HPLCHRF dual-mode communication adaptive coding modulation and anti-noise method based on deep learning

Abstract

The invention discloses an HPLCamp (High Performance Liquid Chromatography) based on deep learning. The invention discloses an HRF dual-mode communication adaptive coding modulation and anti-noise method. The method comprises the following steps: acquiring an optical radio frequency signal amplitude-phase change rate and synchronously sampling and normalizing; calculating a node amplitude-phase residual error to generate a nonlinear mapping coefficient; monitoring coherent change to solve a drift trend, adjusting a modulation coding optimization scheme, compensating distortion and outputting an anti-noise result. According to the method, the instantaneous amplitude and phase of the optical radio frequency dual-mode signal are extracted, a multi-dimensional amplitude-phase characteristic matrix is formed in combination with time domain synchronization and a normalization template, differential residual modeling and nonlinear mapping coefficient calculation are carried out between impedance nodes, and dynamic compensation of amplitude-phase mismatch and envelope offset is achieved. A drift trend quantity is generated based on coherent offset parameter differentiation, feedback is provided for modulation format and coding strategy optimization, amplitude equalization and phase correction are completed, the signal synchronization degree and amplitude-phase consistency are improved, and the steady-state response and anti-disturbance performance of a transmission link are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of transmission line technology, and particularly relates to a deep learning-based adaptive coding modulation and noise reduction method for HPLC & HRF dual-mode communication. Background Technology

[0002] The field of transmission line technology encompasses research on the orderly transmission of electromagnetic energy or electrical signals within conductors and dielectric structures. Its core lies in the efficient transmission and stability control of signals under different media, frequencies, and impedance matching conditions. This technology covers multiple aspects, including high-speed signal transmission, bandwidth optimization, electromagnetic coupling control, noise suppression, and signal integrity maintenance. With the development of optoelectronic integration and intelligent control, transmission line technology has gradually expanded from traditional physical parameter matching design to research on multimodal signal coordination and adaptive coding. This includes systematic analysis of high-frequency link transmission characteristics, nonlinear distortion compensation, and multi-signal coupling interference modeling, providing fundamental support for achieving high-fidelity signal transmission and stable data transmission in complex environments.

[0003] Among them, the deep learning-based adaptive coding modulation and noise reduction method for HPLC & HRF dual-mode communication refers to the use of deep learning models for feature extraction and pattern recognition in high-frequency link communication structures, targeting the collaborative characteristics of high-frequency optical links and high-frequency radio frequency dual-mode transmission. This enables the joint design of dynamic matching and error correction in the modulation and coding stage. This includes joint modeling of HPLC channel parameter variation characteristics and HRF transmission path characteristics, adaptive coding design for signal amplitude and phase coupling characteristics, and a multi-dimensional feature weight allocation mechanism achieved through neural network parameter mapping. The method primarily establishes learnable signal mapping relationships, synchronously optimizes and dynamically modulates signal characteristics of different transmission paths, realizing an adaptive coding modulation process based on feature coupling. Furthermore, it combines this with a noise estimation model to construct the logical framework for generating noise reduction strategies.

[0004] Existing transmission line methods rely heavily on static impedance matching and linear modeling during signal propagation, lacking a joint description of amplitude and phase differences between different nodes for dual-mode signals, and failing to establish a synchronous correction mechanism in the time and frequency domains. Signals are prone to envelope shift and phase drift accumulation during multi-node transmission, leading to decreased coherence and waveform distortion, and making it difficult to maintain consistent modulation depth and carrier amplitude. Under noise disturbance conditions, system feedback lag reduces compensation accuracy, and residual distortion in the transmission path is difficult to suppress in real time, resulting in increased bit errors and signal quality fluctuations, making it difficult to adapt to the dynamic changes required by high-frequency links. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a deep learning-based adaptive coding modulation and noise reduction method for HPLC & HRF dual-mode communication. This method enables continuous calibration of the signal propagation path, significantly reduces phase instability and envelope distortion caused by node differences, improves signal synchronization and amplitude-phase consistency, and enhances the steady-state response and anti-disturbance performance of the transmission link in complex impedance environments.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A deep learning-based adaptive coding modulation and noise immunity method for HPLC & HRF dual-mode communication includes the following steps:

[0008] S1: Obtain the amplitude change rate, phase change rate, and modulation depth of the optical and radio frequency signals, synchronously sample the signals to calculate the instantaneous amplitude and phase, perform filtering and normalization, and obtain the signal preprocessing results;

[0009] S2: Call the signal preprocessing results, calculate the amplitude difference and phase shift between the coaxial cable and the microstrip line based on the amplitude and phase residuals of the optical radio frequency signal at different impedance nodes, generate the intermodulation residual itemset and normalize it, and back-project to generate nonlinear mapping coefficients.

[0010] S3: Based on the signal preprocessing results and the nonlinear mapping coefficients, monitor the time delay difference and coherence ratio change, calculate the time delay difference and instantaneous coherence ratio, generate a coherence offset parameter set, calculate the differential, and obtain the coherence drift trend.

[0011] S4: Based on the coherent drift trend and the nonlinear mapping coefficient, compare the modulation depth and carrier amplitude in the optical domain and the radio frequency domain, perform modulation format and coding scheme adjustment, and optimize the generation of an adaptive scheme;

[0012] S5: Based on the adaptive coding and modulation scheme, the coherent drift trend, and the nonlinear mapping coefficient, compensate and fine-tune the signal amplitude attenuation and phase disturbance, correct the distortion, calculate the stability, and output the noise immunity result.

[0013] As a further aspect of the present invention, the signal preprocessing result includes an amplitude feature matrix, a phase feature matrix, a synchronization index table, and a normalized template; the nonlinear mapping coefficients include an amplitude-phase weight table, an envelope offset vector, a node response matrix, and a mapping reference library.

[0014] As a further aspect of the present invention, the coherent drift trend includes a time delay offset curve, a coherence ratio fluctuation sequence, and a trend stability index; the adaptive coding and modulation scheme includes a modulation format list, a coding strategy table, and an amplitude equalization parameter set; and the noise immunity correction results include an amplitude calibration set, a phase calibration set, and a stability evaluation label.

[0015] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0016] S101: Obtain the amplitude change rate, phase change rate, and modulation depth of the optical and radio frequency signals in the HPLC and HRF link; detect the voltage amplitude and phase angle change rate of the sampled signal; record continuous sample data based on the sampling time series; perform point-by-point calculation on the voltage change and phase angle difference between each sampling point; and generate an instantaneous amplitude and phase dataset.

[0017] S102: Based on the instantaneous amplitude and phase dataset, call the amplitude change rate and phase change rate to perform location judgment on the signal amplitude change point and phase jump point, perform sliding weighted smoothing filtering according to the time interval of continuous sampling interval and amplitude-phase difference, calculate the difference between the mean of the amplitude interval and the mean of the phase interval after filtering, and perform normalization operation according to the amplitude-phase difference interval range to obtain the amplitude-phase filtering normalization matrix.

[0018] S103: Based on the amplitude and phase filtering normalization matrix, the instantaneous amplitude and phase samples of the optical domain and radio frequency signals are aligned and compared according to the sampling time. The amplitude difference interval and phase difference interval of the two signals at the synchronous sampling point are calculated. The correlation ratio is obtained based on the rate of change of the difference interval, and the signal preprocessing result is generated.

[0019] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0020] S201: Call the signal preprocessing results to obtain the sampled voltage sequence and phase angle sequence, use the node impedance value to calculate the amplitude difference between adjacent nodes, use the phase angle sequence to calculate the phase difference component, compare the amplitude and phase difference values ​​point by point and record the difference interval and node index, perform normalization labeling on the difference interval, and generate a node residual information set.

[0021] S202: Based on the node residual information set, obtain the node mapping table of coaxial cable and microstrip line, perform differential operation on the node amplitude difference according to the mapping table, perform subtraction on the phase difference according to the mapping table and calculate the rate of change of the difference interval, extract the intermodulation frequency related components in the difference interval and arrange the sequence, classify and organize the sequence according to the node attributes and label the sequence index to obtain the optical radio frequency intermodulation residual itemset.

[0022] S203: Based on the optical radio frequency intermodulation residual itemset, perform time-domain moving average and normalization operations on the residual item sequence, establish a standardized residual vector and back-project it to the original signal space according to the sampling node position, compare the difference range between the back-projected envelope amplitude and the initial sampling envelope, calculate the difference ratio of each sampling point and convert it into node mapping coefficients, summarize the node mapping coefficients, and generate nonlinear mapping coefficients.

[0023] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0024] S301: Based on the signal preprocessing results and the nonlinear mapping coefficients, collect the time delay value and coherence ratio value of each sampling point of HPLC, collect the time delay value and coherence ratio value of each sampling point of HRF, pair HPLC and HRF data according to the sampling point index, calculate the time delay difference for each pair and record the difference sign and index, calculate the coherence ratio difference for each pair and record the difference interval and index, merge the time delay difference and coherence ratio difference according to the sampling point order to form a sequence, and generate the time delay difference and coherence ratio difference sequence;

[0025] S302: Based on the time delay difference and coherence ratio difference sequence, the difference is calculated for adjacent sampling index pairs to obtain a difference value sequence. A set of difference windows is selected to perform a moving average operation on the difference value sequence. The mean and variance of each window are calculated and the start and end indices of the window are recorded. The difference magnitudes are compared at the difference abrupt change positions within the window and the abrupt change index is marked. The difference statistics are summarized into a parameter vector according to the sampling points and sorted. A parameter vector sequence is established and a coherence offset parameter set is obtained.

[0026] S303: Based on the coherent offset parameter set, the discrete derivative of the parameter vector sequence is calculated in time order to obtain the instantaneous rate of change sequence. The change ratio of adjacent points is calculated for the instantaneous rate of change sequence and the change ratio index is recorded. The change ratio is aggregated according to the sampling points to generate point-level drift. The point-level drift is summarized and the sequence mean and trend coefficient are calculated to obtain the coherent drift trend.

[0027] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0028] S401: Based on the coherent drift trend and the nonlinear mapping coefficient, detect the waveguide section and coupler transmission loss parameters, collect the optical domain modulation depth and carrier amplitude values, collect the radio frequency modulation depth and carrier amplitude values, pair the amplitude and modulation depth of the optical domain and radio frequency domain according to the sampling index, subtract the radio frequency amplitude from the optical domain amplitude and divide the radio frequency depth by the optical domain depth to generate the optical radio frequency modulation comparison coefficient.

[0029] S402: Based on the optical radio frequency modulation contrast coefficient, calculate the arithmetic mean and variance of the amplitude difference sequence, call the waveguide loss rate and coupler loss rate to synthesize the correction coefficient, subtract the correction coefficient from the amplitude difference for each sampling point and normalize it with the maximum absolute value of the sequence as the normalization factor, compare the point-by-point difference of the amplitude difference before and after correction and locate the continuous index interval with the smallest difference, and generate the amplitude difference minimization interval parameter.

[0030] S403: Based on the amplitude difference minimization interval parameter, call the nonlinear mapping coefficient, calculate the coding criterion value according to the ratio of the left and right boundaries of the interval and the rate of change of the coherent drift trend, perform category mapping on the criterion value to select the coding scheme and identify the corresponding modulation depth combination, perform weighted summation on the criterion values ​​of each sampling segment to form a parameter set, and generate an adaptive coding modulation scheme.

[0031] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0032] S501: Based on the adaptive coding modulation scheme, the coherent drift trend and the nonlinear mapping coefficient, detect the amplitude and phase values ​​of the continuously sampled transmission signal, calculate the amplitude attenuation rate of each sampling point as the difference between the current amplitude and the previous amplitude divided by the sampling interval, and count the phase difference between adjacent sampling points as the current phase minus the previous phase divided by the sampling interval to form the amplitude-phase offset ratio as the amplitude attenuation rate divided by the absolute value of the phase difference. Output the ratio sequence according to the sampling order to generate the amplitude-phase offset ratio sequence.

[0033] S502: Based on the amplitude-phase offset ratio sequence, calculate the arithmetic mean and variance of the sequence, call the nonlinear mapping coefficient to calculate the nonlinear offset correction amount for each sampling point by multiplying the mapping coefficient by the corresponding ratio, perform subtraction correction for amplitude and addition / subtraction fine adjustment for phase, and synchronously reconstruct the amplitude-phase correspondence by pairing the corrected amplitude and corrected phase according to the sampling index, and statistically calculate the deviation of the corrected amplitude-phase signal according to the sampling order by summing the squares of the difference between the corrected amplitude and its local mean at each point and taking the average to obtain the nonlinear distortion correction coefficient;

[0034] S503: Based on the nonlinear distortion correction coefficient, calculate the amplitude mean square stability rate of the corrected signal as the ratio of the variance to the mean of the corrected amplitude sequence, calculate the phase correlation as the point-to-point correlation coefficient of the corrected phase sequence, multiply the amplitude mean square stability rate and the phase correlation by the set amplitude and phase weight ratio respectively and sum them, calculate the sum of the stability parameter and the noise threshold difference and perform normalization processing, establish a noise immunity performance index set, and generate noise immunity correction results.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0036] In this invention, the instantaneous amplitude and phase of the optical-RF dual-mode signal are extracted, and a multidimensional amplitude and phase characteristic matrix is ​​formed by combining time-domain synchronization and normalization templates. Differential residual modeling and nonlinear mapping coefficient calculation are introduced between impedance nodes to achieve dynamic compensation for amplitude-phase mismatch and envelope offset. Differential analysis based on coherent offset parameters generates drift trend quantities, providing feedback for modulation format selection and coding strategy optimization, thus completing the amplitude equalization and phase correction process. This logic achieves continuous calibration of the signal propagation path, significantly reducing phase instability and envelope distortion caused by node differences, improving signal synchronization and amplitude-phase consistency, and enhancing the steady-state response and disturbance rejection performance of the transmission link in complex impedance environments. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the steps of the present invention;

[0038] Figure 2 This is a flowchart of the main steps of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0040] Please see Figures 1-2 A deep learning-based adaptive coding modulation and noise immunity method for HPLC & HRF dual-mode communication includes the following steps:

[0041] S1: Obtain the amplitude change rate, phase change rate, and modulation depth of the optical and radio frequency signals in the HPLC and HRF link; perform time-domain synchronous sampling on the optical and radio frequency signals and calculate the instantaneous amplitude and instantaneous phase; perform time-domain filtering and normalization on the signal amplitude and phase data based on the calculation results to obtain the signal preprocessing results.

[0042] S2: Call the signal preprocessing results, and calculate the voltage amplitude difference and phase offset of the signal between the coaxial cable and microstrip line nodes based on the amplitude and phase residuals of the optical radio frequency signal at different impedance nodes in the transmission line. Generate the optical radio frequency intermodulation residual itemset through differential comparison, perform time-domain moving average and normalization operations on the residual itemset to generate a standardized residual vector, back-project the vector to the original signal space, calculate the signal envelope offset and phase difference, and generate nonlinear mapping coefficients based on the differences.

[0043] S3: Based on the signal preprocessing results and nonlinear mapping coefficients, monitor the time delay difference and coherence ratio change during the propagation of HPLC and HRF link signals, calculate the time delay difference and instantaneous coherence ratio of the optical radio frequency signals, perform differential calculation on the two to generate a coherence offset parameter set, and further calculate the derivative to detect the drift trend and obtain the coherence drift trend.

[0044] S4: Based on the coherent drift trend and nonlinear mapping coefficient, the modulation depth and carrier amplitude in the optical domain and the radio frequency domain are compared according to the transmission loss parameters of the waveguide section and the coupler in the transmission line. The modulation format is adjusted and the coding scheme is selected. The coding strategy is optimized by minimizing the carrier amplitude difference to generate an adaptive coding modulation scheme.

[0045] S5: Based on the adaptive coding modulation scheme, coherent drift trend and nonlinear mapping coefficient, amplitude compensation and phase fine-tuning are performed on the amplitude attenuation and phase disturbance in the transmitted signal to correct the nonlinear distortion caused by noise, calculate the amplitude and phase stability of the corrected signal and output the noise correction result.

[0046] Signal preprocessing results include amplitude feature matrix, phase feature matrix, synchronization index table, and normalization template; nonlinear mapping coefficients include amplitude and phase weight table, envelope offset vector, node response matrix, and mapping reference library; coherence drift trend includes time delay offset curve, coherence ratio fluctuation sequence, and trend stability index; adaptive coding and modulation scheme includes modulation format list, coding strategy table, and amplitude equalization parameter set; noise immunity correction results include amplitude calibration set, phase calibration set, and stability evaluation labels.

[0047] Amplitude feature matrix: used to represent the amplitude distribution of a signal in different time or frequency domains, usually a two-dimensional array of signal amplitude and time or frequency; its value reflects the change of signal strength with time or frequency.

[0048] Phase characteristic matrix: Represents the change of signal phase with time or frequency. It is commonly used to describe the phase behavior of signals in the time or frequency domain and to analyze the pattern of phase change.

[0049] Synchronization Index Table: Used to describe the synchronization relationship between different signal sources or processing procedures, ensuring that the data of multiple signals can be correctly aligned in the time domain or frequency domain;

[0050] Normalization template: It is a standardization processing tool used to convert the amplitude or phase values ​​of signal data into a standard range, which is usually used to remove deviations from different signal sources or measurement conditions;

[0051] Amplitude and phase weighting table: In signal processing, different weights are assigned to different amplitude and phase responses to adjust their contribution in the synthesis or modulation process;

[0052] Envelope offset vector: describes the direction and magnitude of the change in the signal envelope (i.e., the outer envelope of the signal), and is usually used to analyze the modulation characteristics and waveform stability of the signal;

[0053] Node response matrix: records the response characteristics of a signal at different transmission nodes (such as between coaxial cables and microstrip lines), and is often used for impedance matching and attenuation analysis during signal transmission;

[0054] Mapping reference library: A standard reference used to record nonlinear mappings, typically used in modulation and demodulation processes to provide reference parameters for calculating mapping coefficients;

[0055] Delay offset curve: describes the time delay change of a signal during propagation, and is usually used for system delay compensation and synchronization problems;

[0056] Coherence ratio fluctuation sequence: describes the fluctuation of the coherence ratio (i.e. the ratio of signal to noise) during signal propagation, and is usually used to evaluate signal quality and stability;

[0057] Trend stability index: Used to measure the stability of a signal and the degree of trend change, typically used for long-term monitoring of signal behavior;

[0058] Modulation format list: describes various signal modulation methods (such as QAM, PSK, etc.) to select the most suitable modulation method for the current transmission environment;

[0059] Encoding Strategy Table: Lists different encoding strategies used to improve signal noise immunity and transmission efficiency;

[0060] Amplitude equalization parameter group: used to balance signal amplitude, minimizing amplitude differences in different transmission paths and system components, thereby improving overall system performance;

[0061] Amplitude calibration set: This includes all the parameters used to correct signal amplitude, typically for correcting signal attenuation or gain non-uniformity.

[0062] Phase calibration set: This includes a set of all parameters that correct the phase of a signal, typically used to adjust phase drift or phase error;

[0063] Stability assessment label: Used to quantitatively assess the stability of a signal, commonly used in long-term data monitoring to help determine whether the signal quality is stable;

[0064] The specific steps of S1 are as follows:

[0065] S101: Obtain the amplitude change rate, phase change rate, and modulation depth of the optical and radio frequency signals in the HPLC and HRF link; detect the voltage amplitude and phase angle change rate of the sampled signal; record continuous sample data based on the sampling time series; perform point-by-point calculation on the voltage change and phase angle difference between each sampling point; and generate an instantaneous amplitude and phase dataset.

[0066] To obtain the amplitude change rate, phase change rate, and modulation depth of the optical and radio frequency signals in the HPLC-HRF link, the optical and radio frequency signals were first sampled independently within a continuous sampling period. The voltage amplitude and phase angle of each sampling point were recorded. The sampling interval was fixed within a resolvable time scale. The amplitude change rate of the optical signal can be expressed as the difference between the amplitudes of two adjacent sampling points divided by the amplitude of the previous sampling. The change rate of the radio frequency signal was obtained in the same way. Then, the modulation depth of the optical and radio frequency signals was set as the ratio of the difference between their maximum and minimum amplitudes. If the optical signal varies between 1.8V and 2.2V, the modulation depth is (2.2-1.8) / 2.2=0.18. The sampling time series was recorded simultaneously, and the voltage and phase change values ​​of each sampling point were labeled as Vi. And θi, and then calculate point by point based on the voltage difference ΔV=Vi-Vi-1 and the phase difference Δθ=θi-θi-1 between adjacent sampling points to generate a set of instantaneous change values ​​of amplitude and phase. During the execution process, when the amplitude change rate of each sampling point exceeds the set change threshold, the point is marked as an amplitude change point, and when the phase change rate exceeds the set angle change threshold, it is marked as a phase change point. The change threshold is set according to 1.5 times the standard deviation of amplitude and the standard deviation of phase within the continuous sampling interval. Then, all sampling points are reordered according to the time series, and the amplitude change rate, phase change rate, modulation depth and time index are merged into a four-dimensional sample set according to the sampling points so that the optical domain and radio frequency signals can be compared and calculated under the same time sequence in subsequent processing, and finally the instantaneous amplitude and phase dataset is generated.

[0067] S102: Based on the instantaneous amplitude and phase dataset, the amplitude change rate and phase change rate are called to locate and judge the signal amplitude change point and phase jump point. The sliding weighted smoothing filter is performed according to the time interval of the continuous sampling interval and the amplitude-phase difference. The difference between the mean of the amplitude interval and the mean of the phase interval after filtering is calculated. The normalization operation is performed according to the range of the amplitude-phase difference interval to obtain the amplitude-phase filtering normalization matrix.

[0068] Based on the instantaneous amplitude and phase dataset, the amplitude change rate and phase change rate are used to locate and determine signal amplitude abrupt changes and phase jump points. First, the amplitude change threshold is set to the mean amplitude change rate plus twice the standard deviation. If the amplitude change rate at any point exceeds this threshold, it is determined to be an amplitude abrupt change point. For phase jump points, the mean phase change rate plus the standard deviation is used as the judgment boundary. For each abrupt change point, its time index and direction of change are recorded. Then, a sliding weighted smoothing filter is performed based on the time interval Δt of the continuous sampling interval and the amplitude-phase difference Δφ. The filter weight coefficient w depends on the reciprocal of the time interval, making the contribution of time-adjacent sampling points greater. The mean amplitude interval Ā after filtering is calculated as ∑(w×Ai) / ∑w. The mean value Φ̄ of the phase interval is calculated as ∑(w×Φi) / ∑w. Based on the difference intervals between amplitude and phase ΔA=Ai-Ā and ΔΦ=Φi-Φ̄, interval normalization is performed on both. The interval maximum and minimum value method is adopted, that is, all ΔA and ΔΦ are subtracted from the minimum value and divided by the difference between the maximum and minimum values. The normalized value range is limited to 0 to 1. In the example, if the mean amplitude after filtering is 2.0V and the amplitude of a certain sampling point is 2.3V, then the normalized amplitude difference is (2.3-1.89) / (2.6-1.8)=0.625. All normalized values ​​are arranged in time index to form a matrix. The matrix rows correspond to the sampling points, and the columns correspond to the amplitude and phase. Finally, the amplitude and phase filtering normalization matrix is ​​obtained.

[0069] S103: Based on the amplitude and phase filter normalization matrix, the instantaneous amplitude and phase samples of the optical domain and radio frequency signals are aligned and compared according to the sampling time. The amplitude difference interval and phase difference interval of the two signals at the synchronous sampling point are calculated. The correlation ratio is obtained based on the rate of change of the difference interval, and the signal preprocessing result is generated.

[0070] Based on the amplitude and phase filtering normalization matrix, the instantaneous amplitude and phase samples of the optical and radio frequency signals are aligned and compared according to the sampling time. First, using the time series of the optical signal as a reference, the sampling time of the radio frequency signal is adjusted to the same sampling interval through linear interpolation to ensure that the synchronous sampling points are aligned. Then, the amplitude difference interval ΔAij = Ai(light) - Aj(radiation) and the phase difference interval ΔΦij = Φi(light) - Φj(radiation) at the same time are calculated. The absolute value of the difference of all sampling points is taken, and the interval change rates rA and rΦ are statistically analyzed. rA is defined as the absolute value of the amplitude difference between adjacent sampling points. The ratio of the change in phase difference to the time interval is defined as rΦ, which is the ratio of the absolute change in phase difference to the time interval. Normalization is performed on both the amplitude and phase change rates, and their average ratio is calculated as the correlation ratio. If the mean value of rA is 0.04 and the mean value of rΦ is 0.05, then the correlation ratio is 0.04 / 0.05 = 0.8. Finally, this ratio is multiplied by the weighting coefficients of the corresponding points in the amplitude-phase filter normalization matrix, and the average is calculated over all samples. The resulting value reflects the degree of coupling between the optical domain and the radio frequency signal at the amplitude-phase synchronization level. This step yields the signal preprocessing result.

[0071] The specific steps of S2 are as follows:

[0072] S201: Call the signal preprocessing results, obtain the sampled voltage sequence and phase angle sequence, use the node impedance value to calculate the amplitude difference between adjacent nodes, use the phase angle sequence to calculate the phase difference component, compare the amplitude and phase difference values ​​point by point and record the difference interval and node index, perform normalization labeling on the difference interval, and generate a node residual information set.

[0073] The signal preprocessing results are retrieved to obtain the sampled voltage and phase angle sequences. First, the amplitude component is extracted as a voltage sequence Vi, and the phase component as a phase angle sequence θi. Each sampling point corresponds to the voltage and phase angle values ​​at the same moment. Then, the amplitude difference ΔVi = |Vi - Vi-1| / Zi between adjacent nodes is calculated based on the node impedance value Zi. The impedance value originates from the node load characteristics, typically obtained by measuring the ratio of the node's input current to voltage. For example, if a node's voltage is 2.0V and its current is 0.01A, its impedance is 200Ω. The impedance of adjacent nodes can be used to correct the amplitude difference to reflect changes in energy distribution. Next, the phase difference component Δθi = |θi - θi-1| is calculated using the phase angle sequence. At each sampling point, the amplitude difference and phase difference are further... The comparison is performed point by point. If the amplitude difference is greater than the phase difference, it is recorded as a positive difference interval; otherwise, it is a negative difference interval. Each interval is marked with node index i and the corresponding difference pair {ΔVi, Δθi}. Then, all difference intervals are normalized and labeled. The difference of each interval is subtracted from the minimum value of the interval and divided by the difference between the maximum and minimum values ​​to ensure that the amplitude and phase differences between different nodes are comparable within the range of 0 to 1. For example, if the amplitude difference of a node is 0.3 and the phase difference is 0.2, while the global amplitude difference interval is 0.1 to 0.5, then the normalized amplitude difference of the node is (0.3-0.1) / (0.5-0.1)=0.5. After normalization, the amplitude and phase difference values, node index and corresponding impedance values ​​of each node are integrated into data recording units and arranged according to the node sequence to finally generate a node residual information set.

[0074] S202: Based on the node residual information set, obtain the node mapping table of coaxial cable and microstrip line, perform differential operation on the node amplitude difference according to the mapping table, perform subtraction on the phase difference according to the mapping table and calculate the rate of change of the difference interval, extract the intermodulation frequency related components in the difference interval and arrange the sequence, classify and organize the sequence according to the node attributes and label the sequence index to obtain the optical radio frequency intermodulation residual itemset.

[0075] Based on the node residual information set, a mapping table of coaxial cable and microstrip line nodes is obtained. First, the physical channel number corresponding to each node in the mapping table is read to establish a one-to-one mapping relationship between optical signal nodes and radio frequency nodes. Then, differential operation is performed on the node amplitude difference according to the mapping table to calculate the amplitude difference ΔAij=ΔVi(light)-ΔVj(radiation) between adjacent mapped nodes. The phase difference is subtracted ΔΦij=Δθi(light)-Δθj(radiation), and the rate of change of the difference interval r=|ΔAij-ΔAij-1| / |ΔΦij-ΔΦij-1| is calculated. The mutual difference within the difference interval between each mapped node is extracted. Intermodulation frequency components are identified by performing discrete frequency distribution analysis on the differential amplitude sequence. For example, when the amplitude difference contains components with frequencies of 2.5MHz and 5MHz, the source node index is recorded and marked accordingly in the node mapping table. Then, the extracted frequency component sequences are classified and organized according to node attributes. If the node is on the optical domain side, it is marked as class L, and if it is on the radio frequency side, it is marked as class R. The organized sequences are arranged in ascending order by node index. Each sequence records the intermodulation frequency value, amplitude difference, phase difference and node attribute identifier, and finally the optical radio frequency intermodulation residual itemset is obtained.

[0076] S203: Based on the optical radio frequency intermodulation residual itemset, perform time-domain moving average and normalization operations on the residual item sequence, establish a standardized residual vector and back-project it to the original signal space according to the sampling node position, compare the difference range between the back-projected envelope amplitude and the initial sampling envelope, calculate the difference ratio of each sampling point and convert it into node mapping coefficients, summarize the node mapping coefficients, and generate nonlinear mapping coefficients.

[0077] Based on the optical radio frequency intermodulation residual itemset, time-domain moving average and normalization operations are performed on the residual item sequence. First, a fixed number of sampling points are selected within a continuous sampling node interval, and the moving average of the residual item amplitude Āi=∑(Aj) / n is calculated. The average amplitude value of all nodes is subtracted from the global minimum average value and divided by the difference between the maximum and minimum average values ​​to obtain the normalized amplitude value A′i. The corresponding phase residual item Φi is also normalized in the same way, forming a normalized residual vector Ri=(A′i,Φ′i). Then, this normalized residual vector is back-projected to the original signal space according to the node position, that is, the sampling points with the same index are located in the original signal sequence, and the corresponding amplitude and phase residuals are replaced. To form a corrected envelope signal, the difference interval ΔEi=|E′i-Ei| between the back-projected envelope amplitude and the initial sampled envelope is compared. The difference ratio Pi=ΔEi / Ei of each sampling point is calculated. All Pi are rearranged according to the node index and converted into node mapping coefficients ki=Pi / (∑Pi). For example, when the difference ratios of the three nodes are 0.2, 0.3, and 0.5, the corresponding node mapping coefficients are 0.2 / 1.0, 0.3 / 1.0, and 0.5 / 1.0, respectively, i.e., 0.2, 0.3, and 0.5. Finally, all node mapping coefficients are weighted and summarized to form a set of parameters reflecting the nonlinear relationship between the amplitude and phase of each node, generating nonlinear mapping coefficients.

[0078] The specific steps for S3 are as follows:

[0079] S301: Based on the signal preprocessing results and nonlinear mapping coefficients, collect the time delay value and coherence ratio value of each sampling point of HPLC, collect the time delay value and coherence ratio value of each sampling point of HRF, pair HPLC and HRF data according to the sampling point index, calculate the time delay difference for each pair and record the difference sign and index, calculate the coherence ratio difference for each pair and record the difference interval and index, merge the time delay difference and coherence ratio difference according to the sampling point order to form a sequence, and generate the time delay difference and coherence ratio difference sequence;

[0080] Based on the signal preprocessing results and nonlinear mapping coefficients, the time delay and coherence ratio values ​​of each sampling point in the HPLC signal are collected. Simultaneously, the time delay and coherence ratio values ​​of each sampling point in the HRF signal are also collected. First, each sampling point in the HPLC signal is numbered i, and its time delay Ti (HPLC) and coherence ratio Ri (HPLC) are recorded. Similarly, the HRF signal is numbered j, and Ti (HRF) and Ri (HRF) are recorded accordingly. Pairing is performed according to the sampling point index, ensuring that each i and j has the same time position. For example, if the 10th sampling point has a time delay of 2.35 in HPLC and 2.40 in HRF, then its corresponding time delay difference ΔTi = 2.35 - 2.40 = -0.05. The sign "-" indicates that the HPLC signal is faster. The coherence ratio difference ΔRi = Ri is calculated simultaneously. (HPLC)-Ri(HRF), if Ri(HPLC)=0.92 and Ri(HRF)=0.87, then ΔRi=0.05. Record this difference and the corresponding index i. Arrange the time delay difference and coherence ratio difference of all sampling points in the order of time index. Establish a data pair {ΔTi,ΔRi} for each group of differences, and label the difference sign information together with the index. For example, when the time delay difference of five consecutive sampling points is -0.04, -0.03, 0.01, 0.02, 0.00, the corresponding sign sequence is {-,-,+,+,0}, indicating that the HPLC signal of the first two points is advanced, the last two points are lagging, and the last point is synchronized. Finally, merge the time delay difference sequence and the coherence ratio difference sequence with the sampling point as the reference to form a joint sequence, and output the time delay difference and coherence ratio difference sequence.

[0081] S302: Based on the time delay difference and coherence ratio difference sequence, the difference is calculated for adjacent sampling index pairs to obtain the difference value sequence. A set of difference windows is selected to perform a moving average operation on the difference value sequence. The mean and variance of each window are calculated and the start and end indices of the window are recorded. The difference magnitude is compared at the difference abrupt change positions within the window and the abrupt change index is marked. The difference statistics are summarized into a parameter vector according to the sampling points and sorted. A parameter vector sequence is established and the coherence offset parameter set is obtained.

[0082] Based on the time delay difference and coherence ratio difference sequence, the difference value sequence is obtained by calculating the difference between adjacent sampling index pairs. First, the adjacent differences ΔTi and ΔRi are calculated in this sequence: Δ2Ti = ΔTi - ΔTi-1 and Δ2Ri = ΔRi - ΔRi-1. The second-order difference corresponding to each sampling point is recorded as a difference vector according to the index. Then, a fixed-length difference window set is selected, and a moving average operation is performed on each window. Assuming that the window contains n difference values, the window mean μk = ∑Δ2Ti / n and the variance σk² = ∑(Δ2Ti - μk)² / n. The start and end indices and statistics of each window are recorded. When determining the position of the difference abrupt change, the difference magnitude is compared between any adjacent difference values ​​within the window. When the difference is greater than a set threshold (the threshold is twice the mean of the window), the position is marked as a mutation index. For example, if the difference values ​​in the sampling point interval 20 to 25 are 0.01, 0.02, 0.05, 0.09, 0.10, and 0.03 respectively, then the window mean μk≈0.05 and the threshold is 0.1. When Δ2Ti=0.10, it is marked as mutation point 25. This process is performed on all windows. The difference statistics are summarized according to the sampling point index to form a parameter vector Vi=(μk,σk²,ik) containing the mean, variance, and mutation index. All parameter vectors are arranged in the sampling order to form a parameter vector sequence, and these sequences are classified as a whole to obtain the coherent offset parameter set.

[0083] S303: Based on the coherent migration parameter set, the discrete derivative of the parameter vector sequence is calculated in time order to obtain the instantaneous rate of change sequence. The change ratio of adjacent points is calculated for the instantaneous rate of change sequence and the change ratio index is recorded. The change ratio is aggregated according to the sampling points to generate point-level drift. The point-level drift is summarized and the sequence mean and trend coefficient are calculated to obtain the coherent drift trend.

[0084] Based on the coherent migration parameter set, the discrete derivative of the parameter vector sequence is calculated in time order to obtain the instantaneous rate of change sequence. Specifically, the mean term μk in adjacent parameter vectors is differencing to obtain the instantaneous rate of change Δμk = μk - μk-1. For each Δμk, the change ratio rk = |Δμk| / |μk-1| between adjacent points is calculated, and the sampling index k corresponding to the change ratio is recorded. The change ratio sequence is locally aggregated using a sliding window method, and the change ratios of several consecutive points are averaged to generate the point-level drift Di for that interval. All point-level drifts are aggregated and averaged in the order of sampling points. The value D̄ = ∑Di / N is used to calculate the drift trend coefficient Ct, which is the correlation coefficient between each Di and the sampling index. Ct = Cov(Di,i) / (σDσi), where Cov represents the covariance, and σD and σi are the standard deviations of the drift and the index, respectively. If the Di sequence in the sample is 0.02, 0.03, 0.04, 0.05, 0.05, then its average drift D̄ = 0.038, and the trend coefficient Ct is positive, indicating that the drift increases with the increase of the sampling index. Finally, D̄ and Ct are jointly labeled under the same sequence index to obtain the coherent drift trend.

[0085] The specific steps of S4 are as follows:

[0086] S401: Based on the coherent drift trend and nonlinear mapping coefficient, detect the transmission loss parameters of the waveguide section and coupler, collect the optical domain modulation depth and carrier amplitude values, collect the radio frequency modulation depth and carrier amplitude values, pair the amplitude and modulation depth of the optical domain and radio frequency domain according to the sampling index, subtract the radio frequency amplitude from the optical domain amplitude and divide the radio frequency depth by the optical domain depth to generate the optical radio frequency modulation comparison coefficient.

[0087] Based on the coherent drift trend and nonlinear mapping coefficient, the transmission loss parameters of the waveguide section and coupler are detected. First, the waveguide section transmission loss LW(i) and coupler loss LC(i) are read at each sampling index position. Both are expressed in decibels. The measured values ​​are usually derived from the power input-output ratio calculation. If the input power is 10mW and the output power is 8mW, then LW = 10 × log10(10 / 8) = 0.97dB. Similarly, LC = 0.55dB is measured. Then, the modulation depth MDL(i) and carrier amplitude AL(i) of the optical domain signal and the modulation depth MDR(i) and carrier amplitude AR(i) of the radio frequency signal are collected. For example, at a certain sampling point, the optical domain modulation depth is 0.20 and the amplitude is 2.1V. With a modulation depth of 0.25 and an amplitude of 1.9V, the amplitude and modulation depth of the optical and radio frequency domains are paired according to the sampling index. For each pair of sampling points, the amplitude difference ΔAi=AL(i)-AR(i)=0.2V is calculated, and the modulation depth ratio Ri=MDL(i) / MDR(i)=0.20 / 0.25=0.8 is calculated. The difference and ratio and their corresponding index i are recorded. This operation is performed on the entire sequence to ensure that the optical and radio frequency have matching datasets under the same sampling time. Then, the mean Ct of the coherent drift trend and the nonlinear mapping coefficient ki are combined for correction and labeling, so that the data pair structure of each sampling point is expanded to {ΔAi,Ri,Ct,ki}. All data points are arranged in index order to form the optical and radio frequency modulation contrast coefficients.

[0088] S402: Based on the optical radio frequency modulation contrast coefficient, the arithmetic mean and variance of the amplitude difference sequence are calculated. The waveguide loss rate and the coupler loss rate are subtracted to synthesize the correction coefficient. For each sampling point, the amplitude difference is subtracted from the correction coefficient and normalized with the maximum absolute value of the sequence as the normalization factor. The point-by-point difference of the amplitude difference before and after correction is compared and the continuous index interval with the smallest difference is located. The amplitude difference minimization interval parameter is generated.

[0089] Based on the optical radio frequency modulation contrast coefficient, the arithmetic mean μA and variance σA² are calculated for the amplitude difference sequence ΔAi, μA=∑ΔAi / N, σA²=∑(ΔAi-μA)² / N, to obtain the overall distribution range of the amplitude difference. Then, the waveguide loss rate ηW and the coupler loss rate ηC are subtracted to obtain the correction coefficient κ=ηW-ηC. If ηW=0.12 and ηC=0.08, then κ=0.04. For each sampling point, the corrected amplitude difference ΔA′i=ΔAi−κ is calculated, and the maximum absolute value Amax=max|ΔA′i| in the sequence is taken as the normalization factor to perform the normalization operation Ai*=ΔA′i / Amax, forming the amplitude difference normalized sequence. A point-by-point comparison is performed between the modified sequence and the unmodified sequence to calculate the absolute difference Δdiffi=|ΔAi-Ai*|. Then, the interval with the smallest consecutive Δdiffi is found according to the index scanning method. When the Δdiffi of three or more consecutive sampling points is lower than 50% of the global mean μdiff, the interval is defined as the interval with the minimum amplitude difference. For example, when Δdiffi is 0.02, 0.01, 0.015, 0.02 and 0.018 respectively between sampling points 15-19, and μdiff=0.04, the interval meets the condition and is identified as {15,19}. The left and right boundary indices of the interval and the corresponding average difference are recorded, and finally the parameters of the interval with the minimum amplitude difference are generated.

[0090] S403: Based on the minimum amplitude difference interval parameter, call the nonlinear mapping coefficient, calculate the coding criterion value according to the ratio of the left and right boundaries of the interval and the rate of change of the coherent drift trend, perform category mapping on the criterion value to select the coding scheme and identify the corresponding modulation depth combination, perform weighted summation of the criterion values ​​of each sampling segment to form a parameter set, and generate an adaptive coding modulation scheme.

[0091] According to the interval parameter that minimizes the amplitude difference, call the non - linear mapping coefficient, and calculate the coding criterion value based on the ratio of the left and right boundaries of the interval and the change rate of the coherent drift trend. First, calculate the boundary ratio β = Ai(il) / Ai*(ir) using the normalized amplitude difference value Ai corresponding to the left boundary index il and the right boundary index ir of the interval. Then, combined with the average change rate of the coherent drift trend within the interval ΔCt=(Ct(ir)-Ct(il)) / (ir - il), calculate the criterion value J = β×(1 + ΔCt). If β = 0.8 and ΔCt = 0.1, then J = 0.8×1.1 = 0.88. Classify the obtained criterion values according to the set threshold intervals. For example, J≤0.7 is judged as the low - code class, 0.7 < J≤0.9 is judged as the medium - code class, and J>0.9 is judged as the high - code class. Each class corresponds to a coding scheme and identifies its modulation depth combination. For example, the low - code class corresponds to (0.2,0.3), the medium - code class corresponds to (0.3,0.4), and the high - code class corresponds to (0.4,0.5). Perform this classification for each sampling segment and record its criterion category, modulation depth group, and index. Subsequently, weighted - sum the criterion values of each sampling segment according to the non - linear mapping coefficient ki to form the parameter set Pi = ∑(ki×Ji) / ∑ki. When the criterion values of three sampling segments are 0.75, 0.85, and 0.95 respectively, and the corresponding weights are 0.3, 0.4, and 0.3, then Pi=(0.75×0.3 + 0.85×0.4 + 0.95×0.3)=0.855. Finally, integrate and output the parameter set in the sampling order to form an adaptive coding modulation scheme;

[0092] The specific steps of S5 are as follows:

[0093] S501: According to the adaptive coding modulation scheme, the coherent drift trend, and the non - linear mapping coefficient, detect the amplitude value and phase value of the continuously sampled transmission signal. Calculate the amplitude attenuation rate of each sampling point as the difference between the current amplitude and the previous amplitude divided by the sampling interval. Statistically, the phase difference between adjacent sampling points is the current phase minus the previous phase and divided by the sampling interval. Form the amplitude - phase offset ratio as the amplitude attenuation rate divided by the absolute value of the phase difference. Output the ratio sequence in the sampling order to generate the amplitude - phase offset ratio sequence;

[0094] According to the adaptive coding modulation scheme, the coherent drift trend, and the non - linear mapping coefficient, first detect the amplitude value and phase value of each sampling point in the transmission signal. For each sampling point i, calculate the amplitude attenuation rate ΔAi=(Ai - Ai - 1) / Δt, where Ai is the amplitude value of the i - th sampling point and Δt is the sampling interval. Record the amplitude attenuation rate of each sampling point. Then calculate the phase difference between adjacent sampling points Δθi=(θi - θi - 1) / Δt, where θi is the phase value of the i - th sampling point. According to the amplitude attenuation rate and the phase difference, calculate the amplitude - phase offset ratio Bi = |ΔAi| / |Δθi|. Output and organize the amplitude - phase offset ratios of all sampling points in the sampling order. Finally, generate the amplitude - phase offset ratio sequence;

[0095] S502: Based on the amplitude-phase offset ratio sequence, calculate the arithmetic mean and variance of the sequence, call the nonlinear mapping coefficient to calculate the nonlinear offset correction amount for each sampling point by multiplying the mapping coefficient by the corresponding ratio, perform subtraction correction for amplitude and addition / subtraction fine adjustment for phase, and synchronously reconstruct the amplitude-phase correspondence by pairing the corrected amplitude and corrected phase according to the sampling index, and statistically calculate the deviation of the corrected amplitude and phase signals according to the sampling order by summing the squares of the difference between the corrected amplitude and its local mean at each point and taking the average to obtain the nonlinear distortion correction coefficient;

[0096] Based on the amplitude-phase offset ratio sequence, the arithmetic mean μB and variance σB² of the sequence are first calculated to understand the overall distribution and fluctuation range of the ratio. Then, the nonlinear mapping coefficient is called to correct the amplitude-phase offset ratio of each sampling point. Specifically, the nonlinear offset correction amount ΔBi=ki×B_i is calculated for each ratio, where ki is the nonlinear mapping coefficient corresponding to the sampling point and B_i is the amplitude-phase offset ratio of the sampling point. After correction, the amplitude is corrected by subtraction ΔAi′=Ai-ΔBi, and the phase is fine-tuned Δθi′=θi±ΔBi. Depending on the specific requirements, the phase can be fine-tuned in the positive or negative direction. Then, the corrected amplitude and phase are paired according to the sampling index to form a new amplitude-phase correspondence. Next, the deviation of the corrected amplitude-phase signal is statistically analyzed. By calculating the difference between each corrected amplitude and its local mean, and taking the average of the sum of squares, the nonlinear distortion correction coefficient is obtained.

[0097] S503: Based on the nonlinear distortion correction coefficient, calculate the amplitude mean square stability rate of the corrected signal as the ratio of the variance to the mean of the corrected amplitude sequence, calculate the phase correlation as the point-to-point correlation coefficient of the corrected phase sequence, multiply the amplitude mean square stability rate and the phase correlation by the set amplitude and phase weight ratio respectively and sum them, calculate the sum of the stability parameter and the noise threshold difference and normalize it, establish a noise immunity performance index set, and generate noise immunity correction results;

[0098] Based on the nonlinear distortion correction coefficient, the amplitude mean square stability rate of the corrected signal is calculated as M_A = σA′² / μA′, where σA′² is the variance of the corrected amplitude sequence and μA′ is the mean of the corrected amplitude sequence. Next, the correlation of the corrected phase sequence is calculated using the point-to-point correlation coefficient formula C_θ = Corr(θi′, θi′+1). Then, according to the set amplitude and phase weight ratios, the amplitude mean square stability rate and phase correlation are weighted and summed to calculate the total stability parameter S = wA × M_A + wθ × C_θ, where wA and wθ are the amplitude and phase weights, respectively. Finally, the sum of the stability parameter and the noise threshold difference is calculated and normalized to obtain the noise immunity performance index set, and the noise immunity correction result is output.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A deep learning-based HPLC&HRF dual-mode communication adaptive coding and modulation and anti-noise method, characterized in that, The method comprises the following steps: S1: obtaining the amplitude variation rate, phase variation rate and modulation depth of the optical and radio frequency signals, synchronously sampling the signals to calculate the instantaneous amplitude and phase, filtering and normalizing to obtain the signal preprocessing result; S2: calling the signal preprocessing result, calculating the amplitude difference and phase offset between the coaxial cable and the microstrip line according to the amplitude residual and phase residual of the optical and radio frequency signals at different impedance nodes, generating the intermodulation residual term set and normalizing, and back projecting to generate the nonlinear mapping coefficient; S3: monitoring the time delay difference and coherence ratio change according to the signal preprocessing result and the nonlinear mapping coefficient, calculating the time delay difference value and instantaneous coherence ratio, generating the coherence offset parameter set, and differentiating to obtain the coherence drift trend; S4: comparing the modulation depth and carrier amplitude in the optical and radio frequency domains based on the coherence drift trend and the nonlinear mapping coefficient, adjusting the modulation format and coding scheme, and optimizing to generate the adaptive coding modulation scheme; S5: compensating and fine tuning the signal amplitude attenuation and phase disturbance according to the adaptive coding modulation scheme, the coherence drift trend and the nonlinear mapping coefficient, correcting the distortion to calculate the stability, and outputting the anti-noise correction result.

2. The deep learning-based HPLC&HRF dual-mode communication adaptive coding and modulation and anti-noise method according to claim 1, characterized in that, In S1, the signal preprocessing result includes the amplitude feature matrix, the phase feature matrix, the synchronization index table and the normalization template; in S2, the nonlinear mapping coefficient includes the amplitude and phase weight table, the envelope offset vector, the node response matrix and the mapping reference library.

3. The deep learning based HPLC & HRF dual-mode communication adaptive coding modulation and anti-noise method according to claim 1, characterized in that, In S3, the coherence drift trend includes the time delay offset curve, the coherence ratio fluctuation sequence and the trend stability index; In S4, the adaptive coding modulation scheme includes the modulation format list, the coding strategy table and the amplitude equalization parameter group; In S5, the anti-noise correction result includes the amplitude calibration set, the phase calibration set and the stability evaluation label.

4. The deep learning based HPLC & HRF dual-mode communication adaptive coded modulation and anti-noise method according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: obtaining the amplitude variation rate, phase variation rate and modulation depth of the optical and radio frequency signals in the HPLC and HRF link, detecting the voltage amplitude and phase angle change rate of the sampled signal, recording the continuous sample data based on the sampling time sequence, performing point-by-point calculation on the voltage variation and phase angle difference between each sampling point, and generating the instantaneous amplitude and phase data set; S102: based on the instantaneous amplitude and phase data set, calling the amplitude variation rate and phase variation rate, performing positioning judgment on the signal amplitude mutation point and phase jump point, performing sliding weighted smoothing filtering according to the time interval and amplitude and phase difference value of the continuous sampling interval, calculating the amplitude interval mean value and phase interval mean value difference after filtering, performing normalization operation according to the amplitude and phase difference value interval range, and obtaining the amplitude and phase filtering and normalizing matrix; S103: according to the amplitude and phase filtering and normalizing matrix, aligning and comparing the instantaneous amplitude and phase samples of the optical and radio frequency signals according to the sampling time, calculating the amplitude difference interval and phase difference interval of the two signals at the synchronous sampling points, obtaining the correlation ratio according to the difference interval variation rate, and generating the signal preprocessing result.

5. The deep learning based HPLC & HRF dual-mode communication adaptive coded modulation and anti-noise method according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: calling the signal preprocessing result, obtaining a sampling voltage sequence and a phase angle sequence, calculating an adjacent node amplitude difference using a node impedance value, calculating a phase difference component using the phase angle sequence, comparing and recording a difference value interval and a node index point by point, performing normalized labeling on the difference value interval, and generating a node residual information set; S202: based on the node residual information set, obtaining a coaxial cable and microstrip line node mapping table, performing difference operation on the node amplitude difference according to the mapping table, performing subtraction on the phase difference according to the mapping table and calculating the difference value interval change rate, extracting the intermodulation frequency related components in the difference interval and arranging the sequence, classifying and arranging the sequence according to the node attribute and labeling the sequence index, and obtaining an optical radio frequency intermodulation residual term set; S203: according to the optical radio frequency intermodulation residual term set, performing time domain moving average and normalization operation on the residual term sequence, establishing a standardized residual vector and back projecting it to the original signal space according to the sampling node position, comparing the difference interval of the back projection envelope amplitude and the initial sampling envelope, calculating the difference proportion of each sampling point and converting it into a node mapping coefficient, and summarizing the node mapping coefficient to generate a nonlinear mapping coefficient.

6. The deep learning based HPLC&HRF dual-mode communication adaptive coded modulation and anti-noise method according to claim 1, characterized in that, The specific steps of S3 are: S301: according to the signal preprocessing result and the nonlinear mapping coefficient, collecting HPLC sampling point time delay value and coherence ratio value, collecting HRF sampling point time delay value and coherence ratio value, pairing HPLC and HRF data according to the sampling point index, calculating the time delay difference value and recording the difference value sign and index, calculating the coherence ratio difference value and recording the difference interval and index, merging the time delay difference value and the coherence ratio difference value in sequence according to the sampling point order, and generating a time delay difference and coherence ratio difference sequence; S302: based on the time delay difference and coherence ratio difference sequence, difference is obtained by difference between adjacent sampling index pairs to obtain a difference value sequence, a sliding average operation is performed on the difference value sequence by selecting a difference window set, the mean and variance of each window are calculated and the window start and end index are recorded, the difference amplitude of the difference mutation position in the window is compared and the mutation index is labeled, the difference statistics are summarized as a parameter vector and sorted according to the sampling point, a parameter vector sequence is established and a coherence shift parameter set is obtained; S303: according to the coherence shift parameter set, the parameter vector sequence is calculated by discrete differentiation in time sequence to obtain an instantaneous change rate sequence, the change ratio of adjacent points is calculated for the instantaneous change rate sequence and the change ratio index is recorded, the change ratio is aggregated to generate a point level drift, the point level drift is summarized and the sequence mean and trend coefficient are calculated, and a coherence drift trend is obtained.

7. The deep learning based HPLC&HRF dual-mode communication adaptive coded modulation and anti-noise method according to claim 1, characterized in that, The specific steps of S4 are: S401: according to the coherence drift trend and the nonlinear mapping coefficient, detecting the waveguide section and the coupler transmission loss parameter, collecting the optical domain modulation depth and the carrier amplitude value, collecting the radio frequency modulation depth and the carrier amplitude value, pairing the amplitude and modulation depth of the optical domain and the radio frequency domain according to the sampling index, and generating an optical radio frequency modulation contrast coefficient by subtracting the radio frequency amplitude from the optical domain amplitude and dividing the radio frequency depth by the optical domain depth. S402: Based on the light radio frequency modulation contrast coefficient, the arithmetic mean and variance of the amplitude difference value sequence are calculated, the corrected coefficient is synthesized by subtracting the waveguide segment loss rate from the coupler loss rate, the amplitude difference value of each sampling point is subtracted by the corrected coefficient and normalized by the maximum absolute value of the sequence, the point-by-point difference of the amplitude difference value before and after correction is compared and the continuous index interval with the smallest difference is located, and the amplitude difference minimization interval parameter is generated; S403: According to the amplitude difference minimization interval parameter, the nonlinear mapping coefficient is called, the encoding criterion value is calculated according to the interval left and right boundary ratio and the coherent drift trend change rate, the criterion value is classified and mapped to select the encoding scheme and identify the corresponding modulation depth combination, the weighted sum of the criterion value of each sampling segment is formed to form a parameter set, and the adaptive encoding modulation scheme is generated. 8.The deep learning based HPLC&HRF dual-mode communication adaptive coded modulation and anti-noise method of claim 1, wherein, The specific steps of S5 are, S501: According to the adaptive encoding modulation scheme, the coherent drift trend and the nonlinear mapping coefficient, the amplitude value and phase value of the continuous sampling of the transmission signal are detected, the amplitude decay rate of each sampling point is calculated as the difference between the current amplitude and the previous amplitude divided by the sampling interval, the phase difference of adjacent sampling points is calculated as the difference between the current phase and the previous phase divided by the sampling interval, the amplitude-phase offset ratio is calculated as the amplitude decay rate divided by the absolute value of the phase difference, the ratio sequence is output in the sampling order, and the amplitude-phase offset ratio sequence is generated; S502: Based on the amplitude-phase offset ratio sequence, the arithmetic mean and variance of the sequence are calculated, the nonlinear mapping coefficient is called to calculate the nonlinear offset correction amount of each sampling point as the mapping coefficient multiplied by the corresponding ratio, the subtraction correction is performed for the amplitude and the plus-minus fine tuning is performed for the phase, the amplitude-phase corresponding relationship is reconstructed synchronously as the corrected amplitude and the corrected phase are paired according to the sampling index, the deviation degree of the corrected amplitude signal is calculated according to the sampling order as the square sum of the difference between each point corrected amplitude and its local mean value and the nonlinear distortion correction coefficient is obtained; S503: According to the nonlinear distortion correction coefficient, the amplitude mean square stability rate of the corrected signal is calculated as the ratio of the variance to the mean value of the corrected amplitude sequence, the phase correlation degree is calculated as the point-to-point correlation coefficient of the corrected phase sequence, the amplitude mean square stability rate and the phase correlation degree are multiplied and summed according to the set amplitude to phase weight ratio respectively, the sum of the stability parameter and the noise threshold difference is calculated and normalized, the anti-noise performance index set is established, and the anti-noise correction result is generated.

Images