Dynamic pre-correction method and system for UHF signals of ground-oriented digital television

Abstract

The application discloses a UHF signal dynamic pre-correction method and system for ground digital television, relates to the field of signal pre-correction, and comprises the following steps: collecting a detection signal sequence of a UHF frequency band in real time, and obtaining an actual output signal processed by a power amplifier; separating the actual output signal into a detection signal component and a television data signal component; performing distortion feature extraction, and calculating and obtaining a nonlinear distortion parameter matrix; inputting the nonlinear distortion parameter matrix into a model as input features for learning and training based on a time domain-frequency domain double channel of a double-loop convolutional neural network, and obtaining a pre-distortion function; inputting the television data signal into the pre-distortion function, and outputting a pre-correction signal; and dynamically adjusting the pre-correction strength through a closed-loop control mechanism. The application has the advantages that: through deep learning technology, signal components are accurately separated, distortion features are extracted, and the pre-correction strength is dynamically adjusted, so that the transmission quality and stability of the ground digital television signal are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal pre-correction, particularly to a UHF signal dynamic pre-correction method and system for ground digital television. BACKGROUND

[0002] The UHF signal dynamic pre-correction method and system for ground digital television is a technical solution designed to address the problem of multipath interference and fading that may occur in modern digital television signal transmission. Through this system, the characteristics of the received UHF signal, such as signal strength, phase shift, and time delay, can be monitored and analyzed in real time, and dynamic correction can be performed using pre-set algorithms and models to improve the signal quality and stability at the receiving end.

[0003] The current market UHF signal dynamic pre-correction method for ground digital television is limited in detecting signal separation and extracting distortion characteristics, usually using basic signal processing technology, which is difficult to accurately capture the nonlinear distortion characteristics of signals in complex environments. This leads to the fact that they may not be able to accurately predict and adjust signal distortion during the pre-correction process, thereby affecting the final transmission quality and stability. In addition, the pre-correction method commonly used in existing systems often lacks effective closed-loop control mechanisms, and cannot monitor and respond to actual performance changes in signal transmission in real time. This makes it difficult for the system to adjust the correction strategy in real time when faced with dynamic channel conditions or external interference, thereby affecting its ability to adapt to complex environments and its sustained stability. In addition, some traditional systems lack the ability to analyze the frequency domain and model the time domain of signals during preprocessing, limiting their accuracy and precision in capturing signal dynamic changes and nonlinear distortion characteristics. SUMMARY

[0004] In order to improve the existing method and system, the UHF signal dynamic pre-correction method and system for ground digital television is provided, which accurately separates signal components and extracts distortion characteristics by real-time acquisition of UHF signals and use of deep learning technology, dynamically adjusts the pre-correction strength, and significantly improves the transmission quality and stability of ground digital television signals.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is:

[0006] The UHF signal dynamic pre-correction method for ground digital television comprises:

[0007] Real-time acquisition of a probe signal sequence in the UHF frequency band, extraction of the pilot sequence in the signal as a reference benchmark, and acquisition of the actual output signal after power amplifier processing;

[0008] Based on the sparse Fourier transform algorithm, the actual output signal is separated into a probe signal component and a television data signal component;

[0009] Based on the separated probe signal components, distortion feature extraction is performed, and a nonlinear distortion parameter matrix is calculated and obtained, including amplitude-phase conversion coefficients, amplitude-amplitude conversion gradients, memory effect depth factors, and intermodulation product distribution weights;

[0010] Based on the time-domain-frequency-domain double-channel of the double-cycle convolutional neural network, the nonlinear distortion parameter matrix is input into the model as an input feature for learning and training to obtain a pre-distortion function;

[0011] Based on the obtained pre-distortion function, the real-time collected television data signal is input into the pre-distortion function, and a pre-correction signal is output, and the pre-correction strength is dynamically adjusted through a closed-loop control mechanism.

[0012] Preferably, the probe signal sequence in the UHF frequency band is collected in real time, and the pilot sequence in the signal is extracted as a reference benchmark, and the actual output signal after being processed by the power amplifier specifically includes:

[0013] A directional coupler is configured at the output end of the power amplifier to extract a mixed feedback signal containing the probe signal and the television signal;

[0014] The collected feedback signal is converted into a digital signal stream;

[0015] According to a predefined frame synchronization header, the time slot starting position of the probe signal is determined, a pure pilot component is separated and obtained through quadrature demodulation, and a reference benchmark signal is output through phase calibration and amplitude normalization;

[0016] The complex analytic form of the power amplifier output signal is obtained through signal alignment processing to obtain the actual output signal.

[0017] Preferably, the actual output signal is separated into a probe signal component and a television data signal component based on the sparse Fourier transform algorithm, specifically including:

[0018] Based on the obtained actual output signal, data preprocessing is performed, and the frequency spectrum of the signal is divided into multiple "buckets", each bucket representing a group of frequency ranges;

[0019] Discrete Fourier transform is performed on each bucket to convert the signal from the time domain to the frequency domain;

[0020] The main frequency components of the signal are obtained based on the matching pursuit algorithm frequency domain representation, and the frequency spectrum components in the signal are separated;

[0021] The probe signal frequency band components that meet the probe signal spectrum distribution in the frequency domain are extracted, and the frequency components related to the television data signal are extracted from the frequency domain;

[0022] Based on the extracted frequency band components, inverse sparse Fourier transform is respectively performed to restore the frequency domain signals to time domain, and the probe signal component and the television data signal component are obtained.

[0023] Preferably, based on the separated probe signal component, distortion feature extraction is performed, and a nonlinear distortion parameter matrix is calculated and obtained, including amplitude-phase conversion coefficient, amplitude-amplitude conversion gradient, memory effect depth factor and intermodulation product distribution weight, specifically including:

[0024] Based on the offset of the output signal phase of the probe signal component relative to the input signal, the amplitude-phase conversion coefficient is obtained by performing cubic polynomial fitting on the data points;

[0025] Based on the amplitude variation rate of the signal of the probe signal component, the amplitude-amplitude conversion gradient is calculated and obtained;

[0026] The memory effect depth factor is calculated based on the third-order cross-correlation algorithm;

[0027] The intermodulation component in the probe signal component is extracted, and the intermodulation product power proportion is calculated by combining the subcarriers of the probe signal two by two, and the intermodulation product distribution weight matrix is obtained;

[0028] Based on the above obtained parameter data, a nonlinear distortion parameter matrix is constructed.

[0029] Preferably, the time-frequency dual channel based on the double-loop convolutional neural network inputs the nonlinear distortion parameter matrix into the model as an input feature for learning and training, and obtains a predistortion function, specifically including:

[0030] A time-frequency dual channel based on a double-loop convolutional neural network is constructed;

[0031] The time domain channel extracts the memory effect time delay feature through the causal convolution layer, and the convolution kernel length is associated with the thermal time constant of the power amplifier;

[0032] The frequency domain channel learns the frequency domain intermodulation response through the complex convolution layer, and fuses the time-frequency features in the output layer;

[0033] Based on the obtained nonlinear distortion parameter matrix, it is input into the time-frequency dual channel for learning and training;

[0034] The predistortion function is generated by the time-frequency dual channel learning.

[0035] Preferably, based on the obtained predistortion function, the real-time collected television data signal is input into the predistortion function, and a pre-correction signal is output, and the pre-correction strength is dynamically adjusted through a closed-loop control mechanism, specifically including:

[0036] Based on the obtained pre-distortion function, the real-time acquisition obtained television data signal is inputted into the pre-distortion function;

[0037] The pre-distortion function generates a corresponding pre-correction signal according to the nonlinear distortion characteristics of the input signal, and preliminarily verifies the characteristics of the output pre-correction signal;

[0038] A closed-loop control mechanism is constructed to monitor the performance of the output signal in real time, and the error between the actual output signal and the target signal is collected through a feedback loop;

[0039] Based on the error analysis, the parameters or pre-correction strength of the pre-distortion function are dynamically adjusted through PID control.

[0040] Further, a UHF signal dynamic pre-correction system for ground digital television is proposed, which comprises:

[0041] The signal acquisition module acquires the probe signal sequence of the UHF frequency band in real time, extracts the pilot sequence as a reference benchmark, and obtains the actual output signal processed by the power amplifier;

[0042] The signal separation module separates the actual output signal into a probe signal component and a television data signal component based on the sparse Fourier transform algorithm;

[0043] The distortion feature extraction module extracts distortion features and calculates a nonlinear distortion parameter matrix based on the separated probe signal component;

[0044] The pre-distortion function module learns the nonlinear distortion parameter matrix through the time-domain-frequency-domain double channel of the double-loop convolutional neural network to generate a pre-distortion function;

[0045] The pre-correction signal module inputs the real-time acquisition of the television data signal into the pre-distortion function to generate a pre-correction signal;

[0046] The closed-loop control module is used to monitor the performance of the output signal in real time, and dynamically adjust the pre-distortion function parameters or pre-correction strength through a feedback loop and PID control;

[0047] The processor is used to process the calculation process of each formula and the construction calculation process of each model.

[0048] Compared with the prior art, the advantages of the present application are:

[0049] By collecting the detection signal of UHF frequency band in real time and extracting the pilot sequence, the actual output condition of the signal can be accurately reflected. By using the sparse Fourier transform algorithm, the detection signal and the television data signal components can be effectively separated, and the nonlinear distortion feature can be extracted, so that the accurate distortion parameter matrix can be obtained. The double-cycle convolutional neural network combines the time domain and frequency domain double-channel mode, so that the pre-distortion function can more accurately adapt to various signal distortion modes and dynamically adjust the correction strength. The closed-loop control mechanism further enhances the real-time performance and adaptability of the system, so that the pre-correction signal can be optimized according to the error feedback in the output process. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The method proposed by the application is shown in the schematic diagram.

[0051] Figure 2 The actual output signal acquisition schematic diagram proposed by the application is shown in the schematic diagram.

[0052] Figure 3 The actual output signal separation schematic diagram proposed by the application is shown in the schematic diagram.

[0053] Figure 4 The distortion feature extraction schematic diagram proposed by the application is shown in the schematic diagram.

[0054] Figure 5 The pre-distortion function acquisition schematic diagram proposed by the application is shown in the schematic diagram.

[0055] Figure 6 The output pre-correction signal schematic diagram proposed by the application is shown in the schematic diagram. DETAILED DESCRIPTION

[0056] The following description is used to disclose the application so that those skilled in the art can implement the application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0057] The UHF signal dynamic pre-correction system for ground digital television comprises:

[0058] The signal acquisition module: the signal acquisition module collects the detection signal sequence of UHF frequency band in real time, extracts the pilot sequence as the reference benchmark, and obtains the actual output signal after the power amplifier processing;

[0059] The signal separation module: the signal separation module separates the actual output signal into detection signal component and television data signal component based on the sparse Fourier transform algorithm;

[0060] The distortion feature extraction module: the distortion feature extraction module extracts the distortion feature and calculates the nonlinear distortion parameter matrix based on the separated detection signal component;

[0061] The pre-distortion function module learns a nonlinear distortion parameter matrix through a time-domain-frequency-domain double channel of a double-loop convolutional neural network to generate a pre-distortion function.

[0062] The pre-correction signal module inputs the real-time collected television data signal into the pre-distortion function to generate a pre-correction signal.

[0063] The closed-loop control module is used to monitor the performance of the output signal in real time, and dynamically adjust the pre-distortion function parameters or the pre-correction strength through a feedback loop and a PID control.

[0064] The processor is used to process the calculation process of each formula and the construction calculation process of each model.

[0065] Referring to Figure 1 The UHF signal dynamic pre-correction method for ground digital television shown in the figure comprises:

[0066] Step one: real-time collection of a probe signal sequence in a UHF frequency band, extraction of a pilot sequence in the signal as a reference benchmark, and acquisition of an actual output signal after processing by a power amplifier;

[0067] Step two: separation of the actual output signal into a probe signal component and a television data signal component based on a sparse Fourier transform algorithm;

[0068] Step three: distortion feature extraction based on the separated probe signal component, calculation and acquisition of a nonlinear distortion parameter matrix, including amplitude-phase conversion coefficients, amplitude-amplitude conversion gradients, memory effect depth factors, and intermodulation product distribution weights;

[0069] Step four: input of the nonlinear distortion parameter matrix into the model as an input feature for learning and training based on a time-domain-frequency-domain double channel of a double-loop convolutional neural network, and acquisition of a pre-distortion function;

[0070] Step five: input of the real-time collected television data signal into the pre-distortion function based on the acquired pre-distortion function, output of a pre-correction signal, and dynamic adjustment of the pre-correction strength through a closed-loop control mechanism.

[0071] Referring to Figure 2 The real-time collection of a probe signal sequence in a UHF frequency band, the extraction of a pilot sequence in the signal as a reference benchmark, and the acquisition of an actual output signal after processing by a power amplifier specifically comprises:

[0072] A directional coupler is configured at the output end of the power amplifier to extract a mixed feedback signal containing the probe signal and the television signal;

[0073] The collected feedback signal is converted into a digital signal stream;

[0074] According to the predefined frame synchronization header, the time slot starting position of the probe signal is determined, the pure pilot component is separated through quadrature demodulation, the reference reference signal is output through phase calibration and amplitude normalization;

[0075] The complex analytic form of the power amplifier output signal is output, and the actual output signal is obtained through signal alignment processing.

[0076] Specifically, a directional coupler is configured at the output end of the power amplifier, which is used to extract a mixed feedback signal containing the probe signal and the television signal from the output signal. The directional coupler extracts part of the output signal through the coupling port while maintaining the main signal transmission. The analog feedback signal extracted through the directional coupler is converted into a digital signal stream for subsequent digital signal processing.

[0077] The predefined frame synchronization header sequence is searched in the digital signal stream, the cross-correlation of the signal and the frame synchronization header is calculated, the peak position is found, and the time slot starting point is determined.

[0078] The pilot component of the probe signal is separated from the mixed signal through quadrature demodulation, and the signal segment of the time slot where the probe signal is located is extracted based on the time slot starting point. The signal is multiplied by the local carrier using quadrature demodulation to separate the in-phase and quadrature components. The two components are low-pass filtered to remove high-frequency components to obtain the baseband pilot signal, where the in-phase component and the quadrature component formula are:

[0079] I[n]=s pilot [n]·cos(2πf c nT s )

[0080] Q[n]=s pilot [n]·sin(2πf c nT s )

[0081] Where I[n] is the in-phase component, Q[n] is the quadrature component, s pilot [n] is the signal segment of the time slot where the probe signal is located, f c is the local carrier, i.e. the frequency, T s is the sampling period, and n is the sampling point index.

[0082] The separated pilot component is phase calibrated and amplitude normalized to generate a reference reference signal. The power amplifier output signal is processed through the complex analytic signal representation to generate the actual output signal in the complex analytic form.

[0083] Referring to Figure 3 , based on the sparse Fourier transform algorithm, the actual output signal is separated into a probe signal component and a television data signal component, which specifically includes:

[0084] Based on the acquired actual output signal, data preprocessing is performed, and the frequency spectrum of the signal is divided into multiple "buckets", each bucket representing a group of frequency ranges;

[0085] A discrete Fourier transform is performed on each bucket to convert the signal from the time domain to the frequency domain;

[0086] The main frequency components of the signal are obtained based on the matching pursuit algorithm in the frequency domain representation, and the frequency components in the signal are separated;

[0087] The frequency band components of the probe signal that match the frequency spectrum distribution of the probe signal in the frequency domain are extracted, and the frequency components related to the television data signal are extracted from the frequency domain;

[0088] Based on the extracted frequency band components, inverse sparse Fourier transforms are respectively performed to restore the frequency domain signal to the time domain, and the probe signal component and the television data signal component are obtained.

[0089] Specifically, the frequency spectrum of the actual output signal is divided into multiple "buckets", each bucket corresponding to a frequency range. According to the frequency range of each bucket, the signal component of the corresponding frequency is extracted through a band-pass filter, and a discrete Fourier transform is applied to each bucket signal segment to obtain a frequency domain representation;

[0090] In the frequency domain representation of each bucket, a dictionary is constructed, containing possible frequency basis functions, and the frequency basis functions that best match the signal are iteratively selected to extract the main frequency components. According to the known spectral distribution characteristics of the probe signal and the television signal, the respective frequency components are separated. The matching pursuit iteration formula is:

[0091]

[0092] where S k [m] is the frequency domain representation of each bucket, m is the frequency index, d i [m] is the i-th frequency basis function in the dictionary, is the complex conjugate of d i [m];

[0093] According to the known spectral characteristics of the probe signal (such as bandwidth, center frequency), the frequency band components of the probe signal are extracted from the frequency domain representation of each bucket, and according to the spectral characteristics of the television signal (such as modulation method, bandwidth), the frequency band components of the television signal are extracted;

[0094] Inverse sparse Fourier transforms are applied to the probe signal frequency domain components and the television signal frequency domain components to restore the corresponding time domain signals, and the time domain signals of all buckets are combined to obtain the complete probe signal component and the television signal component.

[0095] Referring to Figure 4As shown, based on the separated probe signal components, distortion feature extraction is performed, and a nonlinear distortion parameter matrix is calculated, including amplitude-phase conversion coefficients, amplitude-amplitude conversion gradients, memory effect depth factors, and intermodulation product distribution weights, specifically including:

[0096] Based on the offset of the output signal phase of the probe signal component relative to the input signal, the amplitude-phase conversion coefficient is obtained by fitting the data points with a cubic polynomial;

[0097] Based on the amplitude variation rate of the probe signal component, the amplitude-amplitude conversion gradient is calculated;

[0098] The memory effect depth factor is calculated based on the third-order cross-correlation algorithm;

[0099] Extract the intermodulation component in the probe signal component, and calculate the intermodulation product power ratio by combining the subcarriers of the probe signal two by two to obtain the intermodulation product distribution weight matrix;

[0100] Based on the above obtained parameter data, a nonlinear distortion parameter matrix is constructed.

[0101] Specifically, the offset of the probe signal output phase relative to the input signal reflects the phase distortion characteristics of the nonlinear system, and the amplitude-phase conversion coefficient is obtained by calculating the phase difference between the input probe signal and the output signal and fitting a cubic polynomial;

[0102] Based on the first derivative calculation of the input and output amplitude relationship, it can be approximately obtained by fitting method, and the calculation result is the amplitude-amplitude conversion gradient;

[0103] By constructing a third-order cross-correlation function, the delay order corresponding to the maximum correlation value is obtained, which is the memory effect depth factor;

[0104] The probe signal contains multiple subcarriers, the intermodulation frequency is calculated for all combinations, the power of the intermodulation component at the corresponding frequency is extracted, the total intermodulation power is normalized to form an intermodulation product power ratio matrix, and an intermodulation product distribution weight matrix is obtained;

[0105] Integrate all kinds of nonlinear parameters to form a comprehensive modeling matrix that describes the behavior of the power amplifier.

[0106] Referring to Figure 5 As shown, based on the time-domain-frequency-domain double channel of the double-loop convolutional neural network, the nonlinear distortion parameter matrix is input into the model as an input feature for learning and training to obtain a predistortion function, specifically including:

[0107] Construct a time-domain-frequency-domain double channel based on a double-loop convolutional neural network;

[0108] The time domain channel extracts the memory effect time delay feature through a causal convolution layer, and the length of the convolution kernel is associated with the thermal time constant of the power amplifier.

[0109] The frequency domain channel learns the frequency domain intermodulation response through a complex convolution layer, and fuses the time-frequency features in the output layer.

[0110] Based on the obtained nonlinear distortion parameter matrix, it is input into the time domain-frequency domain double channel for learning training.

[0111] The pre-distortion function is generated through the time domain-frequency domain double channel learning.

[0112] Specifically, the purpose of the time domain channel is to capture the nonlinear distortion characteristics of the signal in the time domain, and to extract the memory effect time delay feature of the power amplifier. The input signal is processed through causal convolution operation, and the length of the convolution kernel is associated with the thermal time constant of the power amplifier. Causal convolution means that the output signal depends on the current and previous input, and the formula is:

[0113]

[0114] Where x(t) is the input signal, y(t) is the output signal, w k is the convolution kernel, and L is the length of the convolution kernel.

[0115] The frequency domain channel is used to capture the frequency domain characteristics of the signal, especially the frequency domain intermodulation response of the power amplifier. The frequency domain channel uses complex convolution operation to process the frequency domain characteristics of the input signal. Complex convolution is more suitable for complex signal processing, especially for intermodulation distortion problems, and the formula is:

[0116]

[0117] Where f(f) is the frequency domain representation of the input signal, y f (f) is the frequency domain output signal, W k is the complex convolution kernel, and L f is the length of the frequency domain convolution kernel.

[0118] After the time domain channel and the frequency domain channel extract the time domain and frequency domain features respectively, the time-frequency features are fused in the output layer through weighted summation or splicing.

[0119] Based on the learned nonlinear distortion parameter matrix, a pre-distortion function is generated, which is designed to compensate for the nonlinear distortion of the power amplifier.

[0120] Referring to Figure 6 Based on the obtained pre-distortion function, the real-time collected television data signal is input into the pre-distortion function, and the pre-correction signal is output, and the pre-correction strength is dynamically adjusted through a closed loop control mechanism, which specifically includes:

[0121] Based on the acquired pre-distortion function, the real-time acquisition acquired television data signal is inputted into the pre-distortion function;

[0122] The pre-distortion function generates the corresponding pre-correction signal according to the nonlinear distortion characteristics of the input signal, and preliminarily verifies the characteristics of the output pre-correction signal;

[0123] A closed-loop control mechanism is constructed to monitor the performance of the output signal in real time, and the error between the actual output signal and the target signal is collected through the feedback loop;

[0124] Based on the error analysis, the parameters or pre-correction strength of the pre-distortion function are dynamically adjusted through the PID control.

[0125] Specifically, at the output end of the system, the error between the output signal and the target signal is monitored, and the error information is fed back to the control system in real time through the feedback loop. The error is inputted into the PID controller, and the PID controller generates an adjustment signal according to the error, its derivative and integral part.

[0126] In each cycle, the system continues to monitor the error between the output signal and the target signal, and continuously adjusts the parameters of the pre-distortion function according to the PID control algorithm.

[0127] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. And the above description of the specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0128] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0129] The above is only the preferred embodiment of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A dynamic pre-correction method for UHF signals for terrestrial digital television, characterized in that, include: The system acquires the detection signal sequence in the UHF band in real time, extracts the pilot sequence from the signal as a reference, and obtains the actual output signal after processing by the power amplifier. Based on the sparse Fourier transform algorithm, the actual output signal is separated into the probe signal component and the television data signal component. Based on the separated detection signal components, distortion features are extracted, and the nonlinear distortion parameter matrix is ​​calculated, including amplitude-phase conversion coefficient, amplitude-amplitude conversion gradient, memory effect depth factor, and intermodulation product distribution weight. Based on a dual-recurrent convolutional neural network with time-frequency dual-path, the nonlinear distortion parameter matrix is ​​input into the model as input features for learning and training to obtain the predistortion function. Based on the acquisition predistortion function, the real-time acquired TV data signal is input into the predistortion function, the pre-correction signal is output, and the pre-correction intensity is dynamically adjusted through a closed-loop control mechanism; The method of separating the actual output signal into probe signal components and television data signal components based on the sparse Fourier transform algorithm specifically includes: performing data preprocessing based on the acquired actual output signal and dividing the signal spectrum into multiple "buckets," each representing a set of frequency ranges; performing a discrete Fourier transform on each bucket to convert the signal from the time domain to the frequency domain; obtaining the main frequency components of the signal based on the frequency domain representation of the matching pursuit algorithm and separating the spectral components in the signal; extracting the probe signal frequency band components that conform to the spectral distribution of the probe signal in the frequency domain, and extracting the frequency components related to the television data signal from the frequency domain; and performing inverse sparse Fourier transforms on the extracted frequency band components to restore the frequency domain signal to the time domain, thereby obtaining the probe signal components and television data signal components. The process of extracting distortion features based on the separated probe signal components and calculating a nonlinear distortion parameter matrix, including amplitude-phase conversion coefficients, amplitude-amplitude conversion gradients, memory effect depth factors, and intermodulation product distribution weights, specifically includes: obtaining amplitude-phase conversion coefficients by performing cubic polynomial fitting on data points based on the phase offset of the output signal relative to the input signal of the probe signal components; calculating the amplitude-amplitude conversion gradient based on the amplitude change rate of the probe signal components; calculating the memory effect depth factor based on a third-order cross-correlation algorithm; extracting intermodulation components from the probe signal components and calculating the power proportion of intermodulation products by pairwise combination of the subcarriers of the probe signal to obtain an intermodulation product distribution weight matrix; and constructing a nonlinear distortion parameter matrix based on the obtained parameter data. The aforementioned time-domain-frequency domain dual-path based on a dual-recurrent convolutional neural network, which inputs the nonlinear distortion parameter matrix into the model as input features for learning and training to obtain the predistortion function, specifically includes: constructing a time-domain-frequency domain dual-path based on a dual-recurrent convolutional neural network; the time-domain path extracts memory effect delay features through causal convolutional layers, with the convolutional kernel length correlated with the thermal time constant of the power amplifier; the frequency-domain path learns the frequency domain intermodulation response through complex convolutional layers and fuses time-frequency features at the output layer; based on the obtained nonlinear distortion parameter matrix, it is input into the time-domain-frequency domain dual-path for learning and training; and the predistortion function is generated through learning via the time-domain-frequency domain dual-path.

2. The UHF signal dynamic pre-correction method for terrestrial digital television according to claim 1, characterized in that, The real-time acquisition of the UHF band detection signal sequence, the extraction of the pilot sequence from the signal as a reference, and the acquisition of the actual output signal after power amplifier processing specifically include: A directional coupler is configured at the output of the power amplifier to extract a mixed feedback signal containing both the probe signal and the television signal; The acquired feedback signals are converted into digital signal streams; Based on the predefined frame synchronization header, the start position of the time slot of the probe signal is determined, the pure pilot component is obtained by quadrature demodulation separation, and the reference reference signal is output by phase calibration and amplitude normalization. The actual output signal is obtained by processing the signal alignment through the complex analytical form of the power amplifier output signal.

3. The UHF signal dynamic pre-correction method for terrestrial digital television according to claim 1, characterized in that, The process of acquiring a predistortion function, inputting the real-time acquired television data signal into the predistortion function, outputting a pre-correction signal, and dynamically adjusting the pre-correction intensity through a closed-loop control mechanism specifically includes: Based on the acquired predistortion function, the real-time acquired television data signal is input into it; The predistortion function generates a corresponding pre-correction signal based on the nonlinear distortion characteristics of the input signal, and performs preliminary verification of the characteristics of the output pre-correction signal. A closed-loop control mechanism is constructed to monitor the performance of the output signal in real time, and the error between the actual output signal and the target signal is collected through the feedback loop. Based on error analysis, the parameters of the predistortion function or the pre-correction intensity are dynamically adjusted through PID control.

4. A UHF signal dynamic pre-correction system for terrestrial digital television, used to implement the UHF signal dynamic pre-correction method for terrestrial digital television as described in any one of claims 1-3, characterized in that, include: Signal acquisition module: The signal acquisition module acquires the detection signal sequence of the UHF band in real time, extracts the pilot sequence as a reference, and obtains the actual output signal after processing by the power amplifier; Signal separation module: The signal separation module is based on the sparse Fourier transform algorithm to separate the actual output signal into a probe signal component and a television data signal component; Distortion feature extraction module: The distortion feature extraction module extracts distortion features and calculates the nonlinear distortion parameter matrix based on the separated detection signal components; Predistortion function module: The predistortion function module learns the nonlinear distortion parameter matrix and generates the predistortion function through the time-frequency dual-pathway of the dual recurrent convolutional neural network; Pre-correction signal module: The pre-correction signal module inputs the real-time acquired television data signal into the predistortion function to generate a pre-correction signal; Closed-loop control module: The closed-loop control module is used to monitor the output signal performance in real time and dynamically adjust the predistortion function parameters or pre-correction intensity through feedback loop and PID control; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images