Shortwave amc method based on adaptive quantization and fsmc prediction

Abstract

This invention provides a shortwave AMC method based on adaptive reduction quantization and FSMC prediction. The technical solution includes: obtaining the signal-to-noise ratio (SNR) sequence of the shortwave ionospheric channel; employing a model structure adaptive reduction method to select the optimal SNR sequence to fit the distribution model; using a joint quantization cost function of mean square error and entropy constraint vector to iteratively optimize the SNR sequence using non-uniform vector quantization; calculating the FSMC state transition probability matrix of the SNR; using the FSMC state transition probability matrix of the SNR to predict the channel state at the next frame transmission time, and selecting the MCS that maximizes throughput as the final transmission scheme. This invention balances the effectiveness and implementation complexity of shortwave AMC methods, significantly improving the adaptive capability of shortwave AMC and enabling dynamic optimal allocation of shortwave links.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, and more specifically relates to a shortwave adaptive modulation and coding (AMC) method based on adaptive reduction quantization and finite state Markov chain (FSMC) prediction in the field of electronic data processing technology. This invention balances the effectiveness and implementation complexity of shortwave AMC methods, significantly improving the adaptive adaptability of shortwave AMC, enabling dynamic optimal allocation of shortwave links, and can be used for shortwave AMC data transmission in scenarios such as emergency rescue communication, beyond-line-of-sight tactical communication, and anti-jamming data transmission. Background Technology

[0002] Shortwave communication utilizes ionospheric reflection for beyond-line-of-sight transmission, a core method for building resilient communication networks. However, ionospheric channels are characterized by strong time-varying properties, significant multipath effects, and non-Gaussian noise interference. To improve communication reliability and effectiveness under harsh channel conditions, modern shortwave communication standards generally employ long-frame transmission mechanisms. This long-frame structure inherently conflicts with the highly dynamic and time-varying characteristics of shortwave channels, placing more stringent demands on shortwave AMC (Adaptive Channel Conditioning) technology. However, existing shortwave AMC methods struggle to achieve accurate channel state prediction under shortwave long-frame transmission structures. Traditional shortwave AMC methods primarily employ an open-loop control mechanism of measurement, feedback, and table lookup. The receiver quantifies the channel quality indication from the received signal and feeds it back to the transmitter, which then directly uses this feedback to consult a pre-defined mapping table to determine the transmission strategy for the next frame. This method essentially relies on the assumption that the channel remains stable for a short time within the feedback delay. However, under the long feedback delay introduced by the short-wavelength frame transmission structure, the channel state often changes significantly by the time the feedback information reaches the transmitting end, leading to a mismatch between the transmission strategy and the actual channel state, and severely degrading the communication performance.

[0003] Beijing Rongwei Technology Co., Ltd. disclosed a shortwave AMC method based on a preset fixed signal-to-noise ratio (SNR) threshold in its patent application "Adaptive Coding and Modulation Protocol Communication Method Based on Half-Duplex Communication Mode" (application number: 202411649246.1, publication number: CN 119172800 A). This method includes: acquiring the instantaneous SNR of the signal at the receiving end; comparing the obtained instantaneous SNR with a set of preset discrete switching SNR threshold values; and selecting the corresponding modulation and coding scheme (MCS) based on the threshold interval into which the SNR falls, using it as the result of the shortwave AMC and directly applying it to the transmission of the next frame of data. This method achieves a low-complexity shortwave AMC switching mechanism through preset fixed thresholds and table lookup mapping. However, this method still has shortcomings. Its static SNR interval quantization strategy is difficult to match the highly non-uniform statistical distribution of the shortwave ionospheric SNR, resulting in large quantization distortion in probability-dense intervals and resource waste in sparse intervals. Furthermore, this method is a decision based on the current instantaneous signal-to-noise ratio, ignoring the impact of long feedback delay. When facing short-wavelength frame transmission structures, the selected AMC scheme does not match the current link state, resulting in a sharp decline in transmission performance.

[0004] Dalian University disclosed a shortwave AMC (Automatic Signal-to-Noise Ratio) method combining a Long Short-Term Memory (LSTM) network and maximum likelihood estimation in its patent application "Coded Modulation Switching Method Based on Likelihood Estimation to Correct Signal-to-Noise Ratio" (Application No. 202011615141.6, Publication No. CN 112713966 A). This method includes: acquiring the channel signal-to-noise ratio (SNR) sequence at the initial stage of the shortwave channel and using it as input to an LSTM network to train the network to output the SNR value at any time during link communication; constructing a likelihood function for the error coefficient between the reference SNR and the decision threshold based on the probability distribution characteristics of the shortwave channel SNR, and solving for this error coefficient using a maximum likelihood estimation algorithm; substituting the error coefficient into the reference SNR correction formula to obtain the improved reference SNR value; dividing the improved SNR into several intervals according to different MCS (Multi-Signal Component Scheme) schemes, and selecting the corresponding MCS scheme for data transmission by determining the interval to which the current reference SNR belongs. This method optimizes the switching accuracy of shortwave AMC by fusing timing prediction and maximum likelihood statistical estimation. However, this method has limitations. It primarily focuses on correcting errors in the instantaneous signal-to-noise ratio (SNR) estimate, essentially employing a local optimization strategy for shortwave AMC based on threshold comparison. While SNR accuracy can be improved through prediction and likelihood estimation, its decision-making mechanism still relies on a static, pre-defined SNR interval division. It fails to globally model the overall statistical characteristics and dynamic evolution of the shortwave channel's SNR, resulting in insufficient accuracy in depicting SNR evolution and limited performance improvement for shortwave AMC. Furthermore, this method struggles to adapt to the highly non-stationary and non-uniform fading characteristics of shortwave channels. In short-wavelength feedback delay scenarios, it cannot address the transmission scheme mismatch caused by channel state changes, leading to poor performance of the selected AMC scheme in the link. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the prior art by providing a shortwave AMC method based on adaptive reduction quantization and FSMC prediction. This method aims to solve the problems of signal-to-noise ratio sequence fitting distribution model failure, insufficient accuracy of vector quantization, and link matching failure caused by feedback loop lag when the existing shortwave AMC technology is used for short-wavelength frame transmission structures.

[0006] The technical approach to achieving the objective of this invention is as follows: This invention employs a model structure adaptive reduction criterion fitting analysis and an iterative optimal non-uniform vector quantization mechanism combining mean square error (MSE) and entropy constraints. When constructing the channel state model, this mechanism balances model effectiveness and complexity through the model structure adaptive reduction criterion, adaptively selecting the optimal fitting distribution model based on the average reduction score. Furthermore, by introducing iterative optimal non-uniform vector quantization with joint MSE and entropy constraints, and utilizing the Lagrange multiplier method to construct a joint quantization cost function for MSE and entropy, it densifies partitions in regions with high SNR frequencies and sparsely partitions them in edge regions. This technique features a synergistic optimization balance between the goodness of fit, complexity, and quantization accuracy. This solves the technical problems in existing shortwave AMC technology where fixed-fit distribution models fail due to overfitting or underfitting during channel abrupt changes, leading to model failure and resource waste, and fixed SNR state partitioning results in quantization accuracy imbalance. This invention also employs a channel state prediction technique based on the SNR-based FSMC state transition probability matrix. This technology constructs a signal-to-noise ratio (SNR) FSMC state transition probability matrix, which describes the dynamic evolution of the channel state over time, by statistically analyzing the SNR of historical channel data and combining it with channel physical parameters such as Doppler shift and Doppler spread. Finally, after receiving the SNR representing channel quality, the transmitting end uses this state transition probability matrix to deduce the channel state at the data transmission moment and selects the MCS scheme that maximizes throughput for data transmission. This technology has the substantial characteristic of predicting the channel state at future transmission moments using the SNR FSMC state transition probability matrix. This solves the transmission scheme mismatch problem caused by feedback lag in shortwave communication due to the long frame transmission structure, which forces the transmitting end to make decisions based on outdated feedback information, ultimately achieving dynamic optimal link allocation.

[0007] Based on the above ideas, the technical solution of the present invention includes:

[0008] Step 1: Extract the signal-to-noise ratio sequence of the channel from the channel measurement data of the shortwave ionospheric channel scenario;

[0009] Step 2: Perform adaptive reduction criterion fitting analysis on the extracted signal-to-noise ratio sequence. Based on the reduction results, balance the effectiveness and complexity of the model, and select the fitting distribution model with the largest reduction value that best matches the extracted signal-to-noise ratio sequence.

[0010] Step 3: The extracted signal-to-noise ratio sequence is optimized by iterative optimal non-uniform vector quantization under the joint constraints of mean square error and entropy. This achieves encrypted partitioning in regions where the signal-to-noise ratio occurs frequently, and sparse partitioning in edge regions, until the cost function converges, resulting in the optimized signal-to-noise ratio state partitioning.

[0011] Step 4: Calculate the FSMC state transition probability matrix for signal-to-noise ratio;

[0012] Step 5: During the AMC selection process, after receiving the channel quality indication in real time, the transmitting end predicts the channel state at the time of transmission of the next frame from the FSMC state transition probability matrix of the signal-to-noise ratio, and selects the MCS scheme that maximizes the throughput to achieve dynamic optimal allocation of the link.

[0013] Furthermore, the step of extracting the signal-to-noise ratio sequence of the channel is as follows:

[0014] The first step is for the receiver to use the probe sequence to calculate the channel impulse response in the time domain by convolution, obtaining at least 1000 time-continuous channel impulse responses;

[0015] The second step is to set the response with a maximum power of 75% in the channel impulse response as the noise threshold, and the power less than the noise threshold is recorded as noise, and the power greater than the noise threshold is recorded as useful signal. The instantaneous signal-to-noise ratio in each channel impulse response window is calculated by the power ratio of useful signal to noise, and the instantaneous signal-to-noise ratio representing the channel state is obtained.

[0016] The third step is to store the instantaneous signal-to-noise ratio in the observation buffer in chronological order to obtain a signal-to-noise ratio sequence that reflects the fluctuations of the ionospheric channel over time.

[0017] Furthermore, the steps of the adaptive reduction criterion fitting analysis for the model structure are as follows:

[0018] The first step is to select at least four candidate probability distribution models, namely Rayleigh distribution, Rice distribution, Nakagami distribution, and log-normal distribution. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) reduction criteria are used to calculate the reduction value of each candidate distribution model.

[0019] ;

[0020] in, , Let AIC and BIC be the reduction criteria for the i-th candidate distribution model, respectively. This represents the total number of sequences with signal-to-noise ratio. This represents the logarithmic operation with the natural constant e as the base. This represents the maximum likelihood estimate of the probability density function of the i-th candidate distribution model. This represents the signal-to-noise ratio sequence of the nth generation. This represents the total number of parameters in the candidate distribution model;

[0021] The second step is to normalize the reduced values ​​of AIC and BIC for each candidate distribution model to obtain the normalized weights corresponding to each candidate distribution model:

[0022] ;

[0023] in, This represents the normalized weight value of the i-th candidate distribution model. This represents an exponential function with base e. This represents the difference between the reduced value of the i-th candidate distribution model and the minimum reduced value. This represents the total number of candidate distribution models. This represents the difference between the reduced value and the minimum reduced value of the j-th candidate distribution model;

[0024] The third step is to calculate the average value of the normalized weights of AIC and BIC for each candidate distribution model, and take the candidate distribution model with the largest average weight as the fitted distribution model of the signal-to-noise ratio sequence.

[0025] Furthermore, the balance between model effectiveness and complexity obtained from the specification means that when calculating the specification values ​​of each candidate distribution model, the AIC and BIC specification criteria are introduced, with the AIC specification value constraining the model effectiveness and the BIC specification value constraining the model complexity.

[0026] Furthermore, the steps for determining the best-fit distribution model are as follows:

[0027] The first step is to calculate the AIC normalized weights and BIC normalized weights for each candidate distribution model;

[0028] The second step is to calculate the average value of the normalized weights of AIC and BIC for each candidate distribution model;

[0029] The third step is to select the candidate distribution model with the largest average reduction value of AIC and BIC as the best-fit distribution model that matches the signal-to-noise ratio sequence.

[0030] Furthermore, the steps of the iterative optimal non-uniform vector quantization optimization under the joint constraints of mean square error and entropy are as follows:

[0031] The first step is to establish a joint quantization cost function for mean squared error and entropy;

[0032] The second step is to set the initial signal-to-noise ratio partition boundaries that satisfy the monotonically increasing constraint conditions.

[0033] The third step involves iteratively adjusting the positions of the signal-to-noise ratio partition boundaries in a monotonically increasing manner, with the goal of achieving cost function convergence, until the cost function converges. After convergence, the final signal-to-noise ratio (SNR) interval boundary is obtained, completing the optimization of the SNR state partition.

[0034] The cost function The expression is as follows:

[0035] ;

[0036] in, This represents the boundary signal-to-noise ratio (SNR) value of the m-th SNR state partition, where M represents the total number of SNR state partitions. This represents the total number of sequences with signal-to-noise ratio. This represents the number of signal-to-noise ratio (SNR) sequences within the m-th SNR state partition. This represents the nth signal-to-noise ratio sequence in the mth signal-to-noise ratio state partition. This represents the center signal-to-noise ratio (SNR) of the m-th SNR state partition. This indicates the weighting factor, and its value is 0.5. This represents a logarithmic operation with base 2. This represents the probability that the signal-to-noise ratio sequence falls into the m-th signal-to-noise ratio state partition. These represent the maximum and minimum signal-to-noise ratio values ​​in the signal-to-noise ratio sequence, respectively.

[0037] Furthermore, the FSMC state transition probability matrix of the signal-to-noise ratio is obtained by the following formula:

[0038] ;

[0039] in, This represents the state transition probability from the c-th signal-to-noise ratio (SNR) state partition to the f-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the c-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the f-th SNR state partition. This represents the bivariate joint probability density function of the fitted distribution model.

[0040] Furthermore, the step of predicting the channel state at the time of transmission of the next frame is as follows:

[0041] The first step is that after receiving the signal-to-noise ratio (SNR) sequence reflecting the current channel state, the transmitting end determines the SNR state partition in which the current channel state is located.

[0042] The second step is to calculate the probability of occurrence of each SNR state partition at the time of transmission of the next frame from the FSMC state transition probability matrix of the signal-to-noise ratio, so as to reflect the channel state.

[0043] Furthermore, the steps for selecting the MCS scheme that maximizes throughput are as follows:

[0044] The first step is to preset a number of MCS schemes that is greater than the total number of signal-to-noise ratio (SNR) state partitions. Each MCS scheme has a different transmission rate in different SNR state partitions.

[0045] The second step is to calculate the average throughput of each MCS scheme at the next frame transmission time using the following formula:

[0046] ;

[0047] in, This represents the average throughput of the o-th MCS scheme at the time of transmission of the next frame. This represents the throughput obtained by selecting the o-th MCS scheme when the signal-to-noise ratio (SNR) is in the m-th SNR state partition. This represents the probability that the signal-to-noise ratio of the next frame falls into the m-th signal-to-noise ratio state partition;

[0048] The third step is to compare the average throughput of all MCS schemes and select the MCS scheme with the highest average throughput as the transmission scheme for the next frame of data.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] First, this invention employs an optimization mechanism based on model structure adaptive reduction criterion fitting analysis and iterative optimal non-uniform vector quantization combined with mean square error and entropy constraints. This overcomes the technical challenges of existing technologies where fixed signal-to-noise ratio (SNR) sequence fitting models fail to fit due to overfitting or underfitting during channel abrupt changes, leading to resource waste, and where fixed SNR state partitions result in insufficient quantization accuracy. This invention simultaneously considers multiple parameters such as model fitting distribution goodness, complexity, and quantization accuracy, significantly improving the accuracy and adaptability of the SNR sequence fitting distribution model in the shortwave AMC method, enhancing the quantization accuracy of the SNR state partitions, and allowing the MCS selection of the shortwave AMC strategy to better fit the actual SNR state partitions of the shortwave channel, effectively improving the throughput and transmission efficiency of the shortwave AMC method.

[0051] Second, this invention employs a channel state prediction technique based on the signal-to-noise ratio (SNR) FSMC state transition probability matrix. This overcomes the transmission scheme mismatch problem caused by feedback lag in existing shortwave AMC techniques for short-wavelength frame transmission structures, leading to decisions made by the transmitter based on outdated feedback information. This invention combines channel physical parameters and utilizes the dynamically evolving SNR FSMC state transition probability matrix to predict the MCS scheme for future transmission times. This significantly improves the link reliability and transmission efficiency of the shortwave AMC method in long-frame transmission structures, avoiding problems such as MCS scheme mismatch and link interruption caused by feedback lag. It ensures that the selected shortwave MCS scheme adapts to future channel states, ultimately achieving dynamic optimal link allocation. Attached Figure Description

[0052] Figure 1 This is a flowchart of the present invention;

[0053] Figure 2 This is a flowchart of the adaptive reduction criterion fitting analysis of the present invention;

[0054] Figure 3 This is a flowchart illustrating the FSMC state transition probability matrix for calculating the signal-to-noise ratio in this invention.

[0055] Figure 4 The figures show simulation results of the shortwave AMC method proposed in this invention and existing shortwave AMC methods. Detailed implementation details.

[0056] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort should all fall within the protection scope of the present invention.

[0057] Reference Figure 1 The main steps of the shortwave AMC method based on adaptive reduction quantization and FSMC prediction provided in this embodiment of the invention include the following:

[0058] Step 1: Dataset acquisition and preprocessing.

[0059] The receiver first uses the probe sequence to obtain the channel impulse response in the time domain through convolution. The response with a maximum power of 75% is set as the noise threshold; responses with power less than the noise threshold are considered noise, and responses with power greater than the noise threshold are considered useful signals. The instantaneous signal-to-noise ratio (SNR) within each channel impulse response window is calculated using the power ratio of the useful signal to the noise, thus characterizing the current channel quality. Subsequently, the instantaneous SNRs are stored in the observation buffer in chronological order to obtain an SNR sequence reflecting the fluctuations of the ionospheric channel over time, providing the input data basis for step 2.

[0060] Step 2: Perform model structure adaptive reduction criterion fitting analysis on the signal-to-noise ratio sequence obtained in Step 1.

[0061] The adaptive reduction criterion fitting analysis process refers to... Figure 2 In this process, the system treats the signal-to-noise ratio as a complex statistical whole. The embodiments of this invention employ four candidate probability distribution models: Rayleigh distribution, Rice distribution, Nakagami distribution, and log-normal distribution. These four candidate probability distribution models are selected optimally using the AIC and BIC reduction criteria. The AIC reduction value constrains model effectiveness, while the BIC reduction value constrains model complexity, thereby effectively avoiding overfitting and underfitting. Specifically, BIC penalizes complexity more severely than AIC, and tends to select models with fewer parameters when the sample size is large.

[0062] No. The AIC and BIC reduction values ​​of each candidate model are calculated using the following formula:

[0063] ;

[0064] in, , Let AIC and BIC be the reduction criteria for the i-th candidate distribution model, respectively. This represents the total number of sequences with signal-to-noise ratio. This represents the logarithmic operation with the natural constant e as the base. This represents the maximum likelihood estimate of the probability density function of the i-th candidate distribution model. This represents the signal-to-noise ratio sequence of the nth generation. This represents the total number of parameters in the candidate distribution model;

[0065] To facilitate the comparison and selection of reduced values ​​for candidate distribution models, the AIC and BIC reduced values ​​of each candidate distribution model are normalized to obtain the normalized weights corresponding to each candidate distribution model:

[0066] ;

[0067] in, This represents the normalized weight value of the i-th candidate distribution model. This represents an exponential function with base e. This represents the difference between the reduced value of the i-th candidate distribution model and the minimum reduced value. This represents the total number of candidate distribution models. This represents the difference between the reduced value and the minimum reduced value of the j-th candidate distribution model. Finally, by calculating the average of the normalized weights of AIC and BIC, the distribution model with the maximum average weight is adaptively selected as the most suitable fitted distribution model.

[0068] Step 3: Optimize the state interval of signal-to-noise ratio by iterative optimal non-uniform vector quantization under the joint constraints of mean square error and entropy.

[0069] This step uses a joint cost function of mean squared error and entropy to perform iterative optimal non-uniform vector quantization on the signal-to-noise ratio (SNR) partition boundaries. This enables encrypted partitioning in areas with high SNR frequency and sparse partitioning in edge areas. The joint distribution function formula is as follows:

[0070] ;

[0071] in, This represents the boundary signal-to-noise ratio (SNR) value of the m-th SNR state partition, where M represents the total number of SNR state partitions. This represents the total number of sequences with signal-to-noise ratio. This represents the number of signal-to-noise ratio (SNR) sequences within the m-th SNR state partition. This represents the nth signal-to-noise ratio sequence in the mth signal-to-noise ratio state partition. This represents the center signal-to-noise ratio (SNR) of the m-th SNR state partition. This indicates the weighting factor, and its value is 0.5. This represents a logarithmic operation with base 2. This represents the probability that the signal-to-noise ratio sequence falls into the m-th signal-to-noise ratio state partition. These represent the maximum and minimum signal-to-noise ratio values ​​in the signal-to-noise ratio sequence, respectively.

[0072] Subsequently, convergence optimization is performed on the signal-to-noise ratio (SNR) boundary. During the optimization process, the solver sets an initial SNR partition boundary that satisfies the monotonically increasing boundary constraint, and continuously adjusts the position of each SNR state partition until the cost function... The signal-to-noise ratio (SNR) interval boundaries are obtained upon convergence, thus completing the optimization of the SNR state partitioning.

[0073] Step 4: Calculate the FSMC state transition probability matrix of signal-to-noise ratio based on the channel physical characteristics.

[0074] The calculation process of the FSMC state transition probability matrix for signal-to-noise ratio is referenced. Figure 3 This requires combining channel physical parameters and calculating the autocorrelation coefficient and joint distribution function to finally obtain the FSMC state transition probability matrix of the signal-to-noise ratio. The FSMC state transition probability matrix of the signal-to-noise ratio can be expressed as:

[0075] ;

[0076] in, This represents the state transition probability from the c-th signal-to-noise ratio (SNR) state partition to the f-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the c-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the f-th SNR state partition. Let represent the bivariate joint probability density function of the fitted distribution model. When the Nakagami distribution is chosen in step 2, where Let be the joint binary probability density function of adjacent sampling times under the Nakagami distribution, which can be expressed as:

[0077] ;

[0078] in, This represents the shape parameter of the Nakagami distribution. The autocorrelation coefficient represents the scale parameter of the Nakagami distribution. Used to describe time interval The correlation between adjacent signal-to-noise ratio sampling points Represents the gamma function. express The autocorrelation coefficient of a Bessel function of order -1 can be expressed as:

[0079] ;

[0080] Since ionospheric channels often consist of multiple propagation modes, we assume here that there are two propagation modes, then... It is the power ratio between the two clusters. , It is a two-cluster Doppler extension. and These are the Doppler frequency shifts of the two clusters, where ρ is the autocorrelation function. The value at time.

[0081] Step 5: Based on the FSMC state transition probability matrix of the signal-to-noise ratio, predict the MCS scheme for future transmitted frames by maximizing the expected throughput.

[0082] During the AMC selection process, after receiving the channel quality indication in real time, the transmitting end predicts the channel state at the next frame transmission time from the FSMC state transition probability matrix of the signal-to-noise ratio, and selects the MCS scheme that maximizes throughput to achieve dynamic optimal link allocation. The throughput calculation formula is as follows:

[0083] ;

[0084] in, This represents the average throughput of the o-th MCS scheme at the time of transmission of the next frame. This represents the throughput obtained by selecting the o-th MCS scheme when the signal-to-noise ratio (SNR) is in the m-th SNR state partition. This represents the probability that the signal-to-noise ratio of the next frame falls into the m-th signal-to-noise ratio state partition.

[0085] The effects of this invention can be further illustrated by the following simulation results:

[0086] 1. Simulation experimental conditions.

[0087] The software platform for the simulation experiment of this invention is: Windows 10 operating system and Matlab R2019b.

[0088] The communication system was configured using a single-carrier interleaved frequency division multiple access (FDM) waveform with a bandwidth of 200kHz and a total of 4096 subcarriers. Tail-biting convolutional coding was employed. Simulations were conducted using six typical MCS schemes specified in the shortwave third-generation automatic link standard. The six MCS schemes are shown in Table 1. The modulation methods selected were Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK). The transmission rate doubled progressively from 2.4kbps. The test frame length included the typical bandwidth specified in the shortwave third-generation automatic link standard, ranging from 300ms to 9000ms.

[0089] Table 1 MCS Settings Overview

[0090]

[0091] 2. Simulation content and result analysis.

[0092] The simulation experiment of this invention compares the throughput of this invention and an existing shortwave AMC technology under different frame durations, obtaining throughput results that vary with frame duration, such as... Figure 4 As stated above.

[0093] In the simulation experiment, one existing shortwave AMC technology used is:

[0094] Beijing Rongwei Technology Co., Ltd. proposed a shortwave AMC method based on a preset fixed signal-to-noise ratio threshold in its patent application document "Adaptive Coding and Modulation Protocol Communication Method Based on Half-Duplex Communication Mode" (application number: 202411649246.1, application publication number: CN 119172800 A).

[0095] The following is combined with Figure 4 The simulation diagrams further illustrate the effects of the present invention.

[0096] Figure 4 The horizontal axis represents the frame duration in milliseconds (ms), and the vertical axis represents the average throughput in kbps. The blue curve represents the throughput obtained using existing shortwave AMC technology as a function of frame duration in the simulation experiment of this invention, while the red curve represents the throughput obtained using the technology of this invention as a function of frame duration. The blue area includes the frame lengths specified for the 0th, 1st, 4th, and 5th typical bandwidths in the shortwave third-generation automatic link standard, and the red area includes the frame lengths specified for the 2nd and 3rd typical bandwidths in the shortwave third-generation automatic link standard.

[0097] from Figure 4 As can be seen, across all tested frame durations, the red curve representing the technology of this invention consistently lies above the blue curve representing shortwave AMC technology, indicating a consistently higher average throughput. Specifically, within the blue area (representing the edge of short-duration link control frames at 1300ms), the average throughput is improved by approximately 7%, while within the red area (representing longer-duration link data transmission frames at 8000ms), the average throughput is improved by approximately 24%. This result fully demonstrates that the AMC scheme proposed in this invention can more effectively adapt to changes in channel conditions and significantly outperforms existing shortwave AMC technologies.

Claims

1. A shortwave AMC method based on adaptive reduction quantization and FSMC prediction, characterized in that, The implementation steps of this method include the following: Step 1: Extract the signal-to-noise ratio sequence of the channel from the channel measurement data of the shortwave ionospheric channel scenario; Step 2: Perform adaptive reduction criterion fitting analysis on the extracted signal-to-noise ratio sequence. Based on the reduction results, balance the effectiveness and complexity of the model, and select the fitting distribution model with the largest reduction value that best matches the extracted signal-to-noise ratio sequence. Step 3: The extracted signal-to-noise ratio sequence is optimized by iterative optimal non-uniform vector quantization under the joint constraints of mean square error and entropy. This achieves encrypted partitioning in regions where the signal-to-noise ratio occurs frequently, and sparse partitioning in edge regions, until the cost function converges, resulting in the optimized signal-to-noise ratio state partitioning. Step 4: Calculate the FSMC state transition probability matrix for signal-to-noise ratio; Step 5: During the AMC selection process, after receiving the channel quality indication in real time, the transmitting end predicts the channel state at the time of transmission of the next frame from the FSMC state transition probability matrix of the signal-to-noise ratio, and selects the MCS scheme that maximizes the throughput to achieve dynamic optimal allocation of the link.

2. The shortwave AMC method according to claim 1, characterized in that, The steps for extracting the signal-to-noise ratio sequence of the channel in step 1 are as follows: The first step is for the receiver to use the probe sequence to calculate the channel impulse response in the time domain by convolution, obtaining at least 1000 time-continuous channel impulse responses; The second step is to set the response with a maximum power of 75% in the channel impulse response as the noise threshold, and the power less than the noise threshold is recorded as noise, and the power greater than the noise threshold is recorded as useful signal. The instantaneous signal-to-noise ratio in each channel impulse response window is calculated by the power ratio of useful signal to noise, and the instantaneous signal-to-noise ratio representing the channel state is obtained. The third step is to store the instantaneous signal-to-noise ratio in the observation buffer in chronological order to obtain a signal-to-noise ratio sequence that reflects the fluctuations of the ionospheric channel over time.

3. The shortwave AMC method according to claim 2, characterized in that, The steps of the adaptive reduction criterion fitting analysis of the model structure described in step 2 are as follows: The first step is to select at least four candidate probability distribution models, namely Rayleigh distribution, Rice distribution, Nakagami distribution, and log-normal distribution, and use the AIC and BIC reduction criteria in the following formula to calculate the reduction value of each candidate distribution model. ； in, , Let AIC and BIC be the reduction criteria for the i-th candidate distribution model, respectively. This represents the total number of sequences with signal-to-noise ratio. This represents the logarithmic operation with the natural constant e as the base. This represents the maximum likelihood estimate of the probability density function of the i-th candidate distribution model. This represents the signal-to-noise ratio sequence of the nth generation. This represents the total number of parameters in the candidate distribution model; The second step is to normalize the reduced values ​​of AIC and BIC for each candidate distribution model to obtain the normalized weights corresponding to each candidate distribution model: ； in, This represents the normalized weight value of the i-th candidate distribution model. This represents an exponential function with base e. This represents the difference between the reduced value of the i-th candidate distribution model and the minimum reduced value. This represents the total number of candidate distribution models. This represents the difference between the reduced value and the minimum reduced value of the j-th candidate distribution model; The third step is to calculate the average value of the normalized weights of AIC and BIC for each candidate distribution model, and take the candidate distribution model with the largest average weight as the fitted distribution model of the signal-to-noise ratio sequence.

4. The shortwave AMC method according to claim 2, characterized in that, The balance between model effectiveness and complexity, as described in step 2 based on the results obtained from the reduction, refers to introducing the AIC and BIC reduction criteria when calculating the reduction values ​​of each candidate distribution model. The AIC reduction value constrains the model effectiveness, and the BIC reduction value constrains the model complexity.

5. The shortwave AMC method according to claim 2, characterized in that, The steps for finding the best-fit distribution model in step 2 are as follows: The first step is to calculate the AIC normalized weights and BIC normalized weights for each candidate distribution model; The second step is to calculate the average value of the normalized weights of AIC and BIC for each candidate distribution model; The third step is to select the candidate distribution model with the largest average reduction value of AIC and BIC as the best-fit distribution model that matches the signal-to-noise ratio sequence.

6. The shortwave AMC method according to claim 2, characterized in that, The steps of the iterative optimal non-uniform vector quantization optimization with joint constraints of mean square error and entropy in step 3 are as follows: The first step is to establish a joint quantization cost function for mean squared error and entropy; The second step is to set the initial signal-to-noise ratio partition boundaries that satisfy the monotonically increasing constraint conditions. The third step involves iteratively adjusting the positions of the signal-to-noise ratio partition boundaries in a monotonically increasing manner, with the goal of achieving cost function convergence, until the cost function converges. After convergence, the final signal-to-noise ratio (SNR) interval boundary is obtained, completing the optimization of the SNR state partition.

7. The shortwave AMC method according to claim 6, characterized in that, The cost function The expression is as follows: ； in, This represents the boundary signal-to-noise ratio (SNR) value of the m-th SNR state partition, where M represents the total number of SNR state partitions. This represents the total number of sequences with signal-to-noise ratio. This represents the number of signal-to-noise ratio (SNR) sequences within the m-th SNR state partition. This represents the nth signal-to-noise ratio sequence in the mth signal-to-noise ratio state partition. This represents the center signal-to-noise ratio (SNR) of the m-th SNR state partition. This indicates the weighting factor, and its value is 0.

5. This represents a logarithmic operation with base 2. This represents the probability that the signal-to-noise ratio sequence falls into the m-th signal-to-noise ratio state partition. These represent the maximum and minimum signal-to-noise ratio values ​​in the signal-to-noise ratio sequence, respectively.

8. The shortwave AMC method according to claim 7, characterized in that, The FSMC state transition probability matrix for the signal-to-noise ratio mentioned in step 4 is obtained by the following formula: ； in, This represents the state transition probability from the c-th signal-to-noise ratio (SNR) state partition to the f-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the c-th SNR state partition. This represents the boundary signal-to-noise ratio (SNR) value of the f-th SNR state partition. This represents the bivariate joint probability density function of the fitted distribution model.

9. The shortwave AMC method according to claim 2, characterized in that, The steps in step 5 for predicting the channel state at the time of transmission of the next frame are as follows: The first step is that after receiving the signal-to-noise ratio (SNR) sequence reflecting the current channel state, the transmitting end determines the SNR state partition in which the current channel state is located. The second step is to calculate the probability of occurrence of each SNR state partition at the time of transmission of the next frame from the FSMC state transition probability matrix of the signal-to-noise ratio, so as to reflect the channel state.

10. The shortwave AMC method according to claim 8, characterized in that, The steps for selecting the MCS scheme that maximizes throughput in step 5 are as follows: The first step is to preset a number of MCS schemes that is greater than the total number of signal-to-noise ratio (SNR) state partitions. Each MCS scheme has a different transmission rate in different SNR state partitions. The second step is to calculate the average throughput of each MCS scheme at the next frame transmission time using the following formula: ； in, This represents the average throughput of the o-th MCS scheme at the time of transmission of the next frame. This represents the throughput obtained by selecting the o-th MCS scheme when the signal-to-noise ratio (SNR) is in the m-th SNR state partition. This represents the probability that the signal-to-noise ratio of the next frame falls into the m-th signal-to-noise ratio state partition; The third step is to compare the average throughput of all MCS schemes and select the MCS scheme with the highest average throughput as the transmission scheme for the next frame of data.

Images