A link feature extraction method and system for satellite multi-subcarrier signals

Abstract

The application discloses a link feature extraction method and system for a satellite multi-subcarrier signal and belongs to the technical field of satellite communication. The method comprises the following steps: reading an NPY spectrum data file and a corresponding JSON file; obtaining an effective feature point set sorted by frequency under each timestamp based on the NPY spectrum data file and the corresponding JSON file; forming a plurality of initial subcarrier tracks according to all the effective feature point sets; sorting the frequency positions of the trackers, taking the minimum frequency interval between adjacent trackers as a minimum interval value, and inserting a missing subcarrier track in the initial subcarrier track based on the minimum interval value to obtain a final subcarrier track set; and performing feature extraction based on the final subcarrier track set to obtain a link feature parameter set of the satellite multi-subcarrier signal. The application can overcome defects such as missing detection of feature points, track fracture, missing of intermediate subcarriers and incomplete boundary subcarriers under a low signal-to-noise ratio condition.

Description

Technical Field

[0001] This invention belongs to the field of satellite communication technology, specifically relating to a method and system for extracting link features of satellite multi-subcarrier signals. Background Technology

[0002] The main technique for extracting satellite signal link characteristics involves time-frequency analysis of received spectrum data and detecting and tracking changing signal trajectories in the time-frequency graph. Typically, the receiver first samples the Starlink beacon signal to obtain a complex signal sequence unfolded over time. Then, it identifies the target subcarrier from the spectrum or time-frequency graph and tracks its frequency changes over time. By estimating the center frequency of the target subcarrier at each moment, the Doppler shift process of that subcarrier can be obtained.

[0003] To avoid the problems of spectral peaks being submerged by noise and difficulty in stably locking the target subcarrier when directly estimating the Doppler shift under low signal-to-noise ratio conditions, a local model of the signal is usually established first for a short period of time, and then the target subcarrier is tracked segment by segment using a sliding processing method. It is generally believed that the amplitude of a single subcarrier remains approximately constant over a short period of time, and the Doppler shift can be approximated as a linear change. Therefore, the received signal can be equivalent to a typical linear frequency modulated signal. Since the actual received signal often contains several subcarriers simultaneously, the target subcarrier must be selected from multiple subcarriers before performing Doppler parameter estimation.

[0004] The paper "Doppler Localization Method and Experiment Based on Starlink Beacon Signals" proposes a frequency-domain sliding window estimation method, which first tracks and locks onto the target subcarrier in the frequency domain. The input is a sampled complex signal of a single subcarrier. A coarse estimate of the Doppler frequency shift between the initial and next time moments is first performed using a short-time Fourier transform. Then, a frequency-domain sliding window is constructed, centered on the coarsely estimated frequency and with bandwidth constrained by the subcarrier bandwidth and the spacing between adjacent subcarriers. The next step is to search for spectral peaks within the window, calculate signal power, noise power, and signal-to-noise ratio, and thus determine whether the current window still correctly locks onto the target subcarrier. However, this paper focuses on extracting individual subcarriers rather than extracting a group of subcarriers from a single signal. Since the frequencies of different subcarriers have certain relationships, high real-time requirements may lead to the loss of frequency information. Furthermore, this method is heavily reliant on model assumptions, approximating the long-term Doppler curve as a polynomial. If the actual trajectory exhibits local abrupt changes, nonlinear bending, short-term anomalous disturbances, or if different subcarriers are not strictly isomorphic, model mismatch may occur.

[0005] Chinese patent application CN115291249A proposes a method combining coarse and fine estimation. First, signals are screened based on time frame patterns and ephemeris data. Then, a Doppler frequency shift estimate is obtained using Welch power spectrum estimation. Subsequently, signals with high signal-to-noise ratio and long duration are selected, and corresponding time frames are coarsely screened and truncated in both the time and frequency domains. Finally, a low-pass filter is configured based on the signal bandwidth according to the coarse Doppler estimate. However, it relies heavily on prior auxiliary information, requiring the combination of TLE ephemeris data and elevation angle range to screen reception periods. In the embodiments, the reception elevation angle is typically set to 10°-90°. This indicates that it is not a completely "blind extraction" method, but rather based on prior knowledge of the signal structure and identification. The design process of this patent application is based on opportunistic signals like Iridium satellites, which have fixed time frames and can be screened by channel. For other signal systems, especially those with multiple subcarriers, significant overall drift, unknown or unstable structures, it cannot be assumed that this method is equally suitable; typically, the front-end detection and tracking strategies need to be redesigned. Summary of the Invention

[0006] This invention proposes a link feature extraction method and system for satellite multi-subcarrier signals to overcome the shortcomings of existing methods, such as missed feature points, broken tracks, missing intermediate subcarriers, and incomplete boundary subcarriers, which are prone to occur under low signal-to-noise ratio conditions.

[0007] To achieve the above objectives, the technical solution of the present invention includes the following:

[0008] A method for extracting link features of satellite multi-subcarrier signals, the method comprising: Read the NPY spectrum data file and the corresponding JSON file of the satellite multi-subcarrier wireless signal; wherein, the NPY spectrum data file is used to provide a frequency array and a time-frequency matrix arranged by time, and the JSON file is used to provide auxiliary information, including timestamps; Based on the NPY spectrum data file and the corresponding JSON file, a set of valid feature points sorted by frequency is obtained for each timestamp; Based on all the valid feature point sets, multiple initial subcarrier trajectories are formed; among them, a tracker is used to store the time-series point information of each initial subcarrier trajectory; The trackers are sorted by frequency position, the minimum frequency interval between adjacent trackers is taken as the minimum interval value, and the missing subcarrier trajectories are inserted into the initial subcarrier trajectories based on the minimum interval value to obtain the final set of subcarrier trajectories. Feature extraction is performed based on the final set of subcarrier trajectories to obtain the link characteristic parameters of the satellite multi-subcarrier signal.

[0009] Furthermore, based on the NPY spectrum data file and the corresponding JSON file, a set of valid feature points sorted by frequency for each timestamp is obtained, including: The noise basis is estimated row by row of the time-frequency matrix to obtain the noise reference value corresponding to each timestamp, and the noise basis of each row is obtained based on the noise reference value. Obtain the frequency point amplitude based on the frequency array; The amplitude of a frequency point is compared with the noise floor of the corresponding row, and when the amplitude of a frequency point exceeds the noise floor by more than a preset threshold, the corresponding frequency point is determined to be a valid feature point.

[0010] Furthermore, based on all the valid feature point sets, multiple initial subcarrier trajectories are formed, including: An initial tracker is built based on the valid feature points in the first timestamp; The current valid feature points are associated with the nearest neighbors of existing activity trackers to update the existing activity trackers; for existing activity trackers that are not successfully updated, points are predicted and supplemented based on the relative positional relationship of adjacent trajectories and historical change trends to update the trackers; valid feature points that are not assigned in the nearest neighbor association are analyzed, and a new tracker is generated when the analysis results meet the set conditions. This process continues until all valid feature point sets under all timestamps are updated, resulting in multiple initial subcarrier trajectories.

[0011] Furthermore, based on this minimum interval value, omitted subcarrier trajectories are inserted into the initial subcarrier trajectory to obtain the final set of subcarrier trajectories, including: If the interval between two adjacent trackers is an integer multiple of the minimum interval value, a new tracker is inserted at the corresponding position; wherein, the new tracker is interpolated and reconstructed based on the relative positional relationship between the two adjacent existing tracks; The interval relationship between all adjacent trackers is statistically analyzed. When there is still a gap between two adjacent trackers that meets the minimum interval multiple relationship, the missing trajectory is filled by linear interpolation and a new tracker is generated. Select the tracker with the lowest frequency and the tracker with the highest frequency as the boundary reference trajectories of the left and right boundaries. Search for valid feature points that are not used by the trackers along the direction of decreasing frequency and increasing frequency, respectively. When the number of valid feature points that meet the set conditions reaches the set number, create the corresponding left boundary tracker or right boundary tracker.

[0012] Furthermore, the setting conditions include: The distance or frequency position relationship between the effective feature point and the existing trajectory satisfies the subcarrier spacing rule; and, The effective feature point has a tendency to be continuously correlated in subsequent or adjacent timestamps.

[0013] Furthermore, after inserting omitted subcarrier trajectories into the initial subcarrier trajectory based on this minimum interval value to obtain the final set of subcarrier trajectories, the method further includes: For each timestamp, check whether all trackers have the corresponding data; If the corresponding data for a certain tracker is missing at a certain timestamp, the missing point will be predicted and filled in based on the most recent historical position of the corresponding data, the frequency relationship of adjacent trackers, or the trajectory change trend of the tracker itself.

[0014] Furthermore, after obtaining the link feature extraction results of the satellite multi-subcarrier signal, the method further includes: Based on link characteristic parameters, a trajectory visualization image is generated; wherein, the link characteristic parameters include subcarrier frequency, bandwidth, amplitude, average spacing between adjacent subcarriers, and average frequency shift value.

[0015] A link feature extraction system for satellite multi-subcarrier signals, the system comprising: The data reading module is used to read the NPY spectrum data file and the corresponding JSON file of the satellite multi-subcarrier wireless signal; wherein, the NPY spectrum data file is used to provide a frequency array and a time-frequency matrix arranged by time, and the JSON file is used to provide auxiliary information, including timestamps; The feature point extraction module is used to obtain a set of valid feature points sorted by frequency for each timestamp based on the NPY spectrum data file and the corresponding JSON file. The trajectory generation module is used to generate multiple initial subcarrier trajectories based on all valid feature point sets; among them, a tracker is used to store the time-series point information of each initial subcarrier trajectory; The trajectory completion module is used to sort the trackers by frequency position, take the minimum frequency interval between adjacent trackers as the minimum interval value, and insert the missing subcarrier trajectory into the initial subcarrier trajectory based on the minimum interval value to obtain the final subcarrier trajectory set. The feature extraction module is used to extract features based on the final subcarrier trajectory set to obtain the link feature parameters of the satellite multi-subcarrier signal.

[0016] A computer device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the link feature extraction method for satellite multi-subcarrier signals as described above.

[0017] A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the link feature extraction method for satellite multi-subcarrier signals as described above.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects.

[0019] Existing methods often employ fixed threshold detection, local peak search, or simple proximity connection to process feature points or trajectories in time-frequency maps. When the signal-to-noise ratio (SNR) is high and the trajectory continuity is good, these methods can complete basic trajectory extraction. However, under conditions of low SNR, fluctuating background noise, missing local feature points, and incomplete boundary signals, problems such as missed feature point detection, trajectory breaks, missing intermediate subcarriers, and incomplete boundary subcarrier recovery easily occur. This results in the final output being only a local trajectory fragment, making it difficult to form a complete and continuous description of the link features.

[0020] This invention estimates the noise floor row-by-row from the time-frequency matrix and extracts effective feature points based on this floor. This allows feature point detection to no longer rely on a fixed global threshold, but rather to adaptively discriminate based on local noise backgrounds at different times. Using this technique reduces false positives and false negatives caused by background noise fluctuations, improves the stability of effective feature point detection, and provides a more reliable data foundation for subsequent trajectory recovery. Particularly under low signal-to-noise ratio conditions, this method exhibits good preservation capabilities for weak feature points.

[0021] This invention does not treat each subcarrier trajectory in isolation, but rather tracks and recovers multiple subcarriers as a whole structure. By establishing a tracker set and employing methods such as nearest neighbor association, predictive point filling, and new trajectory generation, the continuity of the trajectory can be maintained even in the event of local detection failure, reducing track breaks caused by anomalies in a single time slice. Compared with existing methods that rely solely on connecting points from the original detection results, this invention demonstrates better performance in terms of trajectory continuity and overall stability.

[0022] For intermediate missing trajectories that are difficult to recover using existing methods, this invention utilizes the relatively stable minimum interval relationship between adjacent trajectories for judgment and interpolation. When it is detected that the interval between a certain adjacent trajectory satisfies an integer multiple of the minimum interval relationship, it can be inferred that there is a missing trajectory, and a corresponding number of intermediate trajectories are inserted accordingly. This processing method can recover intermediate missing subcarriers while maintaining the overall arrangement regularity of the multi-subcarriers. Furthermore, this invention also includes a secondary check and direct line filling stage to correct local gaps that still exist after the initial interpolation, thereby improving the completeness of trajectory recovery.

[0023] For spectral boundary regions, this invention identifies the leftmost and rightmost reference trajectories, searches for unused candidate points outside the boundary, and determines whether there are missing subcarriers based on the interval relationship with the boundary trajectories. When preset conditions are met, boundary completion trajectories are automatically generated. Compared with existing methods that mainly process the middle region, this invention can simultaneously address boundary region recovery, thereby improving the integrity of the overall subcarrier structure.

[0024] In the time dimension, this invention also checks the data integrity of each trajectory at different times. For trajectories with missing moments, the missing points are predicted and filled in based on the most recently known location, trajectory change trends, and relative relationships with reference trajectories. This technique reduces interruptions in the trajectory along the time axis, ensuring that each trajectory remains as continuous as possible throughout the observation period. This is of practical significance for subsequent calculations of Doppler frequency shift, trajectory duration, and other time-varying link parameters.

[0025] Regarding the output results, this invention is not limited to outputting trajectory images. Instead, it further organizes the recovered results into structured link features, including parameters such as frequency point, bandwidth, amplitude, subcarrier spacing, and Doppler frequency shift, and outputs them in a unified format. Compared with methods that only output images or single frequency shift values, the information output by this invention is more complete, making it easier to use directly for subsequent signal identification, link analysis, satellite selection decisions, and positioning processing, reducing the workload of manual processing and secondary conversion.

[0026] Furthermore, this invention supports recursive processing of multiple files, automatically completing the entire process of data reading, feature point detection, trajectory tracking, interval interpolation, boundary completion, trajectory completion, and result output, demonstrating strong automation capabilities. Compared to processing methods that rely on manual image interpretation and line completion, this invention reduces human intervention, improves processing efficiency and result consistency, and is more suitable for engineering applications and batch data processing scenarios.

[0027] In summary, this invention, through techniques such as noise basis estimation, multi-subcarrier overall tracking, minimum interval constraint interpolation, boundary completion, time dimension completion, and structured output, can effectively solve the problems of missed feature points, trajectory breaks, difficulty in recovering missing trajectories, and incomplete output results existing in the prior art, thereby improving the completeness, continuity, accuracy, and practicality of satellite multi-subcarrier signal link feature extraction. Attached Figure Description

[0028] Figure 1 This is a flowchart of the logical processing.

[0029] Figure 2 This is a flowchart of the data preprocessing process.

[0030] Figure 3This is a flowchart of the interval analysis and interpolation process.

[0031] Figure 4 Logical diagram for boundary completion and trajectory completion.

[0032] Figure 5 These are key technical effect diagrams.

[0033] Figure 6 The output diagram is labeled with subcarriers. Detailed Implementation

[0034] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, and to make the objectives, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0035] For multi-subcarrier trajectories in a two-dimensional time-frequency graph formed from spectrum data, this invention achieves automatic detection, continuous tracking, and complete recovery of multi-subcarrier trajectories through noise basis estimation, effective feature point detection, trajectory correlation tracking, missing point completion, missing trajectory interpolation based on subcarrier spacing relationships, and boundary trajectory completion. Furthermore, it outputs link features such as the number of subcarriers, frequency point location, bandwidth, amplitude, subcarrier spacing, and Doppler frequency shift, thereby improving the completeness, continuity, and accuracy of satellite signal link feature extraction and providing structured feature support for subsequent signal identification, link analysis, and positioning applications.

[0036] Specifically, the link feature extraction method for satellite multi-subcarrier signals of the present invention, such as... Figure 1 As shown, it includes the following steps.

[0037] (1) Read the input data and complete the preprocessing initialization.

[0038] In this invention, the NPY spectrum data file to be processed and its corresponding JSON file are first read; the NPY file provides the frequency array and time-frequency matrix data arranged by time, while the JSON file provides auxiliary information such as timestamps. Simultaneously, the data structures required for subsequent processing are initialized, including: A set of trackers used to store the timing information of each subcarrier trajectory; The original feature point set is used to record all valid feature points detected at each timestamp; The set of points already used is used to mark feature points that have been occupied by the tracker; The valid point mapping table is used to store candidate feature points obtained by threshold detection at each timestamp.

[0039] The generation process of the NPY spectrum data file and its corresponding JSON file is as follows: RF hardware acquires satellite multi-subcarrier wireless signals (IQ time-domain baseband sampling data stream); based on the ADC sampling rate, subcarrier spacing, and FFT points, the corresponding frequency array is calculated, and simultaneously, an FFT transformation is performed on the IQ data stream to generate a time-frequency matrix arranged in time; the frequency array and the time-frequency matrix are jointly encapsulated and saved as an NPY spectrum data file, providing a frequency domain reference and time-series spectrum values ​​for subsequent link feature extraction; simultaneously, the acquired satellite communication physical parameters, equipment parameters, channel conditions, and timestamps are extracted to generate a structured JSON file; and the NPY spectrum data file and the JSON file are paired one-to-one.

[0040] (2) Estimate the noise base row by row and extract effective feature points.

[0041] The noise basis is estimated row by row of the input time-frequency matrix to obtain the noise reference value corresponding to each timestamp, and a noise basis for the corresponding row is generated based on the noise reference value. Then, the amplitude of the frequency points obtained from the frequency arrays in each NPY file is compared with the noise basis for that row. When a frequency point exceeds the noise basis by more than a preset threshold, it is determined as a valid feature point. After this step, a set of valid feature points sorted by frequency for each timestamp is obtained, providing input for subsequent trajectory tracking.

[0042] In one embodiment, such as Figure 2 As shown, the preprocessing process first reads the NPY spectrum data and JSON timestamp file, and completes the parsing of the frequency array, time-frequency matrix, and timestamp information. Then, the integrity of the input data is verified. Under the premise that the data is valid, the noise floor is estimated row by row of the time-frequency matrix, and valid feature points are detected based on the threshold method. Finally, the set of valid feature points, noise floor information, and timestamp mapping relationship are output, providing basic data support for subsequent trajectory tracking and link feature extraction.

[0043] (3) Initialize the tracker based on the valid feature points of the first timestamp, and perform trajectory tracking for subsequent timestamps.

[0044] In this invention, an initial tracker is first established based on the valid feature points in the first timestamp; for each subsequent timestamp, the following operations are performed sequentially: First, the effective feature points obtained in the previous step are associated with the nearest neighbors of the existing activity trackers to update the existing trajectories; Second, for trackers that have not been successfully updated, predictions are made to fill in the missing points based on the relative positional relationship of adjacent trajectories and historical trends. Third, the remaining unassigned feature points are analyzed, and new trackers are generated when certain conditions are met. These conditions include: the distance or frequency position relationship between the track and existing trajectories must satisfy the subcarrier spacing rule; and the track must exhibit a continuous correlation trend in subsequent or adjacent timestamps, thus meeting the requirements for generating new trajectories.

[0045] This step can generate multiple initial subcarrier trajectories.

[0046] (4) Perform interval analysis on the initial tracking results and insert the missing trajectories.

[0047] After obtaining the initial tracking results, all valid trackers are sorted by frequency position, and the frequency interval between adjacent trackers is calculated to obtain the minimum interval value. If the interval between two adjacent trackers is found to be an integer multiple of the minimum interval, it is determined that there is a missing trajectory in that area, and a new tracker is inserted at the corresponding position. The inserted tracker performs interpolation and reconstruction based on the relative positional relationship between two adjacent existing trajectories, thereby realizing the recovery of the missing subcarrier in the middle.

[0048] (5) Check the adjacent trajectories again and directly fill in the missing lines.

[0049] After completing the first interval analysis, the interval relationships of all adjacent trackers are statistically analyzed again. When gaps are found between some adjacent tracks that still satisfy the minimum interval multiple relationship, linear interpolation is used to fill in the missing tracks and generate new supplementary line trackers. This step further improves the integrity of the multi-subcarrier structure recovery.

[0050] In one embodiment, such as Figure 3 As shown, the interval analysis and interpolation process is as follows: First, input the initial trajectory data, including the initial tracker set and the frequency position of each trajectory. Then, sort the trajectories by frequency position, count the intervals between adjacent trajectories, and use the smallest interval as the baseline interval. Next, determine if there are any significant large interval gaps. If not, output the current trajectory result directly. If there are, further calculate the interval multiple relationship corresponding to the gap to determine the number of missing trajectories to be inserted, and generate intermediate interpolated trajectories at the gap positions according to the baseline interval to fill in the missing trajectory points. After interpolation, add the newly generated trajectories to the trajectory set and recount the adjacent intervals. If there are still missing intervals, return to the previous steps to continue iterative processing until there are no missing intervals. Finally, output the complete trajectory set and interval analysis results.

[0051] (6) Perform boundary completion checks and restore missing subcarriers on the left and right boundaries.

[0052] In this invention, the tracker with the lowest current frequency and the tracker with the highest current frequency are further selected as the left and right boundary reference trajectories. Then, feature points not used by the trackers are searched along the frequency decreasing direction and the frequency increasing direction, respectively, and it is determined whether the frequency difference between these points and the boundary reference trajectory satisfies the minimum interval integer multiple relationship. When the condition is met and the number of detection points reaches the set requirement, the corresponding left or right boundary tracker is created to supplement the missing subcarriers at the boundary.

[0053] In one embodiment, such as Figure 4 As shown in the figure, this diagram illustrates the overall processing flow of boundary completion and trajectory completion: First, the current tracker data and adjacent trajectory interval information are input. The leftmost and rightmost reference trajectories are determined through boundary trajectory identification, and the boundary frequency positions are extracted. Then, it is determined whether there are any missing subcarriers at the boundaries. If so, untracked points are searched on the left and right boundaries, and boundary completion is performed according to the reference interval. If not, the trajectory integrity check is directly initiated. Next, the time data of each trajectory is checked to determine if there are any missing moments. If so, the missing moment data is completed by predicting the location of the missing points, and the completion result is added to the trajectory set and its integrity is checked again. If missing trajectory data still exists, the iteration process continues until no missing data is found. Finally, the complete trajectory set and the boundary completion and trajectory completion results are output. Figure 6 This is an example of a subcarrier label output diagram.

[0054] (7) Complete all tracker data and ensure that the trajectory is complete at each timestamp.

[0055] Because some trackers may have missing data at certain timestamps, all tracker data is uniformly completed. Specifically, for each timestamp, it is checked whether all trackers have corresponding data; if a tracker is missing at a certain timestamp, the missing data is predicted and completed based on its most recent historical position, the frequency relationship with adjacent reference trackers, or its own trajectory change trend, thereby ensuring that each timestamp contains complete multi-subcarrier trajectory information.

[0056] (8) Calculate the link characteristic parameters and output the structured results.

[0057] After completing the trajectory recovery and completion, the frequency index is converted into the actual frequency value, and the link characteristic parameters at each timestamp are further calculated, including subcarrier frequency, bandwidth, amplitude, average spacing between adjacent subcarriers, and average frequency shift. Finally, the extraction results are output in NPY and JSON formats, and a trajectory visualization image is generated. This achieves unified extraction of link characteristics such as the number of subcarriers, frequency position, bandwidth, amplitude, subcarrier spacing, and Doppler frequency shift of satellite multi-subcarrier signals.

[0058] In summary, most existing satellite multi-subcarrier signal processing methods focus on a single subcarrier or a single time-stamp feature point, typically employing simple threshold detection, local peak search, or proximity connection methods for trajectory recovery. While these methods can achieve basic frequency estimation and feature extraction under conditions of high signal-to-noise ratio and good trajectory continuity, in practical applications, satellite signals often exhibit complex characteristics such as parallel distribution of multiple subcarriers, overall drift, local discontinuities, and boundary truncation. Traditional methods struggle to fully utilize the inherent structural relationships between multiple trajectories. Existing methods generally lack the ability to uniformly model the overall structure of multiple subcarriers, often focusing only on the frequency shift estimation of a single target trajectory while ignoring the relatively stable interval constraints and overall arrangement patterns between adjacent trajectories. When there are large fluctuations in the spectral background or local time-slice detection failures, relying solely on the original detection results for correlation can easily lead to misconnected or broken trajectories, and even distorted judgments regarding the number of subcarriers, frequency positions, and interval relationships. Meanwhile, existing technologies do not pay enough attention to untracked points in the spectral boundary region and lack an effective mechanism to complete data missing in the time dimension. As a result, the final output often only yields incomplete trajectory fragments, making it difficult to form stable, structured link features that can be directly used for subsequent identification and analysis.

[0059] To address the aforementioned shortcomings, this invention first focuses on the fundamental aspects of signal detection, such as... Figure 5 As shown, the reliability of the original trajectory recovery is improved through two key modules: "noise basis estimation and feature point detection" and "multi-subcarrier overall trajectory tracking". On this basis, instead of processing each subcarrier in isolation, the trajectory of the entire multi-subcarrier is tracked. The trajectory is updated through nearest neighbor association, and the continuity and integrity of the trajectory are maintained by predictive point supplementation and new trajectory generation mechanisms, laying the foundation for subsequent overall structure recovery.

[0060] In terms of trajectory structure recovery, this invention introduces a "minimum interval constraint for missing trajectory interpolation" and a "secondary check and direct line completion" mechanism to solve the problem of difficulty in recovering missing intermediate subcarriers and broken trajectories in existing methods. Simultaneously, this invention also sets up two modules: "boundary missing subcarrier completion" and "time dimension trajectory completion," which perform completion processing on the frequency boundary and time dimension, respectively. The former identifies the leftmost and rightmost reference trajectories and searches for untracked points at the boundary. This invention achieves technical effectiveness through "unified output of structured link features" and "automated batch processing capabilities." On the one hand, the recovered complete trajectory set is uniformly converted into structured link features, including key parameters such as frequency, bandwidth, amplitude, subcarrier interval, and Doppler frequency shift, significantly improving the completeness, continuity, and usability of the output results. On the other hand, this invention supports recursive processing of multiple files and automatic result generation, and can operate stably in complex scenarios such as low signal-to-noise ratio, local missed detections, and boundary truncation, enhancing the automated processing capabilities and scenario adaptability in engineering applications.

[0061] Although the present invention has been described through the above preferred embodiments, those skilled in the art will understand that various modifications in form and detail can be made therein without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for extracting link features of satellite multi-subcarrier signals, characterized in that, The method includes: Read the NPY spectrum data file and the corresponding JSON file of the satellite multi-subcarrier wireless signal; wherein, the NPY spectrum data file is used to provide a frequency array and a time-frequency matrix arranged by time, and the JSON file is used to provide auxiliary information, including timestamps; Based on the NPY spectrum data file and the corresponding JSON file, a set of valid feature points sorted by frequency is obtained for each timestamp; Based on all the valid feature point sets, multiple initial subcarrier trajectories are formed; among them, a tracker is used to store the time-series point information of each initial subcarrier trajectory; The trackers are sorted by frequency position, the minimum frequency interval between adjacent trackers is taken as the minimum interval value, and the missing subcarrier trajectories are inserted into the initial subcarrier trajectories based on the minimum interval value to obtain the final set of subcarrier trajectories. Feature extraction is performed based on the final set of subcarrier trajectories to obtain the link characteristic parameters of the satellite multi-subcarrier signal.

2. The method according to claim 1, characterized in that, Based on the NPY spectrum data file and the corresponding JSON file, a set of valid feature points sorted by frequency for each timestamp is obtained, including: The noise basis is estimated row by row of the time-frequency matrix to obtain the noise reference value corresponding to each timestamp, and the noise basis of each row is obtained based on the noise reference value. Obtain the frequency point amplitude based on the frequency array; The amplitude of a frequency point is compared with the noise floor of the corresponding row, and when the amplitude of a frequency point exceeds the noise floor by more than a preset threshold, the corresponding frequency point is determined to be a valid feature point.

3. The method according to claim 1, characterized in that, Based on all valid feature point sets, multiple initial subcarrier trajectories are formed, including: An initial tracker is built based on the valid feature points in the first timestamp; The current valid feature points are associated with the nearest neighbors of existing activity trackers to update the existing activity trackers; for existing activity trackers that are not successfully updated, points are predicted and supplemented based on the relative positional relationship of adjacent trajectories and historical change trends to update the trackers; valid feature points that are not assigned in the nearest neighbor association are analyzed, and a new tracker is generated when the analysis results meet the set conditions. This process continues until all valid feature point sets under all timestamps are updated, resulting in multiple initial subcarrier trajectories.

4. The method according to claim 1, characterized in that, Based on this minimum interval value, missing subcarrier trajectories are inserted into the initial subcarrier trajectory to obtain the final set of subcarrier trajectories, including: If the interval between two adjacent trackers is an integer multiple of the minimum interval value, a new tracker is inserted at the corresponding position; wherein, the new tracker is interpolated and reconstructed based on the relative positional relationship between the two adjacent existing tracks; The interval relationship between all adjacent trackers is statistically analyzed. When there is still a gap between two adjacent trackers that meets the minimum interval multiple relationship, the missing trajectory is filled by linear interpolation and a new tracker is generated. Select the tracker with the lowest frequency and the tracker with the highest frequency as the boundary reference trajectories of the left and right boundaries. Search for valid feature points that are not used by the trackers along the direction of decreasing frequency and increasing frequency, respectively. When the number of valid feature points that meet the set conditions reaches the set number, create the corresponding left boundary tracker or right boundary tracker.

5. The method according to claim 3 or 4, characterized in that, The setting conditions include: The distance or frequency position relationship between the effective feature point and the existing trajectory satisfies the subcarrier spacing rule; and, The effective feature point has a tendency to be continuously correlated in subsequent or adjacent timestamps.

6. The method according to claim 1, characterized in that, After inserting omitted subcarrier trajectories into the initial subcarrier trajectory based on the minimum interval value to obtain the final set of subcarrier trajectories, the method further includes: For each timestamp, check whether all trackers have the corresponding data; If the corresponding data for a certain tracker is missing at a certain timestamp, the missing point will be predicted and filled in based on the most recent historical position of the corresponding data, the frequency relationship of adjacent trackers, or the trajectory change trend of the tracker itself.

7. The method according to claim 1, characterized in that, After obtaining the link feature extraction results of the satellite multi-subcarrier signal, the method further includes: Based on link characteristic parameters, a trajectory visualization image is generated; wherein, the link characteristic parameters include subcarrier frequency, bandwidth, amplitude, average spacing between adjacent subcarriers, and average frequency shift value.

8. A link feature extraction system for satellite multi-subcarrier signals, characterized in that, The system includes: The data reading module is used to read the NPY spectrum data file and the corresponding JSON file of the satellite multi-subcarrier wireless signal; wherein, the NPY spectrum data file is used to provide a frequency array and a time-frequency matrix arranged by time, and the JSON file is used to provide auxiliary information, including timestamps; The feature point extraction module is used to obtain a set of valid feature points sorted by frequency for each timestamp based on the NPY spectrum data file and the corresponding JSON file. The trajectory generation module is used to generate multiple initial subcarrier trajectories based on all valid feature point sets; among them, a tracker is used to store the time-series point information of each initial subcarrier trajectory; The trajectory completion module is used to sort the trackers by frequency position, take the minimum frequency interval between adjacent trackers as the minimum interval value, and insert the missing subcarrier trajectory into the initial subcarrier trajectory based on the minimum interval value to obtain the final subcarrier trajectory set. The feature extraction module is used to extract features based on the final subcarrier trajectory set to obtain the link feature parameters of the satellite multi-subcarrier signal.

9. A computer device, characterized in that, The computer device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the link feature extraction method for satellite multi-subcarrier signals as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the link feature extraction method for satellite multi-subcarrier signals as described in any one of claims 1-7.

Images