A collision frequency hopping signal sorting method based on time-frequency feature matching

Abstract

The application belongs to the field of satellite communication, and discloses a collision frequency hopping signal sorting method based on time-frequency feature matching, which comprises the following steps: preprocessing an input time-frequency graph, namely, sequentially performing adaptive binarization and opening operation; secondly, performing signal detection, performing connected domain judgment on the preprocessed time-frequency graph, and then extracting time-frequency feature parameters; finally, performing signal sorting, clustering the time-frequency feature parameters, and processing the collision frequency hopping signal by using time-frequency feature matching. The time-frequency feature matching introduced in the application introduces a filling factor, an overlap rate and a time interval constraint, and effectively improves the sorting accuracy of the collision frequency hopping signal under noise background.

Description

Technical Field

[0001] This application belongs to the field of satellite communications, specifically relating to a collision frequency hopping signal sorting method based on time-frequency feature matching. Background Technology

[0002] Frequency-hopping signals are widely used in modern wireless communication systems. With the rapid development of on-orbit spectrum monitoring, space-based platforms need to observe a wide range of electromagnetic environments. Compared to fixed ground monitoring stations with limited coverage, space-based platforms have the advantage of a wide monitoring area and can simultaneously receive frequency-hopping signals from multiple regions and multiple service systems. Therefore, in space-based spectrum monitoring scenarios, the receiver typically acquires a mixed signal formed by multiple frequency-hopping radiation sources, requiring sorting to extract useful information.

[0003] Because multiple frequency-hopping radiation sources often operate simultaneously in adjacent time-frequency regions, different frequency-hopping trajectories are prone to collisions on the time-frequency graph, leading to blurred hopping boundaries, instantaneous graphic distortion, and local structural adhesion. When these colliding frequency-hopping signals cannot be effectively sorted, it often results in missing hops in the frequency-hopping sequence, affecting de-hopping and signal reconstruction processes and reducing the reliability of spectrum monitoring and analysis. Therefore, how to achieve reliable sorting of frequency-hopping signals under noisy background and signal collision conditions has become a key technical problem in the field of space-based spectrum monitoring.

[0004] However, existing technologies still have significant shortcomings in collision frequency-hopping signal sorting scenarios, especially in traditional template matching methods. Because the rectangular envelope of the frequency-hopping signal is disrupted under collision conditions, relying solely on geometric similarity can easily lead to incorrect matches in partially overlapping areas, and may also result in repeated matches in already matched areas, thus reducing the accuracy of collision frequency-hopping signal sorting. Therefore, a new method suitable for sorting collision frequency-hopping signals is urgently needed, enabling effective separation and classification of collision frequency-hopping signals in noisy environments. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this application provides a collision frequency hopping signal sorting method based on time-frequency feature matching. This method introduces a fill factor to characterize the signal integrity, sets an overlap rate to suppress erroneous matching results, and uses time interval constraints to ensure the rationality of the hopping timing. This application can accurately sort collision frequency hopping signals, effectively solve the boundary distortion problem, and retain a more complete frequency hopping sequence.

[0006] To achieve the above objectives, this application employs the following technical solution:

[0007] This application discloses a collision frequency hopping signal sorting method based on time-frequency feature matching, characterized in that the collision frequency hopping signal sorting method specifically includes the following steps:

[0008] Step 1, Adaptive Binarization: Perform adaptive binarization on the input time-frequency graph, and obtain the binary time-frequency graph using the triangular thresholding method. ;

[0009] Step 2, Opening Operation: Apply the binary time-frequency graph obtained in Step 1. Perform an opening operation to obtain the time-frequency diagram after the opening operation. ;

[0010] Step 3, Connectivity Component Determination: Analyze the time-frequency graph after the opening operation in Step 2. Perform connectivity determination, identify and mark each independent connected component;

[0011] Step 4, Parameter Extraction: For each connected component obtained in Step 3, extract the minimum bounding rectangle and obtain the width, height and fill factor of each connected component.

[0012] Step 5: Density-based spatial clustering: Using the width, height, and fill factor of connected components as features, the density-based spatial clustering algorithm (DBSCAN) is used to classify each connected component.

[0013] Step 6, Time-Frequency Feature Matching: Based on the time-frequency features of various non-collision frequency hopping signals obtained from clustering in Step 5, templates are constructed. Subgraphs and occupancy matrices are generated for the collision regions. The templates are then slid-matched by combining the matching score, overlap rate, and time interval constraints to sequentially complete the separation and classification of each collision frequency hopping signal.

[0014] A further improvement of this application is that, in step 4, for each connected component, the minimum bounding rectangle corresponding to the connected component is extracted, and a fill factor is defined. :

[0015]

[0016] in, and For the first The width and height of the minimum bounding rectangle of a connected region. , This represents the total number of connected components after the connected component labeling process. For the first The area of ​​a connected region, when When, it means that the connected component is filled with its smallest bounding rectangle. When , it indicates that there is a gap inside the smallest bounding rectangle.

[0017] A further improvement of this application is that in step 5, DBSCAN clustering is performed on the extracted time-frequency feature parameters. Here, the width of the connected component , height and filling factor are used as the features for clustering by the Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Each connected component is classified through the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to obtain clusters corresponding to non-collision frequency-hopping signals and noise points corresponding to collision frequency-hopping signals , , .

[0018] A further improvement of this application is that step 6 specifically includes the following steps:

[0019] Step 6.1, Calculate the average of the widths and heights of each type of non-collision frequency-hopping signal obtained by clustering to construct a template , where and represent the width and height of the template respectively, is a matrix with all elements being 1. Sort the template in descending order according to the area, and subsequent indices all follow this sorting.

[0020] Step 6.2, Let the collision area corresponding to the minimum circumscribed rectangle enclosing the collision frequency-hopping signal in the time-frequency diagram after opening operation be the sub-graph . Introduce an occupancy matrix to mark the positions in the sub-graph that have been successfully matched with the template . Its initial setting is a matrix with the same size as the sub-graph and all elements being 0, where and represent the width and height of the sub-graph respectively;

[0021] Step 6.3, Compare the template with the sub-graph to determine whether the template is available. If the template is much larger than the width or height of the sub-graph in terms of width or height, skip the current template; otherwise, while keeping the aspect ratio unchanged, simultaneously reduce the width and height of the template proportionally until the template can be placed within the sub-graph;

[0022] Step 6.4. Slide the template and traverse the entire sub - graph: For each position in the sub - graph , calculate the matching score between the template and the sub - graph :

[0023]

[0024] where , , represents the time index in the sub - graph, represents the frequency index in the sub - graph;

[0025] Calculate the overlap rate between the template and the occupancy matrix :

[0026]

[0027] When the template finishes sliding over the entire sub - graph, the matching scores corresponding to all positions in the sub - graph and the overlap rate are obtained. If the denominator calculation result of the matching score is zero, discard the corresponding position;

[0028] To ensure the reliability of the time - frequency feature matching, set the threshold of the matching score to the minimum filling factor in the non - collision frequency - hopping signals corresponding to the template category. After the template finishes sliding over the entire sub - graph, collect all candidate positions that satisfy .

[0029] Step 6.5. Sort the candidate positions in ascending order according to the overlap rate ; if the overlap rate is the same, sort them in descending order according to the matching score . The time - interval constraint for each candidate position is defined as:

[0030]

[0031] where represents the starting index of the candidate position, represents the set of starting indices of the non - collision frequency - hopping signals of the th category, is If any element in the cluster satisfies the time interval constraint, the match is considered successful, and the current candidate position is then assigned to the corresponding cluster. and occupy the matrix The element at the corresponding position is set to 1; otherwise, the current candidate position is skipped and the next candidate position is tested using the time interval constraint. If none of the remaining candidate positions satisfy the time interval constraint, the matching of the current template is stopped.

[0032] Step 6.6, Select Use this as a template and repeat steps 6.3 to 6.5 until all templates have been matched;

[0033] Step 6.7: Select subgraph As the collision area to be processed, and selected Use this as a template, and then repeat steps 6.2 to 6.6 until all collision frequency hopping signals have been processed.

[0034] The beneficial effects of this application are:

[0035] This application introduces a fill factor to quantify the density of connected components within their smallest bounding rectangle, which improves the accuracy of identifying collision frequency hopping signals during clustering. Furthermore, this invention also uses the fill factor for adaptive setting of the template matching threshold, avoiding missed detections under fixed threshold conditions.

[0036] This application uses an overlap rate to measure the degree of overlap between the template and the matched region. By prioritizing candidate positions with lower overlap rates, the repeated placement of the template in the matched region can be effectively suppressed, reducing erroneous matches within the collision region and enabling the collision region to be divided into matching results with the smallest possible overlap.

[0037] This application utilizes time interval constraints to verify the temporal consistency of candidate positions using the start-up timing information of non-collision frequency hopping signals. This constraint can fully leverage the temporal repulsion characteristics of frequency hopping signals to eliminate candidate positions with similar geometry but whose start times do not conform to the frequency hopping timing rules. Attached Figure Description

[0038] Figure 1 This is a flowchart of this application.

[0039] Figure 2 This is a schematic diagram of template traversal of subgraphs.

[0040] Figure 3 This is a comparison chart of the sorting results of this application and traditional methods. Detailed Implementation

[0041] The embodiments of the present invention will be disclosed below with reference to the drawings. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the present invention. That is, in some embodiments of the present invention, these practical details are not essential. In addition, for the sake of simplicity, some conventional structures and components will be shown in the drawings in a simple schematic manner.

[0042] like Figure 1 As shown, this application discloses a collision frequency hopping signal sorting method based on time-frequency feature matching, which specifically includes the following steps:

[0043] Step 1, Adaptive Binarization: Perform adaptive binarization on the input time-frequency graph, and obtain the binary time-frequency graph using the triangular thresholding method. Specifically, the process involves: first, calculating the grayscale histogram of the time-frequency graph; then, using the triangular thresholding method, which involves drawing a line between the main peak and the tail endpoint of the grayscale histogram and selecting the grayscale level with the largest vertical distance from this line as the binarization threshold; and finally, using the binarization threshold to obtain the binary time-frequency graph. .

[0044] Step 2, Opening Operation: Apply the binary time-frequency graph obtained in Step 1. Perform an opening operation. Construct a rectangular struct element. For binary time-frequency plots The opening operation is defined as:

[0045]

[0046] in, This represents the time-frequency graph after the opening operation. This indicates the opening operation. This represents the erosion operation. This indicates the expansion operation.

[0047] Step 3, Connectivity Component Determination: Analyze the time-frequency graph after the opening operation in Step 2. Connectivity determination is performed, using the eight-neighbor connectivity rule and a two-pass scanning algorithm to label the connected components. After labeling the connected components, let the total number of connected components be . .

[0048] Step 4: Parameter Extraction: For each connected component obtained in Step 3, extract the minimum bounding rectangle and obtain the width, height, and fill factor of each connected component. Specifically: For each connected component, extract the minimum bounding rectangle corresponding to the connected component and define the fill factor. :

[0049]

[0050] in, and For the first The width and height of the minimum bounding rectangle of a connected region. , This represents the total number of connected components after the connected component labeling process. For the first The area of ​​a connected region, when When, it means that the connected component is filled with its smallest bounding rectangle. When , it indicates that there is a gap inside the smallest bounding rectangle.

[0051] Step 5: Perform DBSCAN clustering on the extracted time-frequency feature parameters. Here, we choose... , and This is a characteristic of density-based spatial clustering (DBSCAN) clustering algorithm. Through clustering processing, we obtain... Each cluster corresponds to a non-collision frequency hopping signal. and Each noise point corresponds to a collision frequency hopping signal. .

[0052] Step 6, Time-Frequency Feature Matching: Based on the time-frequency features of various non-collision frequency hopping signals obtained from clustering in Step 5, templates are constructed. Sub-graphs and occupancy matrices are generated for collision regions. Combining matching scores, overlap rates, and time interval constraints, the templates are subjected to sliding matching to sequentially complete the separation and classification of each collision frequency hopping signal. This includes the following steps:

[0053] Step 6.1: Average the width and height of each clustered non-collision frequency hopping signal to construct a template. ,in, and These represent the width and height of the template, respectively. For a matrix with all elements equal to 1, use the template Sort by area from largest to smallest, and then index... All follow this sorting.

[0054] Step 6.2: Set the surrounding collision frequency hopping signal The time-frequency diagram of the minimum bounding rectangle after opening operation. The corresponding collision region is the subgraph. Introducing the occupancy matrix To mark subgraphs The template is already in use. The position where a match is successfully found is initially set to match the subgraph. Given matrices of the same size containing all zeros, where... and respectively represent the width and height of the sub - figure;

[0055] Step 6.3, Compare the template with the sub - figure to determine whether the template is available. If the template is much larger than the sub - figure in width or height, skip the current template; otherwise, while keeping the aspect ratio unchanged, simultaneously reduce the width and height of the template until the template can be placed into the sub - figure;

[0056] Step 6.4, Slide the template to traverse the entire sub - figure: As Figure 2 shown, for each position in the sub - figure , calculate the matching score between the template and the sub - figure :

[0057]

[0058] where, , , represents the time index in the sub - figure, represents the frequency index in the sub - figure;

[0059] Calculate the overlap rate between the template and the occupancy matrix :

[0060]

[0061] When the template has finished sliding over the entire sub - figure, obtain the matching scores and overlap rates corresponding to all positions in the sub - figure. If the denominator calculation result of the matching score is zero, discard the corresponding position; to ensure the reliability of time - frequency feature matching, set the threshold of the matching score to the minimum filling factor of the non - colliding frequency - hopping signals of the template corresponding category. After the template has finished sliding over the entire sub - figure, collect all candidate positions that satisfy .

[0062] Step 6.5, According to the overlap rate Sort the candidate positions from smallest to largest; if the overlap rate is... If they are the same, then the matching score will be used. Sort from largest to smallest, the time interval constraint for each candidate position is defined as follows:

[0063]

[0064] in, Indicates the starting index of the candidate position. Indicates the first The set of start-up indices for collision-like frequency hopping signals. for If any element in the cluster satisfies the time interval constraint, the match is considered successful, and the current candidate position is then assigned to the corresponding cluster. and occupy the matrix The element at the corresponding position is set to 1; otherwise, the current candidate position is skipped and the next candidate position is tested using the time interval constraint. If none of the remaining candidate positions satisfy the time interval constraint, the matching of the current template is stopped.

[0065] Step 6.6, Select Use this as a template and repeat steps 6.3 to 6.5 until all templates have been matched;

[0066] Step 6.7: Select subgraph As the collision area to be processed, and selected Use this as a template, and then repeat steps 6.2 to 6.6 until all collision frequency hopping signals have been processed.

[0067] Figure 3 The figure shows a comparison of the sorting performance of this application and traditional methods. The frequency hopping signal parameters used in the simulation are shown in Table 1.

[0068] Table 1 Frequency Hopping Signal Parameters

[0069]

[0070] The signal-to-noise ratio ranges from -5dB to 9dB, and 100 Monte Carlo tests are performed for each signal-to-noise ratio.

[0071] To evaluate the performance of the proposed method, the following three sets of comparison methods were set up: (1) Template matching based on occupancy matrix: DBSCAN was first used to cluster the extracted width and height, and then template matching was performed only based on the matching score and occupancy matrix. When the matching score of a position exceeds 0.8, it is determined to be a valid match. (2) Template matching based on normalized cross-correlation: DBSCAN was also used to cluster the extracted width and height, but template matching only relied on the normalized cross-correlation function to calculate the matching score, and the position with the maximum correlation value was taken as a valid match. (3) MeanShift clustering: The MeanShift algorithm was used to cluster the extracted width and height. Figure 3 It can be seen that the method of this application achieves the best sorting performance. Even under low signal-to-noise ratio conditions, the method of this application still maintains strong sorting ability; as the signal-to-noise ratio increases, the F1-score of the method of this application gradually converges to around 100%, and has small fluctuations.

[0072] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A collision frequency hopping signal sorting method based on time-frequency feature matching, characterized in that: The collision frequency hopping signal sorting method specifically includes the following steps: Step 1, Adaptive Binarization: Perform adaptive binarization on the input time-frequency graph, and obtain the binary time-frequency graph using the triangular thresholding method. ; Step 2, Opening Operation: Apply the binary time-frequency graph obtained in Step 1. Perform an opening operation to obtain the time-frequency diagram after the opening operation. ; Step 3, Connectivity Component Determination: Analyze the time-frequency graph after the opening operation in Step 2. Perform connectivity determination, identify and mark each independent connected component; Step 4, Parameter Extraction: For each connected component obtained in Step 3, extract the minimum bounding rectangle and obtain the width, height and fill factor of each connected component. Step 5: Density-based spatial clustering: Using the width, height, and fill factor of connected components as features, a density-based spatial clustering algorithm is used to classify each connected component. Step 6, Time-Frequency Feature Matching: Based on the time-frequency features of various non-collision frequency hopping signals obtained from clustering in Step 5, templates are constructed. Subgraphs and occupancy matrices are generated for the collision regions. The templates are then slid-matched by combining the matching score, overlap rate, and time interval constraints to sequentially complete the separation and classification of each collision frequency hopping signal.

2. The collision frequency hopping signal sorting method based on time-frequency feature matching according to claim 1, characterized in that: In step 4, for each connected component, the minimum bounding rectangle corresponding to the connected component is extracted, and the fill factor is defined. : in, and The first The width and height of the minimum bounding rectangle of a connected region. , The total number of connected components. For the first The area of ​​a connected region, when When, it means that the connected component is filled with its smallest bounding rectangle. When , it indicates that there is a gap inside the smallest bounding rectangle.

3. The collision frequency hopping signal sorting method based on time-frequency feature matching according to claim 2, characterized in that: In step 5, the width of the connected component is selected. ,high and fill factor As a feature of density-based spatial clustering algorithms, density-based spatial clustering algorithms are used to classify each connected component, resulting in... Each cluster corresponds to a non-collision frequency hopping signal. and Each noise point corresponds to a collision frequency hopping signal. , , .

4. The collision frequency hopping signal sorting method based on time-frequency feature matching according to claim 3, characterized in that: Step 6 specifically includes the following steps: Step 6.1: Process the clustered non-collision frequency hopping signals Calculate the average of the width and height respectively, and construct the template. ,in, and These represent the width and height of the template, respectively. For a matrix with all elements equal to 1, use the template Sort by area from largest to smallest; Step 6.2: Set the surrounding collision frequency hopping signal The time-frequency diagram of the minimum bounding rectangle after opening operation. The corresponding collision region is the subgraph. Introducing the occupancy matrix To mark subgraphs The template is already in use. The position where a match is successfully found is initially set to match the subgraph. Given matrices of the same size containing all zeros, where... and These represent the width and height of the subgraph, respectively. Step 6.3, place the template and the sub-graph to compare to determine whether the template is available. If the template is greater than the width or height of the sub-graph in terms of width or height, skip the current template. Otherwise, while keeping the aspect ratio unchanged, simultaneously reduce the width and height of the template until the template can be placed inside the sub-graph ; Step 6.

4. Slide the template and traverse the entire subgraph: For each position in the subgraph , calculate the matching score between the template and the subgraph : in, , , This represents the time index in the subgraph. This represents the frequency index in the subgraph; Calculation template With occupancy matrix Overlap rate between : When template After sliding across the entire subgraph, the matching scores for all positions in the subgraph are obtained. and overlap rate If the match score If the denominator of the fraction is zero, then the corresponding part is discarded. Step 6.5, according to the overlap rate Sort the candidate positions from smallest to largest; if the overlap rate is... If they are the same, then the matching score will be used. Sort from largest to smallest, the time interval constraint for each candidate position is defined as follows: in, Indicates the starting index of the candidate position. Indicates the first The set of start-up indices for collision-like frequency hopping signals. for If any element in the cluster satisfies the time interval constraint, the match is considered successful, and the current candidate position is then assigned to the corresponding cluster. and occupy the matrix The element at the corresponding position is set to 1; otherwise, the current candidate position is skipped and the next candidate position is tested using the time interval constraint. If none of the remaining candidate positions satisfy the time interval constraint, the matching of the current template is stopped. Step 6.6, Select Use this as a template and repeat steps 6.3 to 6.5 until all templates have been matched; Step 6.7: Select subgraph As the collision area to be processed, and selected Use this as a template, and then repeat steps 6.2 to 6.6 until all collision frequency hopping signals have been processed.

5. The collision frequency hopping signal sorting method based on time-frequency feature matching according to claim 4, characterized in that: In step 6.4, to ensure the reliability of time-frequency feature matching, the matching score is... The threshold is set to the minimum fill factor in the non-collision frequency hopping signals of the corresponding category of the template. ,template After sliding over the entire subgraph, collect all that satisfy the condition. Candidate positions.

Images