A communication transmission stability optimization method based on time-frequency precision improvement

Abstract

The application discloses a communication transmission stability optimization method based on time-frequency precision improvement, and belongs to the wireless communication field.The method comprises the following steps: S1, collecting time-frequency synchronization error data and channel state data, and constructing a time-frequency error vector and a channel state vector; S2, establishing a time-frequency error state space model, and performing optimal estimation by using an adaptive Kalman filter, wherein the noise covariance is dynamically adjusted according to the channel state; S3, performing forward-looking prediction on the time-frequency drift by using an exponential weighted moving average method, and adaptively adjusting a forgetting factor according to the channel time-varying rate; S4, adopting a fine, standard or fast three-level compensation mode hierarchical response according to a comprehensive channel quality index; and S5, verifying the compensation effect by using a closed-loop feedback and triggering a parameter adjustment mechanism.The application is suitable for different channel conditions, realizes dynamic optimization of time-frequency synchronization precision, improves the communication transmission stability, and is suitable for satellite communication, high-speed mobile communication and unmanned aerial vehicle cluster scenes.

Description

Technical Field

[0001] This invention relates to wireless communication, and more particularly to a method for optimizing communication transmission stability based on improved time-frequency accuracy. Background Technology

[0002] With the commercialization of 5G and the research and development of 6G, high-precision time-frequency synchronization technology has become a key enabling technology for applications such as high-precision positioning, collaborative communication, and low-latency, high-reliability transmission. In scenarios such as satellite communication, high-speed rail communication, and drone swarm communication, time-frequency synchronization faces severe challenges due to the dynamic changes in the transmitting and receiving parties or the transmission medium.

[0003] In the field of time synchronization technology, Shanghai Jiao Tong University proposed an optical fiber time-frequency transfer system and method for enhancing time synchronization accuracy (patent number: CN114285516B, authorized announcement date March 28, 2023). This system determines the time difference through bidirectional transmission via an optical fiber link using master and slave time-frequency units, and achieves sub-nanosecond time synchronization accuracy using a bidirectional time transfer protocol. However, this technical solution is only designed for optical fiber media and does not address Doppler frequency shift and channel time-varying characteristics in wireless communication scenarios, thus it cannot be directly applied to wireless transmission scenarios such as satellite communication.

[0004] Regarding time-frequency synchronization methods, the signal detection and time-frequency synchronization method based on symmetrical linear frequency modulated (LFM) signals proposed by the 54th Research Institute of China Electronics Technology Group Corporation (Patent No.: CN116319209A, Publication Date: June 23, 2023) achieves time-frequency synchronization by transmitting symmetrical LFM signals and utilizing bilinear frequency modulated (BFM) signal correlation detection. This method exhibits good detection performance in low signal-to-noise ratio (SNR) environments, but its fixed threshold decision mechanism prevents dynamic adjustment of synchronization parameters based on channel conditions, leading to a significant decrease in synchronization performance when channel conditions change drastically.

[0005] Regarding the reliability of time-frequency networks, the 54th Research Institute of China Electronics Technology Group Corporation proposed a highly stable and reliable time-frequency network implementation method based on an optical fiber ring network architecture (patent number: CN110958504A, publication date: April 3, 2020). This method improves the reliability of time-frequency networks through ring networking and redundancy backup mechanisms. This method mainly addresses the network topology redundancy problem, ensuring uninterrupted transmission of time-frequency signals through path backup and fault switching, but it does not address the dynamic optimization of the time-frequency synchronization accuracy itself.

[0006] In the field of time-frequency signal modulation transmission, Xi'an University of Posts and Telecommunications proposed a time-frequency signal modulation transmission method and system (patent number: CN120185993A, publication date: June 20, 2025). This method utilizes an FPGA to implement time-frequency signal modulation and employs a fully digital dual-mixer phase detector for phase compensation. While this method demonstrates good performance in suppressing phase noise, its compensation strategy is passive and can only compensate after detecting a phase deviation, lacking the ability to predict time-frequency drift.

[0007] In terms of 5G / 6G non-terrestrial network (NTN) technology, the current main approach to address the Doppler shift problem is a two-layer strategy combining open-loop pre-compensation and closed-loop fine-tuning. Open-loop pre-compensation calculates the Doppler shift based on the ephemeris model and performs pre-compensation, but the accuracy of pre-compensation depends on the accuracy of the ephemeris model, resulting in residual errors within the ephemeris update cycle. The closed-loop fine-tuning cycle is fixed at approximately 10 seconds, making it impossible to dynamically adjust the compensation frequency according to channel quality, leading to insufficient tracking performance during high-speed movement or drastic channel changes.

[0008] In summary, existing technologies suffer from the following main drawbacks: First, time-frequency synchronization accuracy is limited by fixed parameter settings and cannot be adaptively adjusted according to channel conditions; second, they lack the ability to proactively predict time-frequency drift and can only passively respond to deviations that have already occurred; third, the compensation strategy is simplistic and cannot respond in stages according to channel quality; and fourth, the feedback adjustment mechanism is lagging and cannot achieve adaptive dynamic optimization of parameters. Therefore, there is an urgent need for a time-frequency accuracy improvement method that can adaptively sense channel conditions, proactively predict time-frequency drift, and provide tiered response compensation to meet the application requirements of high-performance communication systems. Summary of the Invention

[0009] Purpose of the invention: The purpose of this invention is to provide a communication transmission stability optimization method based on improved time-frequency accuracy, which aims to solve the technical problems in the prior art such as insufficient time-frequency synchronization accuracy, fixed synchronization parameters that cannot be adaptively adjusted, and lack of forward-looking prediction of time-frequency drift.

[0010] Technical solution: A communication transmission stability optimization method based on improved time-frequency accuracy, comprising the following steps:

[0011] S1. Multi-dimensional time-frequency error perception: Collect time-frequency synchronization error data of the communication link, including time deviation, frequency deviation and phase deviation; simultaneously collect channel state data, including signal-to-noise ratio, Doppler frequency shift and channel delay spread; construct time-frequency error vector and channel state vector;

[0012] S2. Adaptive Kalman filter time-frequency error estimation: Establish a state-space model of the time-frequency error, where the state vector contains the time-frequency deviation and its rate of change; Use an adaptive Kalman filter to perform optimal estimation of the time-frequency error, where the process noise covariance is exponentially adjusted according to the degree of signal-to-noise ratio deviation, and the measurement noise covariance is linearly adjusted according to the relative magnitudes of Doppler frequency shift and time delay spread;

[0013] S3. Time-frequency drift prediction based on exponentially weighted moving average: Using historical time-frequency error estimation sequences, the exponentially weighted moving average method is used to predict the time-frequency drift at the next moment. The forgetting factor is adaptively adjusted according to the ratio of the frequency change rate to the reference change rate.

[0014] S4. Multi-level time-frequency compensation based on channel quality: The comprehensive channel quality index is determined based on the normalized weighted fusion result of each parameter in the channel state vector. The compensation strategy is selected based on the comparison with the dynamic threshold: fine compensation is used when it is not lower than the high threshold, standard compensation is used when it is between the high and low thresholds, and fast acquisition is used when it is lower than the low threshold.

[0015] S5. Closed-loop feedback and compensation effect verification: After performing time-frequency compensation, the error improvement rate is calculated based on the ratio of the time-frequency error amplitude before and after compensation. When the error improvement rate is lower than the preset threshold, the adjustment mechanism is triggered to increase the process noise covariance, increase the forgetting factor, and shorten the synchronization cycle. When the compensation effect meets the standard for multiple consecutive times, the gradient recovery strategy is adopted to gradually restore the parameters to the default value.

[0016] Furthermore, in step S1, the time-frequency synchronization error data is acquired through a two-way time transfer protocol, the time deviation is obtained through round-trip time measurement, the frequency deviation is obtained through phase noise measurement, and the phase deviation is obtained through a carrier phase tracking loop; the channel state data is acquired through a channel estimation module, the signal-to-noise ratio is determined by the ratio of pilot signal power to noise power, the Doppler frequency shift is obtained by estimating the phase rotation rate of the pilot signal, and the channel delay spread is obtained by calculating the root mean square of the power delay spectrum.

[0017] Furthermore, in step S2, the process noise covariance is adjusted by multiplying the basic process noise covariance by an exponential factor related to the signal-to-noise ratio deviation. When the current signal-to-noise ratio is lower than the reference signal-to-noise ratio, the exponential factor is greater than one. The larger the signal-to-noise ratio deviation, the larger the exponential factor. The measurement noise covariance is adjusted by superimposing a weighted term of the relative proportion of Doppler frequency shift and a weighted term of the relative proportion of time delay spread on the basic measurement noise covariance. The larger the proportions of the two, the larger the measurement noise covariance.

[0018] Furthermore, in step S3, the forgetting factor adaptively takes a value between the minimum and maximum values ​​based on the frequency change rate. The larger the frequency change rate, the closer the forgetting factor is to the maximum value, and the smaller the frequency change rate, the closer the forgetting factor is to the minimum value. The exponentially weighted moving average method weights and fuses the latest time-frequency error estimate with the previous prediction value according to the forgetting factor. The larger the forgetting factor, the greater the weight of the latest estimate, and the smaller the forgetting factor, the greater the weight of the historical prediction value.

[0019] Furthermore, in step S4, all three compensation modes are implemented using a numerically controlled oscillator: the fine compensation mode calculates the compensation amount based on the predicted value and makes continuous fine adjustments with the minimum compensation step size; the standard compensation mode calculates the compensation amount based on the current estimated value and makes step compensation with five times the minimum compensation step size; and the fast acquisition mode makes large step compensation with twenty-five times the minimum compensation step size and simultaneously shortens the synchronization period to one-quarter of the normal period to accelerate synchronization convergence.

[0020] Furthermore, the dynamic threshold in step S4 is adaptively adjusted based on the statistical characteristics of the recent comprehensive channel quality index. The high threshold is set as the sum of the mean and standard deviation of the recent comprehensive channel quality index values, and the low threshold is set as the difference between the mean and standard deviation, so that the threshold dynamically floats with the overall trend of channel state changes.

[0021] Furthermore, the gradient recovery strategy in step S5 is activated after the compensation effect reaches the target number of times consecutively. The process noise covariance, forgetting factor and synchronization period are gradually restored to the default value according to the preset gradient. Each recovery amplitude does not exceed 20% of the current value. The compensation effect is continuously monitored during the recovery process. If the compensation effect does not meet the target, the recovery is immediately terminated and the adjustment mechanism is retried.

[0022] A communication transmission system includes a transmitter, a receiver, and a time-frequency synchronization module. The time-frequency synchronization module includes: a time-frequency error sensing unit for performing error sensing in the above method; a time-frequency error estimation unit for performing adaptive Kalman filter estimation in the above method; a time-frequency drift prediction unit for performing drift prediction in the above method; a multi-level time-frequency compensation unit for performing multi-level time-frequency compensation in the above method; and a closed-loop feedback verification unit for performing closed-loop feedback verification in the above method.

[0023] Furthermore, the time-frequency synchronization module also includes a parameter adaptive adjustment unit, which is used to dynamically adjust the process noise covariance, measurement noise covariance, and forgetting factor of the exponentially weighted moving average of the adaptive Kalman filter according to the error improvement rate output by the closed-loop feedback verification unit. When the error improvement rate is lower than a preset threshold, the above parameters are increased to enhance the system's ability to track channel changes. When the compensation effect meets the standard for multiple consecutive times, the parameters are gradually restored to the default value according to the gradient recovery strategy.

[0024] Beneficial Effects: Optimal estimation of time-frequency error is achieved through an adaptive Kalman filter. The process noise covariance and measurement noise covariance are dynamically adjusted according to channel conditions, adapting to estimation requirements under different channel conditions and maintaining good estimation accuracy even under low signal-to-noise ratio, high Doppler, and high delay spread conditions. A time-frequency drift prediction method based on exponentially weighted moving averages enables forward-looking prediction of time-frequency drift, transforming compensation from a passive response to active prevention. Adaptive adjustment of the forgetting factor ensures good performance of the prediction model under both rapidly and slowly changing channel conditions. A multi-level time-frequency compensation strategy based on channel quality, with graded responses according to comprehensive channel quality indicators, employs three compensation modes: fine, standard, and fast, balancing accuracy and speed. A closed-loop feedback and compensation effect verification mechanism enables dynamic adaptive adjustment of parameters. Automatic adjustment is triggered when compensation effect is unsatisfactory, and gradient restoration to default values ​​is achieved when continuous targets are met, ensuring system stability and robustness. Attached Figure Description

[0025] Figure 1 This is an overall flowchart of a communication transmission stability optimization method based on improving time-frequency accuracy;

[0026] Figure 2 This is a diagram of the adaptive Kalman filter time-frequency error estimation and drift prediction structure;

[0027] Figure 3 This is a flowchart of a multi-level time-frequency compensation decision-making process based on channel quality. Detailed Implementation

[0028] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Example 1

[0030] This embodiment applies to a low-Earth orbit satellite communication scenario, where there is a significant relative velocity between the satellite and the ground terminal, resulting in substantial Doppler frequency shift and time-frequency drift. For example... Figure 1 As shown, the entire processing flow of this method is as follows.

[0031] In step 1, the ground terminal collects time-frequency synchronization error data of the communication link. The time deviation is measured using a two-way satellite time-frequency transfer method, resulting in 150 picoseconds; the frequency deviation is obtained through phase noise measurement, resulting in 2.3 kHz; and the phase deviation is obtained through carrier phase tracking, resulting in 0.15 radians. Simultaneously, channel state data is collected: signal-to-noise ratio of 12 dB, Doppler frequency shift of 45 kHz, and channel delay spread of 8 nanoseconds. These data are then used to construct the time-frequency error vector and the channel state vector, respectively.

[0032] In step 2, a state-space model of the time-frequency error is established, such as... Figure 2 The diagram shows the overall structure of the adaptive Kalman filter for time-frequency error estimation and drift prediction. The state vector comprises six components: time-frequency deviation and its rate of change. The reference signal-to-noise ratio (SNR) is set to 15 dB, the maximum Doppler shift to 100 kHz, and the maximum time delay spread to 20 nanoseconds. Given that the current SNR is lower than the reference SNR, the process noise covariance is adjusted to 1.105 times the baseline value to broaden the uncertainty range of the state estimation; based on the relative magnitudes of the Doppler shift and time delay spread, the measurement noise covariance is adjusted to 1.215 times the baseline value to reduce dependence on unreliable measurements. After adaptive Kalman filtering estimation, the estimated time-frequency errors are: time deviation estimated at 148 picoseconds, frequency deviation estimated at 2.28 kHz, and phase deviation estimated at 0.148 radians.

[0033] In step 3, the forgetting factor is set to a minimum of 0.3 and a maximum of 0.9. The current frequency change rate is 0.8 kHz, and the forgetting factor, calculated through adaptive adjustment, is approximately 0.702. Due to the moderate frequency change rate, the forgetting factor is positioned in the middle to high range, balancing tracking speed and smoothness. The exponentially weighted moving average method is used to predict the time-frequency drift at the next moment: the predicted time deviation is 147.4 picoseconds, and the predicted frequency deviation is 2.24 kHz.

[0034] In step 4, the signal-to-noise ratio (SNR) weight is set to 0.4, the Doppler shift weight to 0.35, and the delay spread weight to 0.25. The upper limit for SNR normalization is 30 dB. Figure 3 The diagram illustrates the multi-level time-frequency compensation decision-making process based on channel quality. After normalized weighted fusion calculation, the comprehensive channel quality index is 0.5025. Based on the recent 10 index values, the mean is 0.48 and the standard deviation is 0.05, with a high threshold of 0.53 and a low threshold of 0.43. Since the comprehensive channel quality index of 0.5025 falls between the low and high thresholds, a standard compensation mode is adopted. The compensation amount is calculated based on the current estimate, and a numerically controlled oscillator is used for step compensation at five times the minimum compensation step size, set to 0.5 kHz.

[0035] In step 5, after compensation, time-frequency error data is re-acquired. Before compensation, the time-frequency error amplitude was 150 picoseconds, and after compensation, it was 18 picoseconds, with an error improvement rate of 88%, which is higher than the preset threshold of 60%, thus the compensation effect is deemed satisfactory. After three consecutive satisfactory compensation results, each parameter is gradually restored to its default value according to the gradient recovery strategy, with each restoration increment being 20% ​​of the current value.

[0036] Example 2

[0037] This embodiment is applied to a high-speed rail communication scenario, where the high speed of the train causes time-varying Doppler frequency shift, and the channel conditions change rapidly with the scenario.

[0038] In step 1, the vehicle-mounted terminal collects time-frequency synchronization error data: time deviation 85 picoseconds, frequency deviation 1.2 kHz, and phase deviation 0.08 radians. Channel status data: signal-to-noise ratio 18 dB, Doppler shift 28 kHz, and channel delay spread 5 nanoseconds.

[0039] In step 2, the reference signal-to-noise ratio is set to 20 dB, the maximum Doppler frequency shift to 80 kHz, and the maximum time delay spread to 15 nanoseconds. The current signal-to-noise ratio is close to the reference value, and the process noise covariance is only slightly adjusted to 1.041 times the reference value; the measurement noise covariance is adjusted to 1.138 times the reference value. After adaptive Kalman filtering estimation, the time deviation is estimated to be 83 picoseconds, the frequency deviation to be 1.18 kHz, and the phase deviation to be 0.079 radians.

[0040] In step 3, the frequency change rate is 1.2 kHz, and the forgetting factor is adaptively adjusted to approximately 0.494. The larger frequency change rate results in a larger forgetting factor to accelerate tracking. Prediction results: predicted time deviation is 84.2 picoseconds, and predicted frequency deviation is 1.19 kHz.

[0041] In step 4, the signal-to-noise ratio (SNR) weight is set to 0.45, the Doppler shift weight to 0.30, and the delay spread weight to 0.25, with an upper limit of 25 dB for SNR normalization. The calculated overall channel quality index is approximately 0.686, with a recent mean of 0.65, a standard deviation of 0.04, a high threshold of 0.69, and a low threshold of 0.61. Since the index of 0.686 falls between the high and low thresholds, the standard compensation mode is adopted.

[0042] In step 5, the error improvement rate after compensation was 85%, which is higher than the preset threshold of 60%, and the compensation effect met the standard.

[0043] Example 3

[0044] This embodiment is applied to a drone swarm communication scenario where multiple drones are in relative motion and the channel conditions are complex and variable.

[0045] In step 1, time-frequency error data is collected: time deviation 120 picoseconds, frequency deviation 1.8 kHz, and phase deviation 0.11 radians. Channel conditions: signal-to-noise ratio 15 dB, Doppler shift 35 kHz, and channel delay spread 6 nanoseconds.

[0046] In step 2, the reference signal-to-noise ratio is set to 18 dB, the maximum Doppler frequency shift to 90 kHz, and the maximum time delay spread to 18 nanoseconds. The process noise covariance is adjusted to 1.12 times the reference value, and the measurement noise covariance is adjusted to 1.21 times the reference value. Estimation results: time deviation estimated at 118 picoseconds, frequency deviation estimated at 1.76 kHz, and phase deviation estimated at 0.108 radians.

[0047] In step 3, the frequency change rate is 0.6 kHz, and the forgetting factor is adaptively adjusted to approximately 0.696. Prediction results: predicted time deviation of 119.5 picoseconds and predicted frequency deviation of 1.77 kHz.

[0048] In step 4, the signal-to-noise ratio (SNR) weight is set to 0.42, the Doppler shift weight to 0.33, and the delay spread weight to 0.25, with a normalized SNR upper limit of 28 dB. The calculated overall channel quality index is approximately 0.594, with a recent mean of 0.57, a standard deviation of 0.035, a high threshold of 0.605, and a low threshold of 0.535. Since the index of 0.594 falls between the high and low thresholds, a standard compensation mode is adopted.

[0049] In step 5, the compensation effect met the standard, the error improvement rate was 82%, and the parameters recovered after continuous compliance.

[0050] Example 4

[0051] This embodiment is applied to scenarios where channel conditions change drastically, such as satellite transit handovers or high-speed aircraft traversing different airspaces.

[0052] In step 1, data was collected after the channel conditions deteriorated: the signal-to-noise ratio dropped to 8 dB, the Doppler shift increased to 72 kHz, and the channel delay spread increased to 14 nanoseconds. Time-frequency errors: time deviation 280 picoseconds, frequency deviation 4.5 kHz, and phase deviation 0.32 radians.

[0053] In step 2, under low signal-to-noise ratio conditions, the process noise covariance is adjusted to 1.28 times the baseline value, and the measurement noise covariance is adjusted to 1.52 times the baseline value, significantly reducing the dependence on the measurement values. Estimation results: time deviation estimated at 275 picoseconds, frequency deviation estimated at 4.4 kHz, and phase deviation estimated at 0.31 radians.

[0054] In step 3, the frequency change rate is 2.1 kHz, much larger than the reference change rate. The forgetting factor is adaptively adjusted to approximately 0.51, close to the median but biased towards the larger side to accelerate the tracking speed. Prediction results: predicted time deviation is 278 picoseconds, and predicted frequency deviation is 4.42 kHz.

[0055] In step 4, the calculated comprehensive channel quality index is only 0.28, with a recent mean of 0.42, a standard deviation of 0.06, a high threshold of 0.48, and a low threshold of 0.36. Since the index of 0.28 is lower than the low threshold, the channel conditions are deemed poor, and a fast acquisition mode is adopted. A large step size compensation is performed with a step size 25 times the minimum compensation step size, while the synchronization period is shortened to one-quarter of the normal period to accelerate synchronization convergence.

[0056] In step 5, after performing fast capture compensation, the error improvement rate is 45%, which is lower than the preset threshold of 60%, triggering the adjustment mechanism: increasing the process noise covariance to 1.5 times the current value, increasing the forgetting factor to 0.65, and shortening the synchronization period to half of the current value. After re-executing the compensation process, the error improvement rate reaches 62%, which is higher than the preset threshold, and the compensation effect meets the target.

[0057] As can be seen from the above embodiments, this invention achieves comprehensive optimization of time-frequency accuracy through five steps: multi-dimensional time-frequency error perception, adaptive Kalman filter estimation, exponentially weighted moving average drift prediction, multi-level time-frequency compensation, and closed-loop feedback verification. Fine compensation ensures accuracy under good channel conditions, standard compensation balances accuracy and speed under average channel conditions, and fast acquisition prioritizes synchronization recovery under poor channel conditions. Combined with adaptive parameter adjustment and a closed-loop feedback mechanism, it can adapt to the time-frequency synchronization requirements under different channel conditions, significantly improving the stability of communication transmission.

[0058] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A communication transmission stability optimization method based on improved time-frequency accuracy, characterized in that, Includes the following steps: S1. Multi-dimensional time-frequency error perception: Collect time-frequency synchronization error data of the communication link, including time deviation, frequency deviation and phase deviation; simultaneously collect channel state data, including signal-to-noise ratio, Doppler frequency shift and channel delay spread; construct time-frequency error vector and channel state vector; S2. Adaptive Kalman filter time-frequency error estimation: Establish a state-space model of the time-frequency error, where the state vector contains the time-frequency deviation and its rate of change; Use an adaptive Kalman filter to perform optimal estimation of the time-frequency error, where the process noise covariance is exponentially adjusted according to the degree of signal-to-noise ratio deviation, and the measurement noise covariance is linearly adjusted according to the relative magnitudes of Doppler frequency shift and time delay spread; S3. Time-frequency drift prediction based on exponentially weighted moving average: Using historical time-frequency error estimation sequences, the exponentially weighted moving average method is used to predict the time-frequency drift at the next moment. The forgetting factor is adaptively adjusted according to the ratio of the frequency change rate to the reference change rate. S4. Multi-level time-frequency compensation based on channel quality: The comprehensive channel quality index is determined based on the normalized weighted fusion result of each parameter in the channel state vector. The compensation strategy is selected based on the comparison with the dynamic threshold: fine compensation is used when it is not lower than the high threshold, standard compensation is used when it is between the high and low thresholds, and fast acquisition is used when it is lower than the low threshold. S5. Closed-loop feedback and compensation effect verification: After performing time-frequency compensation, the error improvement rate is calculated based on the ratio of the time-frequency error amplitude before and after compensation. When the error improvement rate is lower than the preset threshold, the adjustment mechanism is triggered to increase the process noise covariance, increase the forgetting factor, and shorten the synchronization cycle. When the compensation effect meets the standard for multiple consecutive times, the gradient recovery strategy is adopted to gradually restore the parameters to the default value.

2. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, In step S1, the time-frequency synchronization error data is acquired through a two-way time transfer protocol, the time deviation is obtained by round-trip time measurement, the frequency deviation is obtained by phase noise measurement, and the phase deviation is obtained through a carrier phase tracking loop. The channel state data is acquired through a channel estimation module, the signal-to-noise ratio is determined by the ratio of pilot signal power to noise power, the Doppler frequency shift is obtained by estimating the phase rotation rate of the pilot signal, and the channel delay spread is obtained by calculating the root mean square of the power delay spectrum.

3. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, In step S2, the process noise covariance is adjusted by multiplying the basic process noise covariance by an exponential factor related to the signal-to-noise ratio deviation. When the current signal-to-noise ratio is lower than the reference signal-to-noise ratio, the exponential factor is greater than one. The larger the signal-to-noise ratio deviation, the larger the exponential factor. The measurement noise covariance is adjusted by superimposing a weighted term of the relative proportion of Doppler frequency shift and a weighted term of the relative proportion of time delay spread on the basic measurement noise covariance. The larger the proportions of the two, the larger the measurement noise covariance.

4. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, In step S3, the forgetting factor adaptively takes a value between the minimum and maximum values ​​based on the frequency change rate. The larger the frequency change rate, the closer the forgetting factor is to the maximum value, and the smaller the frequency change rate, the closer the forgetting factor is to the minimum value. The exponentially weighted moving average method weights and fuses the latest time-frequency error estimate with the previous prediction value according to the forgetting factor. The larger the forgetting factor, the greater the weight of the latest estimate, and the smaller the forgetting factor, the greater the weight of the historical prediction value.

5. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, In step S4, all three compensation modes are implemented by a numerically controlled oscillator: the fine compensation mode calculates the compensation amount based on the predicted value and makes continuous fine adjustments with the minimum compensation step size. The standard compensation model calculates the compensation amount based on the current estimate and performs step compensation in increments of five times the minimum compensation step size; The fast capture mode performs large step compensation with a minimum compensation step size of 25 times and at the same time shortens the synchronization period to one-quarter of the normal period to accelerate synchronization convergence.

6. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, The dynamic threshold in step S4 is adaptively adjusted based on the statistical characteristics of recent integrated channel quality indicators. The high threshold is set as the sum of the mean and standard deviation of the recent integrated channel quality indicator values, and the low threshold is set as the difference between the mean and standard deviation, so that the threshold dynamically floats with the overall trend of channel state changes.

7. The communication transmission stability optimization method based on improved time-frequency accuracy according to claim 1, characterized in that, The gradient recovery strategy in step S5 is activated after the compensation effect reaches the target number of times. The process noise covariance, forgetting factor and synchronization period are gradually restored to the default value according to the preset gradient. Each recovery amplitude does not exceed 20% of the current value. The compensation effect is continuously monitored during the recovery process. If the compensation effect does not meet the target, the recovery is immediately terminated and the adjustment mechanism is retried.

8. A communication transmission system, comprising a transmitter, a receiver, and a time-frequency synchronization module, characterized in that, The time-frequency synchronization module includes: a time-frequency error sensing unit for performing error sensing as described in claims 1 and 2; a time-frequency error estimation unit for performing adaptive Kalman filter estimation as described in claims 1 and 3; a time-frequency drift prediction unit for performing drift prediction as described in claims 1 and 4; a multi-level time-frequency compensation unit for performing multi-level time-frequency compensation as described in claims 1, 5, and 6; and a closed-loop feedback verification unit for performing closed-loop feedback verification as described in claims 1 and 7.

9. The communication system according to claim 8, characterized in that, The time-frequency synchronization module also includes a parameter adaptive adjustment unit, which is used to dynamically adjust the process noise covariance, measurement noise covariance and forgetting factor of the adaptive Kalman filter and the exponentially weighted moving average according to the error improvement rate output by the closed-loop feedback verification unit. When the error improvement rate is lower than a preset threshold, the above parameters are increased to enhance the system's ability to track channel changes. When the compensation effect meets the standard for multiple consecutive times, the parameters are gradually restored to the default value according to the gradient recovery strategy.

Images