Method, device and equipment for generating nonlinear system model

Abstract

The invention provides a method, a device and equipment for generating a nonlinear system model. The method comprises the following steps: establishing a piecewise linear nonlinear system model; obtaining an input signal compression matrix according to a corresponding relation between the input signal sequence of the nonlinear system and a preset LUT; and determining a model coefficient of the nonlinear system model according to the input signal compression matrix and the output signal matrix of the nonlinear system. According to the embodiment of the invention, the parameter calculation efficiency of the nonlinear system model is effectively improved, and the nonlinear system model is efficiently and accurately obtained.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, and device for generating nonlinear system models. Background Technology

[0002] Nonlinear systems in the field of communication electronics, such as power amplifiers, analog amplifiers, mixers, and modems, can generate strong nonlinear effects, leading to signal distortion, harmonics, and intermodulation products, which in turn degrade system performance and affect system stability.

[0003] In related technologies, the mainstream model used for nonlinear system models is the memory polynomial model, which has a better fitting effect. However, in actual implementation, there is the problem of high-order polynomial calculation, which consumes a lot of computing resources. Summary of the Invention

[0004] This invention provides a method, apparatus, and device for generating nonlinear system models, which effectively improves the speed of nonlinear system model parameter calculation and obtains nonlinear system models efficiently and accurately.

[0005] This invention provides a method for generating a nonlinear system model, comprising the following steps.

[0006] Establish a piecewise linear nonlinear system model; The input signal compression matrix is ​​obtained based on the correspondence between the input signal sequence of the nonlinear system and the preset lookup table (LUT). The model coefficients of the nonlinear system model are determined based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0007] According to a method for generating a nonlinear system model provided by the present invention, the nonlinear system model includes:

[0008] in, This represents the output signal of a nonlinear system. These represent the model coefficients of a nonlinear system model; Indicates time delay Input signals per unit; This represents a piecewise function; M represents the preset number of segments; This represents the signal input to the nonlinear system at time point n; Indicates time delay Signal of one unit; This represents the signal input to the nonlinear system at time point n-1; This represents the signal input to the nonlinear system at time point n+1. N is the length of the input and output signals.

[0009] According to a method for generating a nonlinear system model provided by the present invention, the step of obtaining an input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and a preset LUT includes: Obtain the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The short matrix of the input signal is determined based on the matrices corresponding to the multiple target LUT elements; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

[0010] According to a method for generating a nonlinear system model provided by the present invention, the step of compressing the short input signal matrix to obtain a compressed input signal matrix includes: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of each target element is not equal to 0.

[0011] According to a method for generating a nonlinear system model provided by the present invention, the step of determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system includes: Based on the input signal compression matrix, determine the short input signal matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal line in the LUT table. The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the multiple target LUT elements and the output signal matrix of the nonlinear system.

[0012] According to a method for generating a nonlinear system model provided by the present invention, the step of determining the model coefficients of the nonlinear system model based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system includes: A first target value is determined based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. Based on the matrices corresponding to the multiple target LUT elements, a second target value is determined; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first target value and the second target value.

[0013] According to a method for generating a nonlinear system model provided by the present invention, determining a first target value based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system includes: Based on the matrices corresponding to the multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

[0014] According to a method for generating a nonlinear system model provided by the present invention, the step of determining the matrix corresponding to each LUT element in the LUT table based on the matrices corresponding to the plurality of target LUT elements includes: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0015] According to a method for generating a nonlinear system model provided by the present invention, the step of offsetting the matrix corresponding to the target LUT element based on the position of each LUT element in the LUT table to determine the matrix corresponding to each LUT element in the LUT table includes: Based on the position of each LUT element in the LUT table, determine the offset of each LUT element relative to the target LUT element; Based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0016] According to a method for generating a nonlinear system model provided by the present invention, determining the offset of each LUT element relative to a target LUT element based on the position of each LUT element in the LUT table includes: Determine the x-coordinate value of the LUT element in the LUT table and the y-coordinate value of the LUT element in the LUT table; The smaller of the horizontal and vertical coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

[0017] According to a method for generating a nonlinear system model provided by the present invention, the step of determining a second target value based on the matrix corresponding to the plurality of target LUT elements includes: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

[0018] According to a method for generating a nonlinear system model provided by the present invention, determining the second target value based on the matrix corresponding to each LUT element in the LUT table includes: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element in the LUT table whose order is greater than or equal to the matrix corresponding to the first LUT element. The product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by multiplying the matrix corresponding to the first LUT element by the conjugate transpose of the matrix corresponding to the second LUT element, the matrix corresponding to the third LUT element by the conjugate transpose of the matrix corresponding to the third LUT element, and the matrix corresponding to the fourth LUT element.

[0019] According to a method for generating a nonlinear system model provided by the present invention, the step of reusing the conjugate transpose of the matrix corresponding to the first LUT element and the product of the matrix corresponding to the second LUT element to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element includes: When the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0020] The present invention also provides an apparatus for generating a nonlinear system model, comprising the following modules: Establish a module for building piecewise linear nonlinear system models; The compression module is used to obtain the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT; The generation module is used to determine the model coefficients of the nonlinear system model based on the input signal compression matrix.

[0021] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the compression module is specifically used for: Obtain the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The short matrix of the input signal is determined based on the matrices corresponding to the multiple target LUT elements; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

[0022] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the compression module is specifically used for: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of each target element is not equal to 0.

[0023] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Based on the input signal compression matrix, determine the short input signal matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal line in the LUT table. The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the multiple target LUT elements and the output signal matrix of the nonlinear system.

[0024] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: A first target value is determined based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. Based on the matrices corresponding to the multiple target LUT elements, a second target value is determined; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first target value and the second target value.

[0025] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Based on the matrices corresponding to the multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

[0026] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0027] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Based on the position of each LUT element in the LUT table, determine the offset of each LUT element relative to the target LUT element; Based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0028] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Determine the x-coordinate value of the LUT element in the LUT table and the y-coordinate value of the LUT element in the LUT table; The smaller of the horizontal and vertical coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

[0029] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

[0030] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element in the LUT table whose order is greater than or equal to the matrix corresponding to the first LUT element. The product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by multiplying the matrix corresponding to the first LUT element by the conjugate transpose of the matrix corresponding to the second LUT element, the matrix corresponding to the third LUT element by the conjugate transpose of the matrix corresponding to the third LUT element, and the matrix corresponding to the fourth LUT element.

[0031] According to the present invention, a nonlinear system model generation apparatus is provided, wherein the generation module is specifically used for: When the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0032] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for generating a nonlinear system model as described above.

[0033] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for generating a nonlinear system model as described above.

[0034] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for generating a nonlinear system model as described above.

[0035] The method, apparatus, and device for generating nonlinear system models provided by this invention, on the one hand, effectively avoid the calculation of higher-order terms by establishing a piecewise linear nonlinear system model, and the solution matrix of the piecewise linear nonlinear system model itself has a certain sparsity, thereby accelerating the calculation in the specific solution process. On the other hand, by compressing and folding the information of the matrix, unnecessary steps in the calculation process are omitted, reducing calculation time, saving memory space, and effectively improving the speed of parameter calculation, thus enabling the efficient and accurate generation of the nonlinear system model. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is one of the flowcharts illustrating the method for generating nonlinear system models provided by this invention.

[0038] Figure 2 This is the second flowchart illustrating the method for generating nonlinear system models provided by this invention.

[0039] Figure 3 This is a schematic diagram of the lookup table provided by the present invention.

[0040] Figure 4 This is a schematic diagram of the LUT table layout provided by the present invention.

[0041] Figure 5 This is a schematic diagram of the LUT table and input signal matrix provided by the present invention.

[0042] Figure 6 This is a schematic diagram illustrating the generation pattern of the input signal matrix provided by the present invention.

[0043] Figure 7 This is a schematic diagram of the structure corresponding to the second target value calculation process provided by the present invention.

[0044] Figure 8 This is a schematic diagram of sparse matrix calculation provided by the present invention.

[0045] Figure 9 This is a schematic diagram of the structure of the nonlinear system model generation device provided by the present invention.

[0046] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0048] The following is combined Figures 1-10 The present invention describes a method, apparatus, and device for generating nonlinear system models.

[0049] To facilitate a clearer understanding of the technical solutions of the various embodiments of this application, some technical content related to the various embodiments of this application will be introduced first.

[0050] Nonlinear systems in the field of communication electronics, such as power amplifiers, analog amplifiers, mixers, modems, and fiber optic transmission lines, can generate strong nonlinear effects, leading to signal distortion, harmonics, and intermodulation products, which in turn degrade system performance and affect system stability.

[0051] As a key component of a communication system, the linearity of the wireless communication transmitter directly affects the performance of the entire system. Among these components, the power amplifier (PA) is the most representative, and the nonlinearity of other components can be incorporated into the nonlinear distortion of the power amplifier.

[0052] Digital pre-distortion (DPD) technology preprocesses the transmitted signal in the digital domain before it undergoes nonlinear distortion introduced by the power amplifier, producing an "inverse" distortion with the same amplitude but opposite phase as the nonlinear distortion introduced by the power amplifier. The preprocessed transmitted signal can cancel out the nonlinear distortion introduced by the transmitter. As a nonlinear compensation module for power amplifiers, digital pre-distortion technology is widely used in industrial applications.

[0053] It should be noted that the digital predistortion model is the inverse model of nonlinear system models such as power amplifiers, and this inverse model is also nonlinear, belonging to the category of nonlinear system models. Therefore, the generation process of the nonlinear system model based on this application can not only realize the generation of nonlinear system models such as power amplifiers, but also realize the generation of nonlinear system models corresponding to digital predistortion.

[0054] For example, the process of generating a nonlinear system model is as follows: Figure 1 As shown: First, the system input and output signals need to be acquired. This technology is dedicated to constructing a model of a nonlinear system. It requires acquiring the system's input and output signals, and also requires pre-setting the LUT table structure according to the complexity of the nonlinear system (the LUT table plays an "index" role in the entire nonlinear system modeling and calculation, guiding the entire calculation to be carried out efficiently and orderly) in order to construct an accurate nonlinear system model.

[0055] Second, based on the constructed model, a coefficient solution architecture is built using the least squares method. This solution architecture is a classic least squares architecture and needs to include input and output signals.

[0056] Third, this step requires a large amount of memory and computation time, placing very high demands on the hardware. How to effectively reduce the running time and memory usage is a key issue that needs to be addressed.

[0057] Fourth, calculate the model coefficients based on the obtained intermediate variables.

[0058] With the application and popularization of next-generation mobile communication technologies, and based on the urgent needs of practical engineering, a large number of models have emerged in the field of nonlinear model system modeling. Among them, the Volterra series model is a general-purpose memory-based nonlinear model, often used to describe the characteristics of nonlinear systems. However, if a complete series model is used, the system complexity will be very high, making it difficult to meet the real-time computing requirements of high-speed signal processing systems. To reduce model complexity, various pruning techniques can be applied to the Volterra model; therefore, most models are derived from the Volterra model. In existing technologies, the commonly used nonlinear model for digital predistortion is the General Memory Polynomial (GMP) model, which has a better fitting effect. However, in practical implementation, there is the problem of high-order polynomial calculation, which consumes a lot of computing resources. At the same time, in practical applications, as bandwidth increases, the nonlinear system model of power amplifiers becomes increasingly complex. In some applications, due to resource constraints or uneven data distribution, singular matrices may appear in the coefficient solution process of the nonlinear system model, causing the overall predistortion processing flow to be interrupted, resulting in abnormal iteration states of the nonlinear system model and affecting the stability and reliability of the nonlinear system model.

[0059] For nonlinear system models, the form of memory polynomials is generally used for description:

[0060] in, For the depth of signal memory, The memory depth of the signal magnitude. Represents the nonlinear order. This represents the maximum value of the signal memory depth. This represents the maximum value of the memory depth of the signal amplitude. The maximum value of the nonlinear order. The coefficients of the polynomial, The output signal after passing through the nonlinear model. For the input signal of a nonlinear system, It indicates the time when a discrete signal passes through a nonlinear system.

[0061] As the size of such models increases, the difficulty of solving the entire model also increases, and it is necessary to calculate higher-order terms multiple times.

[0062] A key part of model parameter extraction is the selection of an adaptive algorithm. An adaptive algorithm uses the error between the measured signal and the ideal signal to guide the updating of model coefficients. After multiple iterations, it achieves consistency between the measured and ideal signals. It's called an adaptive algorithm because the input-output characteristic curve of a nonlinear system can be affected by external environmental factors. However, the algorithm itself adjusts the model coefficients according to these curve changes, ensuring the fitted result best reflects the current nonlinear characteristics of the system. Using the input and error signals to update the undetermined coefficients provides strong robustness. In practical applications, digital predistortion techniques for nonlinear system parameter extraction must employ adaptive algorithms. Currently, the most commonly used adaptive algorithm is the least squares (LS) algorithm. While it offers better convergence time and avoids steady-state error problems, its implementation requires multiple matrix multiplications, consuming excessive hardware resources and impacting the computational efficiency of digital predistortion. The least squares algorithm is described below: The matrix relationship between the input and output signals of a nonlinear system and the coefficients of the nonlinear system model is known to be:

[0063] Where the matrix For the output signal matrix, The input signal matrix, Let be the coefficient matrix. Solving the above equation using the matrix method yields:

[0064] in, This represents the conjugate transpose of the input signal matrix; however, in engineering implementation, the computational complexity of the matrix increases exponentially with the increase in the matrix size.

[0065] Figure 2 This is one of the flowcharts illustrating the method for generating a nonlinear system model provided by the present invention. The method includes the following: Step 201: Establish a piecewise linear nonlinear system model.

[0066] Specifically, the mainstream model used in existing nonlinear system models is the memory polynomial model, which has a better fitting effect. However, in actual implementation, there is the problem of high-order polynomial calculation, which consumes a lot of computing resources.

[0067] To address the aforementioned issues, this application first establishes a piecewise linear nonlinear system model, replacing the traditional memory polynomial model. Compared to the traditional memory polynomial model, the piecewise linear nonlinear system model has a simpler and easier-to-order expression, and it is also compatible with subsequent least squares matrix calculations, effectively avoiding the operation of higher-order terms. Furthermore, the solution matrix of the piecewise linear nonlinear system model itself has a certain degree of sparsity, which can accelerate computation during the specific solution process.

[0068] Step 202: Obtain the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset lookup table (LUT).

[0069] Specifically, in this embodiment, based on the establishment of a piecewise linear nonlinear system model, an input signal compression matrix is ​​obtained according to the correspondence between the input signal sequence of the nonlinear system and the preset LUT. This allows for the omission of unnecessary steps in the calculation process by compressing the information of the folding matrix, reducing computation time, saving memory space, and effectively improving the speed of parameter updates. Optionally, in this embodiment, the input signal sequence refers to the signals {x(1), x(2), ..., x(N)} input to the nonlinear system at different times, where N is the length of the signal.

[0070] For example, according to the least squares method, the coefficient matrix expression is:

[0071] Let the input signal matrix of the model be represented in one form as follows:

[0072] in, This variable represents The column vectors of the matrix, and

[0073] in, It's a delay. It is the row number of the matrix. It is based on The default spline function, Its magnitude is determined by the amplitude of the signal, therefore its magnitude is The non-zero valid values ​​are... The matrix is ​​sparse because all other elements are 0.

[0074] like Figure 3As shown, the traditional calculation method uses the look-up table (LUT) method, which is the most widely used and simplest nonlinear behavior model in the field of digital predistortion. In essence, the look-up table structure is actually a simple piecewise linear function that divides the entire input range into several sub-units and achieves the final model fitting by linearly weighting each sub-unit. The look-up table model is both a mathematical model and a way of implementation.

[0075] In this application, the LUT table generally needs to be pre-defined based on the complexity of the nonlinear system. The LUT table mainly stores the principal terms of the entire nonlinear model. Each LUT element in the LUT table corresponds to one term, that is, each LUT element in the LUT table corresponds to a part of the input matrix. The LUT table determines the layout rules and size of the entire model, and it plays a guiding role in model implementation and parameter calculation. Figure 4 This represents a LUT arrangement, where This represents the signal delay ( ), The time delay representing the signal amplitude (represented in the GMP model) In the Spline model, it represents ).

[0076] This LUT table contains a total of 54 LUT elements, each of which has an important mapping meaning in the process of solving the nonlinear model. For the input matrix... In other words, each LUT element corresponds to A part of the matrix , matrix The order of the elements is also determined by the order of the LUT elements. The generated coefficients also need to be downloaded to the hardware according to the order of the LUT elements to optimize the input signal to the nonlinear system and reduce signal distortion. In practical engineering, the size of the LUT table can be flexibly changed according to the complexity of the nonlinear system, such as... Figure 5 As shown, this represents the LUT table and The correspondence and size of matrices, in practice The size of the matrix is Where 54 represents the number of LUT elements in the LUT table; This indicates the row number of the matrix corresponding to the LUT element in the LUT table; N represents the length of the input signal.

[0077] In this embodiment, the input signal matrix obtained based on the LUT table and the input signal sequence is compressed to obtain the input signal compression matrix. This allows for the omission of unnecessary steps in the calculation process by compressing the information of the folded matrix, thereby reducing calculation time, saving memory space, and effectively improving the speed of parameter updates.

[0078] Step 203: Determine the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0079] Specifically, after establishing a nonlinear system model and obtaining the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT, the coefficients of the nonlinear system model can be solved based on the nonlinear system model, the input signal compression matrix, and the output signal matrix of the nonlinear system. This effectively avoids the calculation of higher-order terms. Moreover, by compressing and folding the information of the matrix, unnecessary steps in the calculation process are omitted, reducing computation time, saving memory space, and effectively improving the speed of parameter calculation. Thus, the nonlinear system model can be obtained efficiently and accurately. Optionally, in this embodiment, the output signal matrix is ​​directly determined by the output signal sequence of the nonlinear system ({y(1),y(2),.....,y(N)}, where N is the length of the signal) and corresponds one-to-one with the input signal sequence.

[0080] The method described in the above embodiments, on the one hand, effectively avoids the calculation of higher-order terms by establishing a piecewise linear nonlinear system model, and the solution matrix of the nonlinear system model itself has a certain sparsity, thereby accelerating the calculation in the specific solution process. On the other hand, by compressing and folding the information of the matrix, unnecessary steps in the calculation process are omitted, reducing calculation time, saving memory space, and effectively improving the speed of parameter calculation, thus obtaining the nonlinear system model efficiently and accurately.

[0081] In some embodiments, the piecewise linear nonlinear model includes:

[0082] in, This represents the output signal of a nonlinear system. These represent the model coefficients of a nonlinear system model; Indicates time delay Input signals per unit; This represents a piecewise function; M represents the preset number of segments; This represents the signal input to the nonlinear system at time point n; Indicates time delay Signal of one unit; This represents the signal input to the nonlinear system at time point n-1; This represents the signal input to the nonlinear system at time point n+1. N is the length of the input and output signals.

[0083] Specifically, the piecewise linear nonlinear system model established in this application embodiment is simple to construct, easy to use, and has accurate fitting. It can fit complex shapes in curve design and effectively avoids the calculation of higher-order terms.

[0084] For example, the classic memoization polynomial model can be rewritten as:

[0085] in, This represents the output signal of a nonlinear system. These represent the model coefficients of a nonlinear system model; Indicates time delay Input signals per unit; This represents a piecewise function; M represents the preset number of segments; This represents the signal input to the nonlinear system at time point n; Indicates time delay Signal of one unit; This represents the signal input to the nonlinear system at time point n-1; This represents the signal input to the nonlinear system at time point n+1. N is the length of the input and output signals.

[0086] It should be noted that the nonlinear system model established in the embodiments of this application is... In this case, The value of is 0, which means that the nonlinear system model in this embodiment has a certain sparsity, thereby accelerating the calculation and improving the speed of parameter calculation in the specific solution process, so as to obtain the nonlinear system model efficiently and accurately.

[0087] The method described above establishes a piecewise linear nonlinear system model that is simple to construct, easy to use, and accurately fitted. It can also approximate complex shapes in curve design, effectively avoiding the calculation of higher-order terms. Moreover, the established nonlinear system model itself has a certain degree of sparsity, which can accelerate the calculation and improve the speed of parameter calculation in the specific solution process, thus obtaining the nonlinear system model efficiently and accurately.

[0088] In some embodiments, the input signal compression matrix is ​​obtained based on the correspondence between the input signal sequence of the nonlinear system and a preset LUT, including: Based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT, obtain the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; Determine the short matrix of the input signal based on the matrices corresponding to the elements of the multiple target LUTs; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

[0089] Specifically, the nonlinear system model is written in matrix form and solved using the least squares method:

[0090] in,

[0091] in, The input signal matrix of the piecewise model, This is the coefficient matrix of the piecewise model. The output signal matrix is ​​given by the piecewise model; once the coefficients are determined, the polynomial expression for the piecewise linear nonlinear system model can be obtained as follows:

[0092] according to As can be seen from the generation pattern, the input signal matrix U generated by the piecewise model has strong sparsity. Current methods for calculating the input signal matrix U employ a loop-based approach, iterating through each matrix element, and subsequent matrix multiplication operations also use traditional multiplication. This traditional method involves traversing the matrices corresponding to all LUT elements to generate the input signal matrix U, which wastes a significant amount of memory and runtime.

[0093] This application introduces a short input signal matrix and a compressed input signal matrix to represent the input signal matrix U, which greatly reduces memory space, reduces the difficulty of troubleshooting problems later, and improves the computational efficiency of model coefficients and the generation efficiency of nonlinear system models.

[0094] Optionally, in this embodiment, the matrix corresponding to multiple target LUT elements in the LUT table is first obtained; wherein the target LUT elements are located at the first and second positions of each diagonal in the LUT table; then, the short matrix of the input signal is determined based on the matrix corresponding to the multiple target LUT elements, as follows: The coefficient matrix can be expressed as follows using the least squares method:

[0095] in, The matrix is ​​based on Figure 4 The LUT table in the code determines that, in the traditional approach, it is necessary to traverse the matrix corresponding to each LUT element to generate the corresponding LUT. Using matrices would waste a lot of computation time and space.

[0096] This application is combined with Based on the relationship with LUT tables and the sparsity inherent in piecewise models, a method for compressing the input signal matrix is ​​proposed. In this compression method, the input signal matrix is ​​recorded... The key information required ensures the reliability of subsequent coefficient calculations.

[0097] As shown in Figure 6, the matrix corresponding to the first element of the diagonal LUT in the LUT table is... , , , and ,and The matrix corresponding to the LUT elements on the same diagonal is , , , As shown in the diagram, the matrices corresponding to the elements on the diagonal can all be derived from... The matrix can be obtained by removing or adding a small number of elements from the beginning and end of the matrix.

[0098] For example, to obtain the layout structure of the LUT table, iterate through the matrix corresponding to each LUT element, and use the first element of the diagonal in the LUT table. , , , and calculate The matrix is ​​such that the matrix corresponding to the LUT element at the first off-diagonal element position is not calculated, and the final result is... , , , , Five matrices, which will be used in this application The matrix is ​​called the short input signal matrix. This short input signal matrix contains the entire input signal matrix. The information, but it is different from the original input signal matrix. The space has been reduced by 11 times, and the memory space utilization will be even higher as the size of the LUT table continues to increase.

[0099] Optionally, after compressing the input signal matrix and determining the short input signal matrix based on the matrix corresponding to the first element of each diagonal in the LUT table, the embodiments of this application can further compress the short input signal matrix, thereby more effectively saving memory space and running time, and improving the calculation efficiency of model coefficients and the generation efficiency of nonlinear system models.

[0100] The methods described in the above embodiments, traditionally, require traversing all the matrices corresponding to the LUT elements to generate the input signal matrix U, which wastes a significant amount of memory space and runtime. This application, however, combines the relationship between the input signal matrix U and the LUT table with the sparsity unique to piecewise models. It uses the first element of each diagonal in the LUT table to determine a short input signal matrix. This short matrix records the key information required for the input signal matrix U, ensuring the reliability of subsequent coefficient calculations. Simultaneously, it significantly reduces memory space compared to the original input signal matrix U, simplifies later troubleshooting, and improves the efficiency of model coefficient calculation and model generation for nonlinear systems.

[0101] In some embodiments, the short input signal matrix is ​​compressed to obtain a compressed input signal matrix, including: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of the target element is not equal to 0.

[0102] Specifically, after determining the short input signal matrix based on the matrix corresponding to the first element of each diagonal in the LUT table, this embodiment further obtains the input signal compression matrix based on each target element and its position in the short input signal matrix, where the value of each target element is not equal to 0. The specific process is as follows: Suppose the expression for the elements of the short matrix of the input signal corresponding to the LUT table is as follows:

[0103] Then, based on each target element (i.e., non-zero element) in the short matrix of the input signal... By combining the input signal and the positions of each target element, we can obtain the input signal compression matrix:

[0104] The input signal matrix The matrix corresponds to the input signal compression matrix, which is due to the piecewise model. Each column of the matrix contains two data points; the last row is the offset matrix, which represents the position of the first data point in the matrix. The position of the matrix, and the position of the second data point, are incremented by one. This is due to the piecewise model. Two data positions in a matrix are adjacent.

[0105] In the method described above, after determining the short input signal matrix based on the matrix corresponding to the first element of each diagonal in the LUT table, this embodiment further compresses the input signal based on each target element and its position to obtain a compressed input signal matrix. This more effectively reduces memory space and improves the computational efficiency of the model coefficients and the generation efficiency of the nonlinear system model.

[0106] In some embodiments, determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system includes: Determine the short matrix of the input signal based on the input signal compression matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the elements of multiple target LUTs and the output signal matrix of the nonlinear system.

[0107] Specifically, after compressing the input signal matrix U to obtain the compressed input signal matrix, the compressed input signal matrix can be stored, effectively reducing memory space. Optionally, during the calculation of model coefficients for the nonlinear system model, the stored compressed input signal matrix can be used to determine the short input signal matrix, and subsequent calculations do not require saving the short input signal matrix, effectively reducing memory space. Then, by shifting the matrix corresponding to the target LUT element of the short input signal matrix, the matrices corresponding to all LUT elements in the LUT table and the input signal matrix can be quickly and accurately determined without recalculating and generating the input signal matrix U, thereby effectively improving the calculation efficiency of model coefficients for the nonlinear system model.

[0108] The method described in the above embodiments, taking into account the characteristics of nonlinear system models, effectively replaces the input signal matrix with an input signal compression matrix. Compared to traditional methods, it saves storage space for the input signal matrix. The memory space occupied and the calculation of the input signal matrix The time consumed by the matrix can be reduced, thus effectively improving the computational efficiency of model coefficients in nonlinear system models.

[0109] In some embodiments, determining the model coefficients of a piecewise linear nonlinear system model based on the matrix corresponding to multiple target LUT elements includes: The first target value is determined based on the matrices corresponding to multiple target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. The second target value is determined based on the matrix corresponding to the elements of the multiple target LUTs; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first and second objective values.

[0110] Specifically, the least squares method reveals the model coefficients of the nonlinear system model. The expression is:

[0111] That is, the model coefficients of a nonlinear system model. The calculation mainly includes the first target value Overall calculation and second target value In this embodiment of the application, the first target value is determined by calculating the matrix corresponding to multiple target LUT elements. Second target value Then, the model coefficients of the nonlinear system model can be determined. .

[0112] The method in the above embodiments decomposes the calculation of the model coefficients of the nonlinear system model into a first objective value. Calculation and second target value The calculation can be performed using different acceleration methods based on the characteristics of the first and second target values, thereby effectively improving the calculation efficiency of the model coefficients of the nonlinear system model.

[0113] In some embodiments, determining a first target value based on a plurality of target LUT elements includes: Based on the matrices corresponding to multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

[0114] Specifically, the matrix corresponding to the target LUT element is the matrix corresponding to the first element of the diagonal in the LUT table. By offsetting the matrix corresponding to the target LUT element, the matrix corresponding to each LUT element in the LUT table can be determined. Then, based on the matrix corresponding to each LUT element in the LUT table, the input signal matrix U can be determined. Finally, multiplying the conjugate transpose of the input signal matrix U with the output matrix Y of the nonlinear system allows for the rapid and accurate determination of the first target value. .

[0115] The method in the above embodiments calculates the first target value through the input signal compression matrix. Based on sparse matrix acceleration, multiplication calculation is accelerated, saving memory space and greatly reducing matrix calculation time, effectively improving the calculation efficiency of model coefficients of nonlinear system models.

[0116] In some embodiments, determining the matrix corresponding to each LUT element in the LUT table based on the matrices corresponding to multiple target LUT elements includes: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0117] Specifically, since the positions of each LUT element in the LUT table are different, it is necessary to offset the matrix corresponding to the first element of the diagonal in the LUT table differently to accurately determine the matrix corresponding to each LUT element. In other words, to determine the matrix corresponding to each LUT element in the LUT table in this application, it is first necessary to determine the offset of each LUT element relative to the target LUT element based on its position in the LUT table; then, based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset, thus quickly and accurately determining the matrix corresponding to each LUT element in the LUT table.

[0118] In some embodiments, determining the offset of each LUT element relative to the target LUT element based on the position of each LUT element in the LUT table includes: Determine the x-coordinate and y-coordinate values ​​of the LUT element in the LUT table; The smaller of the x-coordinate and y-coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

[0119] Specifically, in this embodiment, the horizontal coordinate value and the vertical coordinate value of the LUT element in the LUT table are first determined. Then, the smaller value between the horizontal and vertical coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element. Thus, each element in the LUT table can be quickly and accurately determined based on the determined offset.

[0120] For example, such as Figure 6 As shown, the x and y coordinates of LUT element 6 in the LUT are 2 and 1, respectively. The smaller of the x and y coordinates of LUT element 6 is 1. Therefore, by shifting the matrix corresponding to LUT element 2, the first element in the LUT table, by one unit, the matrix corresponding to LUT element 6 can be determined. By analogy, the matrices corresponding to all LUT elements in the LUT can be determined quickly and accurately, thereby effectively improving the calculation efficiency of the model coefficients of the nonlinear system model.

[0121] The method described in the above embodiment determines the offset of each LUT element relative to the target LUT element based on the position of each LUT element in the LUT table. Then, based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset, which can quickly and accurately determine the matrix corresponding to each LUT element in the LUT table and effectively improve the calculation efficiency of model coefficients of nonlinear system models.

[0122] In some embodiments, determining a second target value based on a plurality of target LUT elements includes: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

[0123] Specifically, the matrix corresponding to the target LUT element is the matrix corresponding to the first element of the diagonal in the LUT table. By offsetting the matrix corresponding to the target LUT element, the matrix corresponding to each LUT element in the LUT table can be determined. Then, the second target value can be determined based on the matrix corresponding to each LUT element in the LUT table.

[0124] The method in the above embodiments calculates the second target value through the input signal compression matrix. Based on sparse matrix acceleration, multiplication calculation is accelerated, saving memory space and greatly reducing matrix calculation time, effectively improving the calculation efficiency of model coefficients of nonlinear system models.

[0125] In some embodiments, determining the second target value based on the matrix corresponding to each LUT element in the LUT table includes: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element whose order in the LUT table is greater than or equal to the matrix corresponding to the first LUT element. The product of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by the product of the conjugate transpose of the matrix corresponding to the first LUT element, the product of the matrix corresponding to the second LUT element, the product of the conjugate transpose of the matrix corresponding to the third LUT element, and the product of the matrix corresponding to the fourth LUT element.

[0126] Specifically, calculate the second target value. The specific process is as follows: (1) Initialization Matrix (noted as) ) is a matrix consisting entirely of zeros.

[0127]

[0128] (2) Construct the compression matrix.

[0129] (3) Calculation A matrix, specifically expressed as:

[0130] like Figure 7 As shown, in order to reduce the amount of computation and improve the computational efficiency of the second target value in this embodiment of the application, the second target value can be... The calculation process is simplified to Figure 7 For ease of representation, we can... Equivalent to .

[0131] Optionally, during the reuse process, firstly, the matrix corresponding to the target LUT element is used as the matrix corresponding to the first LUT element, that is, the matrix corresponding to the first element of each diagonal in the LUT table is used as the matrix corresponding to the first LUT element. Then, the matrices corresponding to the LUT elements in the LUT table whose order is greater than or equal to the matrix corresponding to the first LUT element are used as the matrices corresponding to the second LUT elements. Finally, the product of the matrices corresponding to the first LUT element and the matrices corresponding to the second LUT element is determined, that is, the product of the matrices corresponding to the first elements of each diagonal in the LUT table (the matrices corresponding to LUT element 1, LUT element 2, LUT element 3, LUT element 4, and LUT element 8) with the matrices corresponding to each LUT element (such as the product of matrix 1 corresponding to LUT element 1 and the matrix corresponding to LUT element 2, and the product of matrix 8 corresponding to LUT element 9). Then, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the product of the matrix corresponding to the second LUT element is reused to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0132] In some embodiments, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element, including: If the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0133] Specifically, when the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0134] For example, if the matrix corresponding to the third LUT element is the matrix corresponding to LUT element 2 in the LUT table, and the matrix corresponding to the fourth LUT element is the matrix corresponding to LUT element 1 in the LUT table, then the product of the matrix corresponding to LUT element 2 and the matrix corresponding to LUT element 1 can reuse the product of the matrix corresponding to LUT element 1 and the matrix corresponding to LUT element 2. That is, reuse the conjugate transpose of the matrix corresponding to the first LUT element and the product of the matrix corresponding to the second LUT element (the product of the matrix corresponding to LUT element 1 and the matrix corresponding to LUT element 2) to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element (the product of the matrix corresponding to LUT element 2 and the matrix corresponding to LUT element 1).

[0135] For example, taking the product of the matrix corresponding to LUT element 1 and the matrix corresponding to LUT element 2 as an example, the specific expression is:

[0136] The specific process of determining the product of the matrix corresponding to LUT element 2 and the matrix corresponding to LUT element 1 by reusing the product of the matrix corresponding to LUT element 1 and LUT element 2 is as follows:

[0137] but

[0138] Optionally, in this embodiment of the application, when the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0139] For example, taking the matrix corresponding to the third LUT element as the matrix corresponding to LUT element 5 in the LUT table, and the matrix corresponding to the fourth LUT element as the matrix corresponding to LUT element 5 in the LUT table as well, for instance... Figure 4 As shown, the matrix corresponding to LUT element 5 is offset by 1 relative to the matrix corresponding to the first element, LUT element 1, in the LUT table. Therefore, offsetting the matrix corresponding to LUT element 1 will determine the matrix corresponding to LUT element 5. Thus, in calculating the product of the conjugate transpose of the matrix corresponding to the third LUT element (the matrix corresponding to LUT element 5) and the matrix corresponding to the fourth LUT element (the matrix corresponding to LUT element 5), the conjugate transpose of the matrix corresponding to the first LUT element (the matrix corresponding to LUT element 1) and the product of the matrix corresponding to the second LUT element (the matrix corresponding to LUT element 1) can be reused. Based on the offsets of the matrix corresponding to the third LUT element (the matrix corresponding to LUT element 5) relative to the matrix corresponding to the first LUT element (the matrix corresponding to LUT element 1) and the offsets of the matrix corresponding to the fourth LUT element (the matrix corresponding to LUT element 5) relative to the matrix corresponding to the second LUT element (the matrix corresponding to LUT element 1), the matrix can be calculated as follows:

[0140] in,

[0141] In other words, the matrix corresponding to LUT elements on the same diagonal can be determined by processing the first element of the diagonal. It is possible The process involves deleting the first element and adding an additional element to the last line. Therefore... Available The solution can be quickly obtained by deleting or adding certain elements based on the given information.

[0142]

[0143] in,

[0144] Taking into account the characteristics of the piecewise model, the size of the compression matrix constructed in this application is... The matrix actually used in the matrix multiplication is The calculation process is as follows: Figure 8 As shown.

[0145] Taking a LUT with 54 elements as an example, the efficiency of the algorithm in this application is analyzed from the perspectives of time complexity and space complexity. It is also compared with traditional algorithms, as shown in Table 1 (time complexity represents the number of main numerical operations, and space complexity represents the variable space used for the main numerical operations).

[0146] Table 1

[0147] Note: C represents the memory space occupied by a double-precision complex number.

[0148] Meanwhile, the absolute error is used as the final measure, and the absolute error is calculated using the following formula:

[0149] in, The result of matrix calculations in the Python system. The final result calculated by this algorithm .

[0150] In the methods described above, the time consumed in calculating model coefficients in the prior art is the main time cost of the nonlinear modeling process. Faced with large LUT table configurations, traditional algorithms require allocating huge memory spaces for generating the U matrix and consume a significant amount of time. First target value Second target value The computational overhead of traditional methods wastes significant time, preventing digital predistortion from updating parameters in real-world business scenarios and hindering its adaptability to constantly evolving business environments, thus severely impacting user experience. This application addresses this by proposing a nonlinear system model and employing algorithms to compress and reuse the input signal matrix, effectively saving runtime memory, significantly reducing reliance on hardware memory, and greatly improving the computational efficiency of digital predistortion. This, in turn, enhances the computational efficiency of the nonlinear system model's coefficients and the generation efficiency of the nonlinear system model itself.

[0151] The apparatus for generating nonlinear system models provided by the present invention will be described below. The apparatus described below corresponds to the method described above for generating nonlinear system models. The apparatus for generating nonlinear system models according to embodiments of this application is as follows: Figure 9 As shown, it includes: Module 910 is established to build a piecewise linear nonlinear system model; Compression module 920 is used to obtain the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT; The generation module 930 is used to determine the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0152] In some embodiments, the compression module 920 is specifically used for: Retrieve the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; Determine the short matrix of the input signal based on the matrices corresponding to the elements of the multiple target LUTs; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

[0153] In some embodiments, the compression module 920 is specifically used for: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of the target element is not equal to 0.

[0154] In some embodiments, the generation module 930 is specifically used for: Determine the short matrix of the input signal based on the input signal compression matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the elements of multiple target LUTs and the output signal matrix of the nonlinear system.

[0155] In some embodiments, the generation module 930 is specifically used for: The first target value is determined based on the matrices corresponding to multiple target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. The second target value is determined based on the matrix corresponding to the elements of the multiple target LUTs; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first and second objective values.

[0156] In some embodiments, the generation module 930 is specifically used for: Based on the matrices corresponding to multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

[0157] In some embodiments, the generation module 930 is specifically used for: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0158] In some embodiments, the generation module 930 is specifically used for: Based on the position of each LUT element in the LUT table, determine the offset of each LUT element relative to the target LUT element; Based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0159] In some embodiments, the generation module 930 is specifically used for: Determine the x-coordinate and y-coordinate values ​​of the LUT element in the LUT table; The smaller of the x-coordinate and y-coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

[0160] In some embodiments, the generation module 930 is specifically used for: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

[0161] In some embodiments, the generation module 930 is specifically used for: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element whose order in the LUT table is greater than or equal to the matrix corresponding to the first LUT element. The product of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by the product of the conjugate transpose of the matrix corresponding to the first LUT element, the product of the matrix corresponding to the second LUT element, the product of the conjugate transpose of the matrix corresponding to the third LUT element, and the product of the matrix corresponding to the fourth LUT element.

[0162] In some embodiments, the generation module 930 is specifically used for: If the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0163] It should be noted that the apparatus provided in this embodiment of the invention can implement all the method steps implemented in the above method embodiment and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0164] Figure 10 A schematic diagram of the physical structure of an electronic device is provided. This electronic device may include a processor 1010, a communications interface 1020, a memory 1030, and a communication bus 1040. The processor 1010, communications interface 1020, and memory 1030 communicate with each other via the communication bus 1040. The processor 1010 can call logical instructions in the memory 1030 to execute a method for generating a nonlinear system model. This method includes: establishing a piecewise linear nonlinear system model; obtaining an input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and a preset lookup table (LUT); and determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0165] In some embodiments, obtaining the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and a preset LUT includes: Retrieve the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; Determine the short matrix of the input signal based on the matrices corresponding to the elements of the multiple target LUTs; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

[0166] In some embodiments, the short input signal matrix is ​​compressed to obtain a compressed input signal matrix, including: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of the target element is not equal to 0.

[0167] In some embodiments, determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system includes: Determine the short matrix of the input signal based on the input signal compression matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the elements of multiple target LUTs and the output signal matrix of the nonlinear system.

[0168] In some embodiments, the model coefficients of the nonlinear system model are determined based on the matrices corresponding to multiple target LUT elements and the output signal matrix of the nonlinear system, including: The first target value is determined based on the matrices corresponding to multiple target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. The second target value is determined based on the matrix corresponding to the elements of the multiple target LUTs; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first and second objective values.

[0169] In some embodiments, determining a first target value based on the matrices corresponding to multiple target LUT elements and the output signal matrix of the nonlinear system includes: Based on the matrices corresponding to multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

[0170] In some embodiments, determining the matrix corresponding to each LUT element in the LUT table based on the matrices corresponding to multiple target LUT elements includes: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0171] In some embodiments, the matrix corresponding to the target LUT element is offset according to the position of each LUT element in the LUT table to determine the matrix corresponding to each LUT element in the LUT table, including: Based on the position of each LUT element in the LUT table, determine the offset of each LUT element relative to the target LUT element; Based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

[0172] In some embodiments, determining the offset of each LUT element relative to the target LUT element based on the position of each LUT element in the LUT table includes: Determine the x-coordinate and y-coordinate values ​​of the LUT element in the LUT table; The smaller of the x-coordinate and y-coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

[0173] In some embodiments, determining a second target value based on a matrix corresponding to multiple target LUT elements includes: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

[0174] In some embodiments, determining the second target value based on the matrix corresponding to each LUT element in the LUT table includes: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element whose order in the LUT table is greater than or equal to the matrix corresponding to the first LUT element. The product of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by the product of the conjugate transpose of the matrix corresponding to the first LUT element, the product of the matrix corresponding to the second LUT element, the product of the conjugate transpose of the matrix corresponding to the third LUT element, and the product of the matrix corresponding to the fourth LUT element.

[0175] In some embodiments, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element, including: If the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

[0176] Furthermore, the processor 1010 described above can be a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or a Complex Programmable Logic Device (CPLD), and the processor can also adopt a multi-core architecture. The processor executes any of the methods described in the embodiments of this application according to the obtained executable instructions by calling a computer program stored in memory. The processor and memory can also be physically separated.

[0177] The logical instructions in the aforementioned memory 1030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0178] It should be noted that the electronic device provided in this embodiment of the invention can implement all the method steps implemented in the above method embodiment and achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0179] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the nonlinear system model generation method provided by the above methods. The method includes: establishing a piecewise linear nonlinear system model; obtaining an input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and a preset lookup table (LUT); and determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0180] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for generating a nonlinear system model provided by the above methods. The method includes: establishing a piecewise linear nonlinear system model; obtaining an input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and a preset lookup table (LUT); and determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

[0181] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0182] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a nonlinear system model, characterized in that, include: Establish a piecewise linear nonlinear system model; The input signal compression matrix is ​​obtained based on the correspondence between the input signal sequence of the nonlinear system and the preset lookup table (LUT). The model coefficients of the nonlinear system model are determined based on the input signal compression matrix and the output signal matrix of the nonlinear system.

2. The method for generating a nonlinear system model according to claim 1, characterized in that, The nonlinear system model includes: in, This represents the output signal of a nonlinear system. These represent the model coefficients of a nonlinear system model; Indicates time delay Input signals per unit; This represents a piecewise function; M represents the preset number of segments; This represents the signal input to the nonlinear system at time point n; Indicates time delay Signal of one unit; This represents the signal input to the nonlinear system at time point n-1; This represents the signal input to the nonlinear system at time point n+1. N is the length of the input and output signals.

3. The method for generating a nonlinear system model according to claim 1, characterized in that, The step of obtaining the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT includes: Based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT, obtain the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal in the LUT table; The short matrix of the input signal is determined based on the matrices corresponding to the multiple target LUT elements; The short matrix of the input signal is compressed to obtain the compressed matrix of the input signal.

4. The method for generating a nonlinear system model according to claim 3, characterized in that, The step of compressing the short matrix of the input signal to obtain the compressed matrix of the input signal includes: The input signal compression matrix is ​​obtained based on each target element and its position in the short input signal matrix; the value of each target element is not equal to 0.

5. The method for generating a nonlinear system model according to claim 1, characterized in that, The step of determining the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system includes: Based on the input signal compression matrix, determine the short input signal matrix; Based on the short matrix of the input signal, determine the matrix corresponding to multiple target LUT elements in the LUT table; the target LUT elements are located at the first and second positions of each diagonal line in the LUT table. The model coefficients of the nonlinear system model are determined based on the matrices corresponding to the multiple target LUT elements and the output signal matrix of the nonlinear system.

6. The method for generating a nonlinear system model according to claim 5, characterized in that, The step of determining the model coefficients of the nonlinear system model based on the matrices corresponding to the multiple target LUT elements and the output signal matrix of the nonlinear system includes: A first target value is determined based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system; the first target value is the product of the conjugate transpose of the input signal matrix and the output signal matrix. Based on the matrices corresponding to the multiple target LUT elements, a second target value is determined; the second target value is the product of the conjugate transpose of the input signal matrix and the input signal matrix. The model coefficients of the nonlinear system model are determined based on the first target value and the second target value.

7. The method for generating a nonlinear system model according to claim 6, characterized in that, The step of determining the first target value based on the matrices corresponding to the plurality of target LUT elements and the output signal matrix of the nonlinear system includes: Based on the matrices corresponding to the multiple target LUT elements, determine the matrix corresponding to each LUT element in the LUT table; The input signal matrix is ​​determined based on the matrix corresponding to each LUT element in the LUT table; The first target value is determined based on the input signal matrix and the output signal matrix.

8. The method for generating a nonlinear system model according to claim 7, characterized in that, The step of determining the matrix corresponding to each LUT element in the LUT table based on the matrices corresponding to the plurality of target LUT elements includes: Based on the position of each LUT element in the LUT table, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

9. The method for generating a nonlinear system model according to claim 8, characterized in that, The step of offsetting the matrix corresponding to the target LUT element based on the position of each LUT element in the LUT table to determine the matrix corresponding to each LUT element in the LUT table includes: Based on the position of each LUT element in the LUT table, determine the offset of each LUT element relative to the target LUT element; Based on the offset of each LUT element relative to the target LUT element, the matrix corresponding to the target LUT element is offset to determine the matrix corresponding to each LUT element in the LUT table.

10. The method for generating a nonlinear system model according to claim 9, characterized in that, The step of determining the offset of each LUT element relative to the target LUT element based on the position of each LUT element in the LUT table includes: Determine the x-coordinate value of the LUT element in the LUT table and the y-coordinate value of the LUT element in the LUT table; The smaller of the horizontal and vertical coordinate values ​​is determined as the offset of the LUT element relative to the target LUT element.

11. The method for generating a nonlinear system model according to any one of claims 6-10, characterized in that, The step of determining the second target value based on the matrix corresponding to the plurality of target LUT elements includes: Based on the matrix corresponding to the target LUT element, determine the matrix corresponding to each LUT element in the LUT table; The second target value is determined based on the matrix corresponding to each LUT element in the LUT table.

12. The method for generating a nonlinear system model according to claim 11, characterized in that, Determining the second target value based on the matrix corresponding to each LUT element in the LUT table includes: Determine the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element; the matrix corresponding to the first LUT element is any one of the matrices corresponding to the target LUT element; the matrix corresponding to the second LUT element is the matrix corresponding to the LUT element in the LUT table whose order is greater than or equal to the matrix corresponding to the first LUT element. The product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is used to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. The second target value is determined by multiplying the matrix corresponding to the first LUT element by the conjugate transpose of the matrix corresponding to the second LUT element, the matrix corresponding to the third LUT element by the conjugate transpose of the matrix corresponding to the third LUT element, and the matrix corresponding to the fourth LUT element.

13. The method for generating a nonlinear system model according to claim 12, characterized in that, The step of reusing the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element includes: When the matrix corresponding to the third LUT element is equal to the matrix corresponding to the second LUT element, and the matrix corresponding to the fourth LUT element is equal to the matrix corresponding to the first LUT element, the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element is conjugate transposed to determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element. Otherwise, determine the first offset of the matrix corresponding to the third LUT element relative to the matrix corresponding to the first LUT element, and the second offset of the matrix corresponding to the fourth LUT element relative to the matrix corresponding to the second LUT element; reuse the product of the conjugate transpose of the matrix corresponding to the first LUT element and the matrix corresponding to the second LUT element, and based on the first offset and the second offset, determine the product of the conjugate transpose of the matrix corresponding to the third LUT element and the matrix corresponding to the fourth LUT element.

14. A device for generating a nonlinear system model, characterized in that, include: Establish a module for building piecewise linear nonlinear system models; The compression module is used to obtain the input signal compression matrix based on the correspondence between the input signal sequence of the nonlinear system and the preset LUT; The generation module is used to determine the model coefficients of the nonlinear system model based on the input signal compression matrix and the output signal matrix of the nonlinear system.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for generating a nonlinear system model as described in any one of claims 1 to 13.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for generating a nonlinear system model as described in any one of claims 1 to 13.

Images