model is an inverse behavioral model of the target power amplifier device PA used to power amplify the time series signal to be processed.

[0036] In an optional embodiment of the present application, the first continuous timing signals include a first preset number of timing signals before the timing signal to be processed, and a second preset number of timing signals after the timing signal to be processed.

[0037] In an optional embodiment of the present application, the normalization module is specifically configured to:

[0038] Obtaining a training data set, and obtaining a mean and a standard deviation of time series data samples in the training data set, wherein the training data set includes a third preset number of PA input time series signal samples and corresponding PA output time series signal samples of the target PA;

[0039] Based on the mean and the standard deviation, obtaining a normalized time series signal corresponding to each time series signal in the first continuous time series signal;

[0040] A second continuous time series signal is acquired based on each normalized time series signal.

[0041] In an optional embodiment of the present application, the predistortion processing module is specifically configured to:

[0042] receiving, through the input layer, a second continuous time series signal, and converting the second continuous time series signal into an initial feature vector;

[0043] Through the hidden layer, based on the initial feature vector, the corresponding abstract feature vector is obtained;

[0044] Through the output layer, the pre-distorted time series signal is output based on the abstract feature vector.

[0045] In an optional embodiment of the present application, the device further includes a first training module for:

[0046] The BP neural network model is trained using the training data set to obtain a pre-trained BP neural network model.

[0047] In an optional embodiment of the present application, the first training module is specifically used to:

[0048] Input each PA output timing signal sample into the BP neural network model and output the corresponding predicted PA input timing signal;

[0049] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding PA input timing signal sample;

[0050] Based on the mean square errors, the network parameters of the BP neural network model are updated using the back propagation algorithm to obtain a pre-trained BP neural network model.

[0051] In an optional embodiment of the present application, the device further includes a second training module for:

[0052] After obtaining pre-distorted timing signals corresponding to a fourth preset number of timing signals to be processed, inputting each pre-distorted timing signal into a target PA to obtain at least one corresponding PA output timing signal;

[0053] Based on each predistorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network model are updated again.

[0054] In an optional embodiment of the present application, the second training module is specifically used to:

[0055] Input each PA output signal into the BP neural network model and output the corresponding predicted PA input timing signal;

[0056] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding predistorted timing signal;

[0057] The network parameters of the BP neural network model are updated based on the mean square errors.

[0058] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor;

[0059] The memory stores a computer program;

[0060] A processor is used to execute a computer program to implement the method provided in the embodiment of the first aspect or any optional embodiment of the first aspect.

[0061] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method provided in the embodiment of the first aspect or any optional embodiment of the first aspect is implemented.

[0062] The beneficial effects of the technical solution provided by this application are:

[0063] Before the target PA amplifies the time series signal to be processed, the time series signal to be processed is pre-distorted by the inverse behavior model based on the BP neural network, and the time series signal input to the inverse behavior model based on the BP neural network is normalized before the pre-distortion processing. Using the inverse behavior model based on the BP neural network and normalizing the time series signal of the input model can better improve the nonlinear factors of the overall system and improve the accuracy of the pre-distortion processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments of the present application.

[0065] Figure 1 Schematic diagram of the LUT table search process in the prior art;

[0066] Figure 2 A flowchart of a pre-distortion processing method provided in an embodiment of the present application;

[0067] Figure 3 A schematic diagram of the structure of a BP neural network model in an example of an embodiment of the present application;

[0068] Figure 4a Schematic diagram comparing a normalized IQ timing signal and a non-normalized timing signal in an example of an embodiment of the present application;

[0069] Figure 4b This is a schematic diagram comparing NMSE curves of the model training process using normalized IQ timing signals and non-normalized timing signals in an example of an embodiment of the present application;

[0070] Figure 5 This is a schematic diagram of the results of a BP neural network model in another example of an embodiment of the present application;

[0071] Figure 6 This is a feature combination method of the input layer in an example of an embodiment of the present application;

[0072] Figure 7 A schematic diagram of a training process of a BP neural network model in an example of an embodiment of the present application;

[0073] Figure 8 A structural block diagram of a pre-distortion processing device provided in an embodiment of the present application;

[0074] Figure 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0075] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and are not to be construed as limiting the present application.

[0076] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or wireless couplings. The term "and / or" used herein includes all or any units and all combinations of one or more associated listed items.

[0077] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0078] The digital predistortion modeling solution in existing base stations can be expressed as the following formula using the GMP model:

[0079]

[0080] In formula (1), when establishing the inverse mode of the power amplifier (i.e., the inverse behavior model), x is the power amplifier output signal, and y is the power amplifier input signal. The three parameters i, j, and k represent the signal memory depth, signal modulus memory depth, and signal modulus nonlinearity order, respectively, and b is the model coefficient. In a real-time system, for speed and resource considerations, the trained GMP model is usually implemented using a lookup table (LUT). The LUT table is looked up based on the signal amplitude, and the value in the table is the cumulative value of the multiplication of the distortion coefficient and the signal modulus nonlinear term. The number of LUT tables is determined by the model memory depth i and j. Each (i, j) combination corresponds to a LUT table. The values ​​stored in the table are calculated according to formula (2), where AMP is the quantized signal amplitude.

[0081]

[0082] like Figure 1 The figure shows the LUT table search process with a signal memory depth of i and an amplitude memory depth of j. The Z in the figure represents the delay. For example, in a set of time series signals, there is x(n). After Z -j After the change, we get x(nj), abs represents the modulus value, abs(x(nj)) is the |AMP| in formula (2), and combined with the delay Z -i Get x(ni), and then get:

[0083]

[0084] Then, all the y ij(n) The sum is y(n).

[0085] The traditional GMP model is simplified from the Volterra series model. As the bandwidth and frequency increase, it often requires a larger memory depth and nonlinear order, and has potential numerical instability. When applied to power amplifier models with strong memory effects and strong nonlinearities, the modeling accuracy is not high. In other words, in this context, the pre-distortion processing accuracy of the traditional GMP model is low.

[0086] To address the above problems, embodiments of the present application provide a pre-distortion processing method, apparatus, electronic device, and computer-readable storage medium. The solution provided in the present application will be described in detail below.

[0087] Figure 2 A flow chart of a pre-distortion processing method provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the method may include the following steps:

[0088] Step S201 : Acquire a first continuous time series signal including a time series signal to be processed.

[0089] Among them, the time series signal to be processed is the time series signal that is pre-distorted before entering the power amplifier. Among them, the continuous time series signal can be understood as multiple time series signals obtained by sampling at a certain time interval. If the time series signal to be processed corresponds to the current sampling moment, the time series signal collected at the sampling moment before the current sampling moment can be called the time series signal before the time series signal to be processed, and the time series signal collected at the sampling moment after the current sampling moment can be called the time series signal after the time series signal to be processed. Then the first continuous time series signal is the time series signal including the time series signal to be processed, the time series signal before the time series signal to be processed, and the time series signal after the time series signal to be processed. In the subsequent steps, the BP neural network model needs to be used to obtain the pre-distorted time series signal of the time series signal to be processed based on the characteristics of the time series signals before and after the time series signal to be processed. Therefore, it is necessary to obtain the first continuous time series signal here.

[0090] Step S202 : performing normalization processing on the first continuous time series signal to obtain a corresponding second continuous time series signal.

[0091] Specifically, since the system involves a large amount of data, whether in training or in the online pre-distortion processing process, and may also introduce amplitude order information, different features in the data will have different orders of magnitude. When solving the coefficients, the gradient weights of large features will be larger, and the training speed will be reduced when using the loss function to find the optimal value. In addition, for some activation functions commonly used in neural networks, such as sigmoid, tanh, etc., when the data is too large, the derivative will appear in the saturation region of the activation function, resulting in extremely small gradients of the parameters, which is not conducive to error back propagation. Therefore, before the time series data is input into the input layer of the BP network model, the data needs to be normalized. Specifically, it is necessary to normalize each time series signal contained in the first continuous time series signal to obtain the corresponding second continuous time series signal.

[0092] In step S203, the second continuous time series signal is input into a pre-trained back propagation BP (Back Propagation) neural network model, and a pre-distorted time series signal corresponding to the time series signal to be processed is output, wherein the BP neural network model is an inverse behavioral model of the target power amplifier PA used to power amplify the time series signal to be processed.

[0093] Among them, the structure of the BP neural network model is as follows Figure 3 As shown in Figure 1, it can consist of three parts: an input layer, a hidden layer, and an output layer. The input layer contains the structured data required to build the inverse behavior model. The number of hidden layers can be arbitrary. The hidden layers generate nonlinear features through nonlinear activation functions and linear weighting, propagating useful feature information to the next layer in the form of dense feature vectors. The initial coefficients of the neural network are generally initialized with random values, and the training method is gradient descent. This training method requires multiple iterations on the dataset for each model training.

[0094] Specifically, the normalized second continuous data is input into a pre-trained BP neural network model, and a corresponding pre-distorted time series signal is output. The corresponding pre-distorted time series signal is input into a corresponding target PA to achieve amplification of the time series signal to be processed.

[0095] The solution provided in the present application performs pre-distortion processing on the time series signal to be processed through an inverse behavior model based on a BP neural network before amplifying the time series signal to be processed through a target PA, and normalizes the time series signal input to the inverse behavior model based on a BP neural network before the pre-distortion processing. The use of the inverse behavior model based on a BP neural network and the normalization processing of the time series signal of the input model can better improve the nonlinear factors of the overall system and improve the accuracy of the pre-distortion processing.

[0096] In an optional embodiment of the present application, the first continuous timing signals include a first preset number of timing signals before the timing signal to be processed, and a second preset number of timing signals after the timing signal to be processed.

[0097] The first preset number and the second preset number can be set according to actual needs. For example, the first preset number and the second preset number can be set to the same number.

[0098] In an optional embodiment of the present application, normalizing the first continuous time series signal to obtain a corresponding second continuous time series signal includes:

[0099] Obtaining a training data set, and obtaining a mean and a standard deviation of time series data samples in the training data set, wherein the training data set includes a third preset number of PA input time series signal samples and corresponding PA output time series signal samples of the target PA;

[0100] Based on the mean and the standard deviation, obtaining a normalized time series signal corresponding to each time series signal in the first continuous time series signal;

[0101] A second continuous time series signal is acquired based on each normalized time series signal.

[0102] The training data set is also used to train the aforementioned pre-trained BP neural network model. In other words, while training the pre-trained BP neural network model, the mean and standard deviation required for subsequent normalization processing of the time series signal are determined.

[0103] The number of time series signal samples included in the training data set can be set according to actual needs, that is, the third preset number can be set according to actual needs.

[0104] Specifically, the process of normalizing each time series signal included in the first continuous time series signal can be expressed as the following formula:

[0105]

[0106] Where x k is the timing signal contained in the first continuous timing signal, x′ k The corresponding normalized time series signal is the time series signal contained in the second continuous time series signal. μ is the mean of the sample data, σ is the standard deviation of the sample data, and the normalized data conforms to the normal distribution, that is, the mean is 0 and the standard deviation is 1.

[0107] Furthermore, for a time series signal data set, each time series signal can be represented as a real part (In-phase, referred to as I) component and an imaginary part (Quadrature, referred to as Q) component. Then the time series signal can also be called an IQ time series signal, or IQ data or IQ format data. Figure 4a As shown in the figure, the horizontal axis in the left and right figures represents the real component of the time series signal (i.e., I), and the vertical axis represents the imaginary component of the time series signal (i.e., Q). The left figure is an unnormalized data set, and the right figure is a normalized data set. By comparing the left and right figures, it can be seen that the order of magnitude of the normalized data is significantly reduced (i.e., the order of magnitude of the corresponding horizontal and vertical axes is significantly reduced), which is convenient for model processing.

[0108] like Figure 4b As shown, the BP neural network model is trained using an unnormalized data set (left figure), and the BP neural network model is trained using a normalized data set (right figure). Specifically, the data is divided into a training set (corresponding to curve 401 in the two figures), a validation set (corresponding to curve 402 in the two figures), and a test set (corresponding to curve 403 in the two figures). The horizontal axes in the two figures represent the number of training times, and the vertical axes indicate the corresponding normalized mean square error (NMSE). The corresponding curves can be called NMSE curves. Comparing the left and right figures, it can be seen that using the normalized data set for training makes the model converge faster, and the obtained model pre-distortion processing effect is better. In summary, during the training and online pre-distortion processing processes, the time series signal can be normalized first, and then the normalized time series signal can be used for training or online pre-distortion processing.

[0109] In an optional embodiment of the present application, the BP neural network includes an input layer, a third preset number of hidden layers, and an output layer, inputs the second continuous time series signal into a pre-trained back-propagation BP neural network model, and outputs a pre-distorted time series signal corresponding to the time series signal to be processed, including:

[0110] receiving, through the input layer, a second continuous time series signal, and converting the second continuous time series signal into an initial feature vector;

[0111] Through the hidden layer, based on the initial feature vector, the corresponding abstract feature vector is obtained;

[0112] Through the output layer, the pre-distorted time series signal is output based on the abstract feature vector.

[0113] Among them, the BP neural network model is as follows Figure 5As shown, it includes an input layer, a hidden layer (i.e., a hidden layer) and an output layer. Specifically, a normalizer is provided before the input layer for normalizing the first continuous time series data. The input layer mainly involves how to combine the input time series signal into the input format required by the BP neural network model (i.e., converting from the initial feature vector). The main function of the hidden layer is to abstract the input information and discover implicit features so as to represent the input information as a feature vector (i.e., obtain an abstract feature vector), wherein the number of hidden layers (i.e., depth) can be set according to demand. In the embodiment of the present application, the number of hidden layers is 2, which are hidden layer 1 and hidden layer 2 in the figure respectively. The main function of the output layer is to convert the abstract feature information into the target IQ format after pre-distortion, that is, to output the pre-distorted time series signal corresponding to the time series signal to be processed.

[0114] Specifically, the input layer converts the input time series signal into the input of the BP neural network. Figure 6 The matrix shown in the figure lists the information of the input layer. From top to bottom, each row of the matrix corresponds to the memory depth of the signal, the value of the signal at the current moment, and the delay information. I and Q represent the imaginary and real part information corresponding to the time series signal. A^n represents the order information of the signal amplitude. "Other" is other expandable information to facilitate the expansion of the model input. When using the BP neural network model, before the feature matrix enters the hidden layer, it is necessary to Figure 6 The matrix representation in is converted into a preset vector form (i.e., initial eigenvector), and the vector dimension is 1*n.

[0115] The depth and size of the hidden layer determine the upper limit of the learning of the BP network model. However, simply increasing the depth and size does not directly improve the model effect. Instead, it makes the model training difficult to converge. Through experimental testing, the present invention uses a two-layer hidden layer structure. The calculation expression of each hidden layer is shown in (4):

[0116] h=f(X input *W h +b) (4)

[0117] Among them, X input is the input of this layer, W h It is an n*m matrix, where n represents the output dimension of the upper layer, m represents the number of neurons in this layer, f(x) is the activation function, and b is the bias.

[0118] Activation function: The introduction of the activation function is to give the network model the ability to learn nonlinear functions. In this embodiment of the application, the ReLU function can be used as the activation function. The formula is shown in (5):

[0119] a=max(0,z) (5)

[0120] Among them, a is the output of the activation function, and z is the calculation result of the linear change of the neuron.

[0121] The output layer uses a fully connected linear unit for output, as shown in formula (6):

[0122] y output =h*W o +b (6)

[0123] Wo is an m*2 matrix, where m represents the output dimension of the hidden layer and 2 represents the 2D vector output of the real and imaginary parts.

[0124] In an optional embodiment of the present application, the training process of the BP neural network model is as follows:

[0125] The BP neural network model is trained using the training data set to obtain a pre-trained BP neural network model.

[0126] Specifically, if Figure 7 The figure shows a schematic diagram of an online real-time system corresponding to an embodiment of the present application. The training of the BP neural network model can be divided into two stages. The first is an offline training stage, which is the stage of obtaining the pre-trained BP neural network model in the basic application, and the second is a gradual optimization stage during online use. Each training process requires a large number of iterative calculations. Considering that the base station has a long training time for the first training after power-on, the application embodiment adopts the above two-stage training schedule, selects a large amount of data with different power as a pre-training data set, and trains a model in an offline state. In the online real-time system, the pre-trained model will be used as the training starting point for gradual optimization. Among them, the structure of the pre-trained BP neural network model can be set according to different power amplifier models.

[0127] It should be noted that during the offline training phase, the parameters of the normalizer are determined at the same time. In other words, it can be understood as determining the mean and standard deviation in the normalization process. Then, the parameters of the normalizer remain unchanged during the subsequent online use process and the online update phase.

[0128] Furthermore, during the offline training phase, the BP neural network model is trained using a pre-acquired training data set, where the training data used is the PA input timing signal samples and the corresponding PA output timing signal samples of the target PA. Specifically, each PA output timing signal sample is input into the BP neural network model, which outputs the corresponding predicted PA input timing signal. Based on each predicted PA input timing signal and the corresponding PA input timing signal sample, the corresponding mean square error is obtained. Based on each mean square error, the network parameters of the BP neural network model are updated using the backpropagation algorithm, thereby obtaining a pre-trained BP neural network model.

[0129] In an optional embodiment of the present application, the method further includes:

[0130] After obtaining pre-distorted timing signals corresponding to a fourth preset number of timing signals to be processed, inputting each pre-distorted timing signal into a target PA to obtain at least one corresponding PA output timing signal;

[0131] Based on each predistorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network model are updated again.

[0132] Specifically, the above process is the second-stage training process. Specifically, each PA output signal is input into the BP neural network model, and a corresponding predicted PA input timing signal is output; based on each predicted PA input timing signal and the corresponding pre-distorted timing signal, a corresponding mean square error is obtained; and based on each mean square error, the network parameters of the BP neural network model are updated. It will be understood that the obtained PA output signal can also be referred to as feedback data in the figure, that is, the network parameters of the BP neural network model are updated using multiple feedback data. The network parameter update frequency can be set according to actual needs, for example, network parameter updates can be performed at intervals of a preset duration. It will be understood that the fourth preset number can be set according to actual needs.

[0133] It should be noted that the objective function used in the two-stage training process can be to minimize the mean square error, as shown in formula (7):

[0134]

[0135] y represents the IQ data before the power amplifier (i.e., the PA input timing signal sample or the pre-distorted timing signal), which is the expected value of the power amplifier inverse behavior model. output is the output of the BP neural network model (the predicted PA input timing signal). Since the BP network is solved using the gradient descent algorithm, the cutoff condition for training an available model iteration is when the number of iterations exceeds the preset maximum number of iterations or the objective function does not decrease within the specified number of iterations.

[0136] Figure 8 This is a structural block diagram of a pre-distortion processing device provided in an embodiment of the present application, such as Figure 8 As shown, the apparatus 800 may include: a timing signal acquisition module 801, a normalization module 802, and a pre-distortion processing module 803, wherein:

[0137] The timing signal acquisition module 801 is used to acquire a first continuous timing signal including a timing signal to be processed;

[0138] The normalization module 802 is used to perform normalization processing on the first continuous time series signal to obtain a corresponding second continuous time series signal;

[0139] The pre-distortion processing module 803 is used to input the second continuous time series signal into a pre-trained back-propagation BP neural network model, and output a pre-distorted time series signal corresponding to the time series signal to be processed, wherein the BP neural network model is an inverse behavioral model of the target power amplifier device PA used to power amplify the time series signal to be processed.

[0140] The solution provided in the present application performs pre-distortion processing on the time series signal to be processed through an inverse behavior model based on a BP neural network before amplifying the time series signal to be processed through a target PA, and normalizes the time series signal input to the inverse behavior model based on a BP neural network before the pre-distortion processing. The use of the inverse behavior model based on a BP neural network and the normalization processing of the time series signal of the input model can better improve the nonlinear factors of the overall system and improve the accuracy of the pre-distortion processing.

[0141] In an optional embodiment of the present application, the first continuous timing signals include a first preset number of timing signals before the timing signal to be processed, and a second preset number of timing signals after the timing signal to be processed.

[0142] In an optional embodiment of the present application, the normalization module is specifically configured to:

[0143] Obtaining a training data set, and obtaining a mean and a standard deviation of time series data samples in the training data set, wherein the training data set includes a third preset number of PA input time series signal samples and corresponding PA output time series signal samples of the target PA;

[0144] Based on the mean and the standard deviation, obtaining a normalized time series signal corresponding to each time series signal in the first continuous time series signal;

[0145] A second continuous time series signal is acquired based on each normalized time series signal.

[0146] In an optional embodiment of the present application, the predistortion processing module is specifically configured to:

[0147] receiving, through the input layer, a second continuous time series signal, and converting the second continuous time series signal into an initial feature vector;

[0148] Through the hidden layer, based on the initial feature vector, the corresponding abstract feature vector is obtained;

[0149] Through the output layer, the pre-distorted time series signal is output based on the abstract feature vector.

[0150] In an optional embodiment of the present application, the device further includes a first training module for:

[0151] The BP neural network model is trained using the training data set to obtain a pre-trained BP neural network model.

[0152] In an optional embodiment of the present application, the first training module is specifically used to:

[0153] Input each PA output timing signal sample into the BP neural network model and output the corresponding predicted PA input timing signal;

[0154] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding PA input timing signal sample;

[0155] Based on the mean square errors, the network parameters of the BP neural network model are updated using the back propagation algorithm to obtain a pre-trained BP neural network model.

[0156] In an optional embodiment of the present application, the device further includes a second training module for:

[0157] After obtaining pre-distorted timing signals corresponding to a fourth preset number of timing signals to be processed, inputting each pre-distorted timing signal into a target PA to obtain at least one corresponding PA output timing signal;

[0158] Based on each predistorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network model are updated again.

[0159] In an optional embodiment of the present application, the second training module is specifically used to:

[0160] Input each PA output signal into the BP neural network model and output the corresponding predicted PA input timing signal;

[0161] Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding predistorted timing signal;

[0162] The network parameters of the BP neural network model are updated based on the mean square errors.

[0163] Reference below Figure 9 , which shows an electronic device suitable for implementing the embodiments of the present application (for example, Figure 2The electronic devices in the embodiments of the present application may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (such as in-vehicle navigation terminals), wearable devices, and fixed terminals such as digital TVs and desktop computers. Figure 9 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0164] The electronic device includes: a memory and a processor, wherein the memory is used to store a program for executing the methods described in each of the above method embodiments; and the processor is configured to execute the program stored in the memory. The processor here may be referred to as the processing device 901 described below, and the memory may include at least one of the read-only memory (ROM) 902, the random access memory (RAM) 903, and the storage device 908 described below, as shown below:

[0165] like Figure 9 As shown, the electronic device 900 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 into a random access memory (RAM) 903. Various programs and data required for the operation of the electronic device 900 are also stored in the RAM 903. The processing device 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0166] Typically, the following devices may be connected to the I / O interface 905: an input device 906 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 907 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 908 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 909. The communication device 909 may allow the electronic device 900 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 9 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0167] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory 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 the network through the communication device 909, or installed from the storage device 908, or installed from the ROM 902. When the computer program is executed by the processing device 901, the above-mentioned functions defined in the method of the embodiment of the present application are performed.

[0168] It should be noted that the computer-readable storage medium mentioned above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples of computer-readable storage media can 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), 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 of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. 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. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0169] In some embodiments, the client and server can communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0170] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0171] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device:

[0172] A first continuous time series signal including a time series signal to be processed is obtained; the first continuous time series signal is normalized to obtain a corresponding second continuous time series signal; the second continuous time series signal is input into a pre-trained back-propagation BP neural network model, and a pre-distorted time series signal corresponding to the time series signal to be processed is output, wherein the BP neural network model is an inverse behavioral model of a target power amplifier PA used for power amplification of the time series signal to be processed.

[0173] 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, but not limited to, 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., via the Internet using an Internet service provider).

[0174] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of 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 box can also occur in a different order than that marked in the accompanying drawings. For example, two 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 box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0175] The modules or units described in the embodiments of the present application may be implemented in software or hardware. The name of a module or unit does not, in some cases, limit the unit itself. For example, a timing signal acquisition module may also be described as a "module for acquiring timing signals."

[0176] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0177] In the context of the present application, a machine-readable medium can be a tangible medium that can contain or store a program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, 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 of the foregoing.

[0178] The device provided in the embodiments of the present application can implement at least one of the multiple modules through an AI model, and can execute functions associated with the AI ​​through a non-volatile memory, a volatile memory, and a processor.

[0179] The processor may include one or more processors. In this case, the one or more processors may be general-purpose processors, such as a central processing unit (CPU), an application processor (AP), etc., or pure graphics processing units, such as a graphics processing unit (GPU), a visual processing unit (VPU), and / or an AI-specific processor, such as a neural processing unit (NPU).

[0180] The one or more processors control processing of input data according to predefined operating rules or artificial intelligence (AI) models stored in non-volatile memory and volatile memory. The predefined operating rules or artificial intelligence models are provided by training or learning.

[0181] Here, providing by learning means obtaining predefined operating rules or an AI model with desired characteristics by applying a learning algorithm to a plurality of learning data. This learning can be performed in the device itself in which the AI ​​according to the embodiment is executed, and / or can be implemented by a separate server / system.

[0182] The AI ​​model can include multiple neural network layers. Each layer has multiple weight values, and the calculation of one layer is performed by combining the calculation results of the previous layer and the multiple weights of the current layer. Examples of neural networks include, but are not limited to, convolutional neural networks (CNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), bidirectional recurrent deep neural networks (BRDNNs), generative adversarial networks (GANs), and deep Q networks.

[0183] A learning algorithm is a method for training a predetermined target device (e.g., a robot) using multiple learning data to enable, allow, or control the target device to make a determination or prediction. Examples of such learning algorithms include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0184] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific method implemented when the computer-readable medium described above is executed by an electronic device can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0185] The technical solution provided in the embodiment of the present application can be applicable to a variety of systems, especially 5G systems. For example, the applicable system can be a global system of mobile communication (GSM) system, a code division multiple access (CDMA) system, a wideband code division multiple access (WCDMA) general packet radio service (GPRS) system, a long term evolution (LTE) system, a LTE frequency division duplex (FDD) system, a LTE time division duplex (TDD) system, an advanced long term evolution (LTE-A) system, a universal mobile telecommunication system (UMTS), a world-wide interoperability for microwave access (WiMAX) system, a 5G new air interface (NR) system, etc. These various systems include terminal equipment and network equipment. The system may also include a core network part, such as an evolved packet system (EPS), a 5G system (5GS), etc.

[0186] The terminal device involved in the embodiments of the present application may be a device that provides voice and / or data connectivity to a user, a handheld device with wireless connection function, or other processing devices connected to a wireless modem. In different systems, the name of the terminal device may also be different. For example, in a 5G system, the terminal device may be called a user equipment (UE). A wireless terminal device can communicate with one or more core networks (CN) via a radio access network (RAN). The wireless terminal device can be a mobile terminal device, such as a mobile phone (or "cellular" phone) and a computer with a mobile terminal device. For example, it can be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device that exchanges language and / or data with a radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, an access point, a remote terminal device, an access terminal device, a user terminal device, a user agent, or a user device, but is not limited in the embodiments of the present application.

[0187] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0188] The above are only some of the implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A predistortion processing method, characterized in that: include: Acquire a first continuous time series signal including a time series signal to be processed; Normalizing the first continuous time series signal to obtain a corresponding second continuous time series signal; The second continuous time series signal is input into a pre-trained back-propagation BP neural network model, and a pre-distorted time series signal corresponding to the time series signal to be processed is output, wherein the BP neural network model is an inverse behavioral model of the target power amplifier PA used to power amplify the time series signal to be processed.

2. The method according to claim 1, characterized in that The first continuous timing signals include a first preset number of timing signals before the timing signal to be processed and a second preset number of timing signals after the timing signal to be processed.

3. The method according to claim 1, characterized in that The normalizing the first continuous time series signal to obtain a corresponding second continuous time series signal includes: Obtaining a training data set, and obtaining a mean and a standard deviation of time series data samples in the training data set, wherein the training data set includes a third preset number of PA input time series signal samples and corresponding PA output time series signal samples of the target PA; Based on the mean and the standard deviation, obtaining a normalized time series signal corresponding to each time series signal in the first continuous time series signal; Based on each normalized time series signal, the second continuous time series signal is acquired.

4. The method according to claim 1, wherein The BP neural network includes an input layer, a third preset number of hidden layers, and an output layer. Inputting the second continuous time series signal into a pre-trained back-propagation BP neural network model and outputting a pre-distorted time series signal corresponding to the time series signal to be processed includes: receiving the second continuous time series signal through the input layer, and converting the second continuous time series signal into an initial feature vector; Obtaining a corresponding abstract feature vector based on the initial feature vector through the hidden layer; The pre-distorted time series signal is outputted through the output layer based on the abstract feature vector.

5. The method according to claim 3, characterized in that The training process of the BP neural network model is as follows: The BP neural network model is trained using the training data set to obtain the pre-trained BP neural network model.

6. The method according to claim 5, characterized in that The step of training the BP neural network model using the training data set to obtain the pre-trained BP neural network model includes: Input each PA output timing signal sample into the BP neural network model, and output the corresponding predicted PA input timing signal; Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding PA input timing signal sample; Based on each mean square error, the network parameters of the BP neural network model are updated using a back propagation algorithm to obtain the pre-trained BP neural network model.

7. The method according to claim 1, characterized in that The method further comprises: After obtaining pre-distorted timing signals corresponding to a fourth preset number of timing signals to be processed, inputting each pre-distorted timing signal into the target PA to obtain at least one corresponding PA output timing signal; Based on each pre-distorted timing signal and the corresponding PA output timing signal, the network parameters of the BP neural network model are updated again.

8. The method according to claim 7, characterized in that The updating of the network parameters of the BP neural network based on each pre-distorted timing signal and the corresponding PA output timing signal comprises: Input each PA output signal into the BP neural network model, and output the corresponding predicted PA input timing signal; Obtaining a corresponding mean square error based on each predicted PA input timing signal and a corresponding predistorted timing signal; The network parameters of the BP neural network model are updated based on the mean square errors.

9. A predistortion processing device, characterized in that: include: A timing signal acquisition module, configured to acquire a first continuous timing signal including a timing signal to be processed; a normalization module, configured to perform normalization processing on the first continuous time series signal to obtain a corresponding second continuous time series signal; A predistortion processing module is used to input the second continuous time series signal into a pre-trained back-propagation BP neural network model, and output a predistorted time series signal corresponding to the time series signal to be processed, wherein the BP neural network model is an inverse behavioral model of the target power amplifier device PA used to power amplify the time series signal to be processed.

10. An electronic device, characterized in that: including memory and processor; The memory stores a computer program; The processor is configured to execute the computer program to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images