Ship communication signal quality detection method capable of performing detection automatically

Abstract

The invention relates to the technical field of ship communication signal processing, and discloses a ship communication signal quality detection method for automatic detection. The method comprises the following steps: hierarchically decomposing a captured original communication signal to obtain a discrete signal data unit group; constructing a spatial distribution map of the signals through topological mapping of the vector features, and identifying a main signal core area according to a density gradient and a vector field streamline in the map; tracking a signal energy diffusion path along a vector field streamline, positioning discontinuous nodes on the path, and aggregating to form a suspected interference signal cluster according to the discontinuous nodes; performing morphological consistency verification on the signal clusters, and separating interference components and effective signal components from the atlas; the effective signal components are recombined to reconstruct a communication waveform, and quality measurement and analysis are performed based on the waveform to generate a quantized quality report. According to the invention, accurate positioning and separation of interference components can be realized, and objectivity and accuracy of ship communication signal quality evaluation are improved.

Description

Technical Field

[0001] This invention relates to the field of ship communication signal processing technology, specifically to an automatic method for detecting the quality of ship communication signals. Background Technology

[0002] Accurate detection of ship communication signal quality is crucial for ensuring safe maritime navigation and reliable information transmission. Existing technologies typically employ time-domain or frequency-domain energy analysis methods based on fixed thresholds, or use pre-set filter banks to distinguish between signals and noise. These methods are effective in relatively stable channel environments and rely on prior statistical assumptions about signals and interference.

[0003] In complex maritime communication environments, signals are frequently affected by multipath fading, sudden impulse noise, and non-stationary co-channel interference, characterized by high aliasing in the time and frequency domains. Conventional methods struggle to dynamically adapt to these complex variations, easily misidentifying strong interference as signal components or incorrectly filtering out distorted valid signals, leading to inaccurate signal integrity assessments. Existing technologies lack in-depth exploration of the diffusion patterns and spatial structure characteristics of signal energy in the joint time and frequency domains.

[0004] A method is needed to accurately remove interference and restore the intrinsic structure of signals from complex aliased signals. This requires a technical solution that can not only identify the core region of the signal, but also trace its natural energy propagation path and accurately locate the anomalies caused by interference along this path, thereby achieving essential separation of interference and signal at the morphological level and improving the objectivity and automation of quality assessment. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic detection method for ship communication signal quality, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides an automatic detection method for ship communication signal quality, the method comprising: The original communication signal is captured by at least one signal processing antenna on the ship, generating a raw signal sequence containing multiple dimensions; The original signal sequence is subjected to hierarchical decomposition to generate discrete signal data unit groups within the intersection dimension of a preset frequency domain sub-band and a preset time domain window; A topological mapping based on vector features is performed on the signal data unit group to construct a spatial distribution map of the signal data; In the spatial distribution map, the core region of the main signal is identified based on the density gradient and vector field streamlines between the signal data units; Using the core region of the main signal as the root node, trace the diffusion path of the signal energy along the vector field streamline and locate the discontinuous nodes that appear on the energy diffusion path; The signal data units of the discontinuous nodes and their neighborhoods are aggregated to form a cluster of suspected interference signals to be determined; A morphological consistency check is performed on the suspected interference signal clusters, and the interference component data and the effective signal component data are separated from the spatial distribution map based on the check results; The effective signal component data are recombined to reconstruct a continuous ship communication signal waveform; Based on the reconstructed ship communication signal waveform, signal integrity measurement and channel distortion analysis are performed. By combining the signal integrity measurement results with the channel distortion analysis results, a quantitative quality report for the original communication signal is generated.

[0007] Preferably, the step of performing hierarchical decomposition on the original signal sequence to generate a discrete group of signal data units within the intersection dimension of a preset frequency domain sub-band and a preset time domain window includes: Set a set of preset frequency domain sub-bands covering the target communication frequency band, and set a set of preset time domain windows with overlapping areas; The original signal sequence is passed sequentially through the filter group corresponding to the preset frequency domain sub-band to obtain a set of sub-band signal sequences; Each sub-band signal sequence is input into the corresponding preset time-domain window for truncation, generating multiple time-frequency two-dimensional signal segments; The mean energy, zero-crossing rate statistic, and spectral centroid coordinates of each signal segment are extracted as the basic feature vector of the signal data unit. Each signal segment and its corresponding basic feature vector are bound together and encapsulated into an independent signal data unit; All the encapsulated signal data units are indexed and arranged according to their time-frequency coordinates to form the discrete signal data unit group.

[0008] Preferably, the step of performing topological mapping based on vector features on the signal data unit group to construct a spatial distribution map of the signal data includes: The basic feature vector of the signal data unit is used as the coordinate point in the multidimensional space; Calculate the geodesic distance between each pair of coordinate points in the multidimensional space, and establish a nearest neighbor network between the coordinate points based on the geodesic distance; Each edge in the nearest neighbor network is assigned a direction, which points from a coordinate point with a lower energy value to a coordinate point with a higher energy value, forming a directed edge; By summing all the coordinate points and all the directed edges, a topological network graph with vector directions is generated; The topological network diagram is projected onto a three-dimensional space with signal strength, frequency offset, and time delay spread as axes to form a visualized spatial distribution map of the signal data.

[0009] Preferably, identifying the main signal core region in the spatial distribution map based on the density gradient and vector field streamlines between signal data units includes: In the three-dimensional space of the spatial distribution map, the cell density value in the neighborhood of the spatial location of each signal data unit is calculated; Based on the cell density values ​​of all signal data units, draw the density isosurface in the three-dimensional space; Identify the high-density region that is closed and has the largest enclosed volume in the density isosurface; Extract the direction data of all directed edges within the high-density region, and generate the vector field streamlines describing the mainstream direction of signal energy through integration. The spatial sub-region with the highest convergence of vector field streamlines within the high-density region is designated as the main signal core region.

[0010] Preferably, the step of tracing the diffusion path of signal energy along the vector field streamlines, using the main signal core region as the root node, and locating discontinuous nodes appearing on the energy diffusion path includes: Starting from the geometric center point of the main signal core region, trace the vector field streamlines in both the positive and negative directions; Record the spatial coordinates of each signal data unit through which the vector field streamline passes, forming multiple energy diffusion paths; Calculate the characteristic distance between adjacent signal data units on each of the energy diffusion paths; When multiple intervals with a feature distance greater than a preset jump threshold appear consecutively on a certain energy diffusion path, it is determined that there is a path break in this segment of the energy diffusion path; Mark the signal data units on both sides of the path break as path endpoints; The path endpoints that are not directly connected to any other path endpoints on the energy diffusion path are selected and defined as the non-continuous nodes.

[0011] Preferably, the aggregation of signal data units within the discontinuous nodes and their neighborhoods to form a cluster of suspected interference signals to be determined includes: A spherical neighborhood is defined in the spatial distribution map, centered on each of the non-continuous nodes and bounded by a preset clustering radius. Collect all signal data units that fall within each of the spherical neighborhoods; For each collected signal data unit, calculate its similarity to the central discontinuous node in the feature vector space; Retain signal data units with a similarity greater than a preset similarity threshold, and discard signal data units with a similarity less than or equal to the preset similarity threshold; Each of the non-contiguous nodes is merged with all the signal data units reserved for it, forming an independent candidate set; A union operation is performed on multiple candidate sets whose spatial distance is less than a preset merging threshold. The merged set is the whole set of suspected interference signals to be determined.

[0012] Preferably, the step of performing morphological consistency verification on the suspected interference signal cluster, and separating the interference component data and the valid signal component data from the spatial distribution map based on the verification result, includes: For each of the suspected interference signal clusters, extract the spatial coordinates of all signal data units on its boundary contour; Based on the extracted spatial coordinates, the ratio of the convex hull volume to the surface area of ​​the suspected interference signal cluster is calculated as its morphological compactness parameter. Analyze the directional distribution of the directed edges within the suspected interference signal cluster and calculate its directional entropy value; Set a binary decision threshold for the morphological compactness parameter and the directional entropy value; The morphological compactness parameter and the directional entropy value of the suspected interference signal cluster are input into the binary decision threshold for comparison; If the comparison result meets the interference feature criteria, the suspected interference signal cluster is determined to be an interference component, and all signal data units contained therein are marked as interference component data. If the comparison result does not meet the interference feature criterion, the suspected interference signal cluster is determined to be a distorted part of the valid signal, and all signal data units contained therein are marked as valid signal component data. The signal data units marked as interference components are removed from the spatial distribution map.

[0013] Preferably, the reassembly of the effective signal component data to reconstruct a continuous ship communication signal waveform includes: In the spatial distribution map after removing the interference component data, only the signal data units marked as valid signal component data are retained; Based on the time-frequency coordinates of the retained effective signal component data, it is restored to the corresponding time-frequency two-dimensional signal segment; All the restored time-frequency two-dimensional signal segments are spliced ​​together according to their time-frequency coordinates to fill the gaps that may be caused by the removal of interference components. The gaps are filled by bilinear interpolation of adjacent segments. An inverse transformation is performed on the spliced ​​and filled complete time-frequency matrix to synthesize the continuous ship communication signal waveform in the time domain.

[0014] Preferably, the step of performing signal integrity measurement and channel distortion analysis based on the reconstructed ship communication signal waveform includes: Align the reconstructed ship communication signal waveform with the preset ideal reference signal waveform in the time domain; The sum of squared point-by-point errors between the reconstructed ship communication signal waveform after alignment and the ideal reference signal waveform is calculated as the waveform fidelity loss. The reconstructed ship communication signal waveform is subjected to segmented autocorrelation calculation to obtain the distribution of its multipath delay spread; Based on the distribution of the multipath delay spread, calculate the root mean square delay spread parameter and coherence bandwidth parameter of the channel; The waveform fidelity loss, the root mean square delay spread parameter, and the coherence bandwidth parameter together constitute the signal integrity measurement result and the channel distortion analysis result.

[0015] Preferably, the step of combining the signal integrity metric results and the channel distortion analysis results to generate a quantitative quality report for the original communication signal includes: A first weighting coefficient is set for the waveform fidelity loss, a second weighting coefficient is set for the root mean square delay spread parameter, and a third weighting coefficient is set for the coherence bandwidth parameter. The waveform fidelity loss is weighted using the first weighting coefficient to obtain a fidelity score item; The root mean square delay spread parameter is weighted using the second weighting coefficient, and a delay scoring term is used. The coherent bandwidth parameters are weighted using the third weighting coefficient to obtain the bandwidth score item; The fidelity score, the delay score, and the bandwidth score are added together to obtain the overall signal quality score. The signal quality comprehensive score, the waveform fidelity loss, the root mean square delay spread parameter, the coherence bandwidth parameter, and the identification information of the original communication signal are integrated into a structured data record. The structured data record is output as a quantization quality report for the original communication signal.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By performing topological mapping on discrete signal data units based on vector features, a spatial distribution map of the signal data is constructed, and the core region of the main signal is identified based on the density gradient and vector field streamlines. This method transforms the energy distribution of the signal in the time-frequency domain into a map with spatial topological relationships. The identification of the core region depends on the aggregation pattern of the data units in this space and the macroscopic direction of the vector field, rather than a threshold judgment of a single-dimensional energy value. This enables differentiation based on the essential differences in their spatial structures in complex scenarios where the time-frequency distributions of signals, noise, and interference overlap. This technique avoids the insufficient adaptability of traditional methods that rely on fixed thresholds or typical interference templates, thus improving the accuracy of core signal component extraction.

[0017] Starting from the identified core region of the main signal, the energy diffusion path is traced along the vector field streamlines. Discontinuous nodes on the path are located, and their neighborhoods are aggregated to form clusters of suspected interference signals for morphological consistency verification. This process establishes the analysis path based on the continuity and directionality of signal energy diffusion in the time and frequency domains. Discontinuous nodes indicate abnormal interruptions in the path, typically corresponding to injection points of external interference or distortion points in the signal itself. By analyzing the morphological characteristics of the clusters formed by these nodes, interference components with structural differences can be distinguished from signal distortion caused by channel distortion. This method, based on the path continuity of signal energy propagation and the morphological characteristics of signal components, reduces reliance on prior knowledge of interference and achieves more refined separation of mixed signals. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the automatic detection method for ship communication signal quality according to the present invention. Figure 2 A flowchart for generating a group of signal data units; Figure 3 A flowchart for identifying the core region of the main signal; Figure 4 A bar chart comparing the quality indicators of interpolation strategies for reconstructing ship communication signals; Figure 5 A bar chart showing the weighting of signal quality indicators for different communication scenarios on ships. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides an automatic method for detecting the quality of ship communication signals. The method includes: capturing raw communication signals using at least one signal processing antenna on the ship to generate a multi-dimensional raw signal sequence; performing hierarchical decomposition on the raw signal sequence to generate discrete signal data unit groups within the intersection dimension of a preset frequency domain sub-band and a preset time domain window; performing topological mapping based on vector features on the signal data unit groups to construct a spatial distribution map of the signal data; identifying the main signal core region in the spatial distribution map based on the density gradient and vector field streamlines between the signal data units; tracing the signal energy diffusion path along the vector field streamlines with the main signal core region as the root node to locate discontinuous nodes appearing on the energy diffusion path; aggregating the signal data units in the discontinuous nodes and their neighborhoods to form suspected interference signal clusters to be judged; performing morphological consistency verification on the suspected interference signal clusters, and separating interference component data and valid signal component data from the spatial distribution map based on the verification results; reconstructing the continuous ship communication signal waveform by reassembling the valid signal component data; and performing signal integrity measurement and channel distortion analysis based on the reconstructed ship communication signal waveform. By combining the signal integrity measurement results with the channel distortion analysis results, a quantitative quality report for the original communication signal is generated.

[0021] In one embodiment of the present invention, see [reference] Figure 2 A set of preset frequency domain sub-bands covering the target communication frequency band is defined, along with a set of preset time domain windows with overlapping regions. The original signal sequence is sequentially passed through the filter banks corresponding to the preset frequency domain sub-bands to obtain a set of sub-band signal sequences. Each sub-band signal sequence is input into its corresponding preset time domain window for truncation, generating multiple time-frequency two-dimensional signal segments. The mean energy, zero-crossing rate statistics, and spectral centroid coordinates of each signal segment are extracted as the basic feature vectors of the signal data unit. Each signal segment and its corresponding basic feature vector are bound together and encapsulated into an independent signal data unit. All encapsulated signal data units are indexed and arranged according to their time-frequency coordinates to form a discrete signal data unit group.

[0022] The fundamental feature vectors of signal data units are used as coordinate points in a multi-dimensional space. The geodesic distance between each pair of coordinate points in the multi-dimensional space is calculated, and a nearest neighbor network is established based on the geodesic distance. Each edge in the nearest neighbor network is assigned a direction, pointing from a coordinate point with lower energy value to a coordinate point with higher energy value, forming directed edges. All coordinate points and all directed edges are summarized to generate a topological network graph with vector directions. This topological network graph is projected onto a three-dimensional space with signal strength, frequency offset, and time delay spread as axes, forming a visualized spatial distribution map of the signal data.

[0023] In practice, at least one signal processing antenna on the ship captures a raw signal from a VHF communication band containing complex background noise and impulse interference, generating a raw signal sequence with multiple dimensions including amplitude, phase, and instantaneous frequency. To achieve refined analysis of the raw signal sequence, a hierarchical decomposition process is performed. A set of preset frequency domain sub-bands covering the 156MHz to 162MHz communication band is defined, with eight sub-bands and each sub-band bandwidth of 750kHz. Simultaneously, a set of preset time-domain windows with overlapping regions is defined, with a window length of 256 sampling points and an overlap rate of 50%. The raw signal sequence is sequentially passed through bandpass filter banks corresponding to the eight preset frequency domain sub-bands, resulting in eight sub-band signal sequences. Each sub-band signal sequence is input into its corresponding preset time-domain window for truncation, generating multiple time-frequency two-dimensional signal segments based on a sliding window mechanism.

[0024] The mean energy, zero-crossing rate statistic, and spectral centroid coordinates of each signal segment are extracted as the fundamental feature vectors of the signal data unit. The mean energy of the signal segment reflects the local signal strength, the zero-crossing rate statistic reflects the complexity of the signal frequency components, and the spectral centroid coordinates indicate the concentrated location of signal energy in the frequency domain. Each signal segment and its corresponding fundamental feature vector are bound and encapsulated into an independent signal data unit. All encapsulated signal data units are indexed and arranged according to their time-frequency coordinates, which are determined by the preset frequency domain sub-band number and the start time point of the preset time domain window, forming a discrete signal data unit group. A topological mapping based on vector features is performed on the discrete signal data unit group. The fundamental feature vectors of the signal data units serve as coordinate points in a multi-dimensional space, the dimension of which is equal to the number of components of the fundamental feature vector. The geodesic distance between each pair of coordinate points in the multi-dimensional space is calculated. The geodesic distance reflects the true proximity relationship of the coordinate points in the feature space, and a nearest neighbor network is established based on the geodesic distance. In the nearest neighbor network, each coordinate point is connected only to the K coordinate points with the smallest geodesic distance. Each edge in the nearest neighbor relationship network is assigned a direction, which points from the coordinate point with the lower energy mean to the coordinate point with the higher energy mean, forming a directed edge. The energy mean represents the signal strength attribute of the signal data unit.

[0025] By summing all coordinate points and all directed edges, a topological network diagram with vector directions is generated. This diagram depicts the distribution structure and energy flow trend of signal data units in the feature space. The topological network diagram is projected onto a three-dimensional space with signal strength, frequency offset, and time delay spread as axes. During projection, the signal strength axis corresponds to the mean energy component of the coordinate points, the frequency offset axis corresponds to the spectral centroid component of the coordinate points, and the time delay spread axis is derived from the time span of the signal segment to which the coordinate point belongs in the original sequence, forming a visualized spatial distribution map of the signal data. In some embodiments, the formula for calculating the geodesic distance between coordinate points is:

[0026] in: Represents coordinate points with coordinate points Geodetic distance between them Indicates from coordinate point to coordinate point The shortest path The feature vectors of the coordinate points along the route, Describes the Euclidean norm. This represents the total number of coordinate points traversed along the shortest path.

[0027] It is understood that the number and bandwidth of the preset frequency domain sub-bands can be adjusted according to the specific standard of the target communication signal, and the length and overlap rate of the preset time domain window can also be configured according to the non-stationary characteristics of the signal. In some embodiments, the basic feature vector may also include the spectral flatness and peak-to-average power ratio of the signal segment to enrich the descriptive dimensions of the signal data unit.

[0028] In one embodiment of the present invention, see [reference] Figure 3 In the three-dimensional space of the spatial distribution map, the cell density value within the neighborhood of each signal data unit is calculated. Based on the cell density values ​​of all signal data units, density isosurfaces are plotted in three-dimensional space. The high-density regions with the largest closed enclosing volume within the density isosurfaces are identified. The direction data of all directed edges within the high-density regions are extracted, and vector field streamlines describing the mainstream energy direction of the signal are generated through integration. The spatial sub-region with the highest convergence of vector field streamlines within the high-density region is marked as the main signal core region.

[0029] Starting from the geometric center of the main signal core region, vector field streamlines are traced in both forward and reverse directions. The spatial coordinates of each signal data unit traversed by the vector field streamlines are recorded, forming multiple energy diffusion paths. The characteristic distance between adjacent signal data units on each energy diffusion path is calculated. When multiple consecutive characteristic distances exceeding a preset jump threshold occur on an energy diffusion path, a path break is determined. The signal data units on either side of the path break are marked as path endpoints. Path endpoints that are not directly connected to any other path endpoints are selected and defined as discontinuous nodes.

[0030] In practical implementation, for a constructed three-dimensional spatial distribution map with signal strength, frequency offset, and time delay spread as axes, it contains a large number of signal data units representing different signal segments. The process of identifying the core region of the main signal is based on the analysis of the spatial aggregation state of the signal data units. In the three-dimensional space of the spatial distribution map, the unit density value in the neighborhood of each signal data unit is calculated. During the calculation, a spherical space with a fixed radius is set as the neighborhood, centered on the spatial coordinates of the signal data unit. The total number of signal data units falling into this spherical space is counted and divided by the volume of the spherical space to obtain the unit density value at that location. Based on the unit density values ​​of all signal data units, density isosurfaces in three-dimensional space are drawn. Density isosurfaces are surfaces formed by connecting all spatial points with the same unit density value. The high-density region with the largest closed enclosing volume in the density isosurface is identified, specifically by comparing the spatial volume enclosed by the closed isosurfaces generated under different density thresholds.

[0031] The directional data of all directed edges within the identified high-density region are extracted. These directed edges are defined by the energy direction in the nearest neighbor network. Vector field streamlines describing the mainstream energy direction of the signal are generated through integration. The integration operation starts from multiple random starting points within the high-density region and performs path tracing and smooth connection along the directions indicated by the directed edges. The spatial sub-region with the highest convergence of vector field streamlines within the high-density region is marked as the main signal core region. The convergence of vector field streamlines is quantified by calculating the difference between the number of vector field streamlines flowing into and out of this sub-region. In some embodiments, the calculation of the unit density value adopts a kernel function-based method. In this kernel function-based unit density value calculation method, the specific implementation of the kernel function involves weighted density estimation of the spatial distribution of signal data units. The kernel function, as a non-negative weighting function, is used to calculate the contribution of each signal data unit in the neighborhood of the target spatial location. The Gaussian kernel function is a commonly used implementation method, which assigns weights based on the Euclidean distance between the spatial coordinates of the signal data unit and the target point; the closer the unit, the higher its weight value. The bandwidth parameter defines the spatial range of the weighted neighborhood. Its value directly affects the resolution and smoothness of the density estimation. By appropriately setting the bandwidth parameter, the signal density characteristics under different ship communication scenarios can be adapted, thereby accurately calculating the cell density value at each location. The formula is expressed as:

[0032] in: Indicates spatial location The cell density value at that location, This indicates the total number of signal data units. Indicates the first The spatial coordinate vector of each signal data unit Represents Euclidean distance. It is a non-negative kernel function. It is a bandwidth parameter that controls the neighborhood range.

[0033] Starting from the geometric center of the main signal core region, the vector field streamlines are traced in both forward and reverse directions. Forward tracing follows the energy increase direction indicated by the directed edge, while reverse tracing proceeds against the direction indicated by the directed edge. The spatial coordinates of each signal data unit traversed by the vector field streamlines are recorded, forming multiple energy diffusion paths. The characteristic distance between adjacent signal data units on each energy diffusion path is calculated. The characteristic distance refers to the Euclidean distance between two adjacent signal data units in their fundamental eigenvector space. When multiple consecutive intervals with characteristic distances greater than a preset jump threshold appear on an energy diffusion path, it is determined that this energy diffusion path segment has a path break. The preset jump threshold is set based on the statistical values ​​of the distance distribution throughout the entire eigenvector space.

[0034] The signal data units on both sides of the path break are marked as path endpoints; one path break generates two path endpoints. Path endpoints on all energy diffusion paths that are not directly connected to any other path endpoints are defined as discontinuous nodes. Direct connection means that there is a directed edge between the two path endpoints or that they belong to the same continuous vector field streamline. It is understood that the specific value of the preset jump threshold can be adjusted based on prior knowledge of the signal environment; a smaller preset jump threshold can be used in a stable channel environment to improve detection sensitivity. In some embodiments, a maximum step size limit can be set for tracking the energy diffusion path; the tracking will automatically terminate when the path length exceeds the maximum step size limit to avoid meaningless tracking in areas with weak signal energy. Optionally, in addition to Euclidean distance, Mahalanobis distance can also be used to calculate the feature distance to consider the correlation between feature dimensions. It is understood that the identification of discontinuous nodes depends on the determination of path breaks, and the determination of path breaks is strictly based on the comparison between the feature distance and the preset jump threshold, avoiding subjective judgment.

[0035] In one embodiment of the present invention, a spherical neighborhood is defined in the spatial distribution map, centered on each discontinuous node and bounded by a preset clustering radius. All signal data units falling within each spherical neighborhood are collected. For each collected signal data unit, its similarity to the central discontinuous node in the feature vector space is calculated. Signal data units with a similarity greater than a preset similarity threshold are retained, while those with a similarity less than or equal to the preset similarity threshold are discarded. Each discontinuous node is merged with all the retained signal data units to form an independent candidate set. A union operation is performed on multiple candidate sets whose spatial distance is less than a preset merging threshold; the merged set constitutes a suspected interference signal cluster to be determined.

[0036] For each suspected interference signal cluster, the spatial coordinates of all signal data units on its boundary contour are extracted. Based on the extracted spatial coordinates, the ratio of the convex hull volume to the surface area of ​​the suspected interference signal cluster is calculated as its morphological compactness parameter. The directional distribution of the directed edges inside the suspected interference signal cluster is analyzed, and its directional entropy value is calculated. A binary decision threshold is set for the morphological compactness parameter and the directional entropy value. The morphological compactness parameter and the directional entropy value of the suspected interference signal cluster are compared with the binary decision threshold. If the comparison result meets the interference feature criteria, the suspected interference signal cluster is determined to be an interference component, and all signal data units contained in it are marked as interference component data. If the comparison result does not meet the interference feature criteria, the suspected interference signal cluster is determined to be a distorted part of the valid signal, and all signal data units contained in it are marked as valid signal component data. The signal data units marked as interference component data are removed from the spatial distribution map.

[0037] In practice, based on the identified discontinuous nodes, the aggregation operation aims to group discrete signal components belonging to the same interference source. A spherical neighborhood is defined in the spatial distribution map, centered on each discontinuous node and bounded by a preset clustering radius. The preset clustering radius is set based on the average distribution density of signal data units in space. All signal data units falling within each spherical neighborhood are collected. This collection process involves traversing the spatial coordinates of all signal data units in the spatial distribution map and calculating their Euclidean distance to the coordinates of the central discontinuous node. For each collected signal data unit, its similarity to the central discontinuous node in the feature vector space is calculated. The feature vector space is composed of the basic feature vectors of the signal data units, and cosine similarity is used for similarity calculation. Signal data units with a similarity greater than a preset similarity threshold are retained, while those with a similarity less than or equal to the preset similarity threshold are removed. The preset similarity threshold controls the degree of homogeneity in features between the signal components included in the aggregation range and the core discontinuous node. Each discontinuous node is then merged with all the signal data units retained for it, forming an independent candidate set. A union operation is performed on multiple candidate sets whose spatial distance is less than a preset merging threshold. The spatial distance is defined as the shortest Euclidean distance between all signal data units in two candidate sets. The merged whole is a suspected interference signal cluster to be judged.

[0038] For each suspected interference signal cluster to be judged, morphological consistency verification is used to determine its attributes. The spatial coordinates of all signal data units on the boundary contour of the suspected interference signal cluster are extracted. The boundary contour is obtained by calculating the convex hull of the spatial coordinates of all signal data units in the suspected interference signal cluster. Based on the extracted spatial coordinates, the ratio of the convex hull volume to the surface area of ​​the suspected interference signal cluster is calculated as its morphological compactness parameter. A higher morphological compactness parameter value indicates that the suspected interference signal cluster's spatial distribution tends towards a compact spherical shape. The directional distribution of directed edges inside the suspected interference signal cluster is analyzed, and its directional entropy value is calculated. The directional entropy value is used to quantify the degree of disorder in the energy flow direction inside the suspected interference signal cluster. A binary decision threshold is set for the morphological compactness parameter and the directional entropy value. The binary decision threshold is jointly defined by the lower limit of the morphological compactness parameter and the upper limit of the directional entropy value. The morphological compactness parameter and the directional entropy value of the suspected interference signal cluster are input into the binary decision threshold for comparison. If the comparison result meets the interference characteristic criteria, i.e., the morphological compactness parameter is below the lower limit and the directional entropy value is above the upper limit, then the suspected interference signal cluster is determined to be an interference component, and all signal data units contained therein are marked as interference component data. If the comparison result does not meet the interference characteristic criteria, then the suspected interference signal cluster is determined to be a distorted part of the valid signal, and all signal data units contained therein are marked as valid signal component data. The signal data units marked as interference component data are removed from the spatial distribution map. The removal operation directly deletes the records of these signal data units and their associated directed edges in all data structures. In some embodiments, the directional entropy value... The calculation formula is defined as follows:

[0039] in: Represents the directional entropy value. This represents the total number of azimuth intervals into which the three-dimensional directional space is uniformly divided. This indicates that the directed edge direction within the suspected interference signal cluster falls into the first... The probability of each directional interval.

[0040] It is understandable that the specific numerical combinations of the preset clustering radius, preset similarity threshold, and preset merging threshold need to adapt to the diffusion and distribution characteristics of interference signals under different ship communication scenarios. In some embodiments, the alpha shape algorithm can be used instead of the convex hull algorithm for boundary contour extraction in order to capture the non-convex contour features of suspected interference signal clusters. Optionally, other geometric measures can be used to calculate the morphological compactness parameter, such as the ratio of the eigenvalues ​​of the covariance matrix of the coordinates of signal data units in the suspected interference signal cluster. It is understandable that the lower and upper limits of the binary decision threshold need to be determined by analyzing the prior statistical differences in morphological and directional characteristics between typical interference signals and effective signal distortions.

[0041] In one embodiment of the present invention, in the spatial distribution map after removing interference component data, only signal data units marked as valid signal component data are retained. Based on the time-frequency coordinates of the retained valid signal component data, they are restored to corresponding time-frequency two-dimensional signal segments. All restored time-frequency two-dimensional signal segments are spliced ​​together according to their time-frequency coordinates to fill the gaps caused by the removal of interference component data; the gaps are filled by bilinear interpolation of adjacent segments. An inverse transform is performed on the spliced ​​and filled complete time-frequency matrix to synthesize a continuous ship communication signal waveform in the time domain.

[0042] The reconstructed ship communication signal waveform is aligned with a preset ideal reference signal waveform in the time domain. The sum of squared point-by-point errors between the aligned reconstructed ship communication signal waveform and the ideal reference signal waveform is calculated as the waveform fidelity loss. Segmented autocorrelation is calculated on the reconstructed ship communication signal waveform to obtain its multipath delay spread distribution. Based on the multipath delay spread distribution, the root mean square delay spread parameter and coherence bandwidth parameter of the channel are calculated. The waveform fidelity loss, root mean square delay spread parameter, and coherence bandwidth parameter together constitute the signal integrity measurement result and the channel distortion analysis result.

[0043] In practice, after removing interference components, only signal data units marked as valid signal components are retained in the spatial distribution map. These valid signal components are time-frequency components that have been determined not to be interference after morphological consistency verification. Based on the time-frequency coordinates of the retained valid signal components, each valid signal component is restored to its corresponding time-frequency two-dimensional signal segment. The restoration process relies on the original signal segment data stored during signal data unit encapsulation, as well as its preset frequency domain sub-band and preset time domain window information. All restored time-frequency two-dimensional signal segments are then spliced ​​together according to their time-frequency coordinates, which are determined by the frequency domain sub-band index and the start time point of the time domain window, forming a complete time-frequency matrix. The splicing process needs to fill the gaps caused by the removal of interference components. The gaps are filled by bilinear interpolation of adjacent segments. Adjacent segments refer to the valid time-frequency two-dimensional signal segments in the four directions (up, down, left, and right) that are closest to the gap position on the time-frequency coordinates. An inverse transformation is performed on the spliced ​​and filled complete time-frequency matrix. The inverse transformation corresponds to the reverse process of the forward time-frequency analysis transformation used to generate the signal segment, and a continuous ship communication signal waveform in the time domain is synthesized.

[0044] Based on the reconstructed ship communication signal waveform, signal integrity measurement and channel distortion analysis are performed. The reconstructed ship communication signal waveform is aligned with a preset ideal reference signal waveform in the time domain. This alignment is achieved by calculating the cross-correlation function of the two waveforms and finding the peak position. The sum of squared point-by-point errors between the aligned reconstructed ship communication signal waveform and the ideal reference signal waveform is calculated as the waveform fidelity loss, reflecting the overall distortion degree of the reconstructed waveform relative to the ideal waveform. Segmented autocorrelation calculations are performed on the reconstructed ship communication signal waveform, dividing the waveform into multiple continuous or overlapping time segments. The signal autocorrelation function for each time segment is calculated to obtain the distribution of multipath delay spread. Based on the distribution of multipath delay spread, the root mean square delay spread parameter and coherence bandwidth parameter of the channel are calculated. The root mean square delay spread parameter quantifies the delay dispersion caused by multipath propagation, and the coherence bandwidth parameter characterizes the frequency range where the channel frequency response has strong correlation. The waveform fidelity loss, root mean square delay spread parameter, and coherence bandwidth parameter together constitute the signal integrity measurement result and the channel distortion analysis result. In some embodiments, the coherence bandwidth parameter... One calculation method is based on the root mean square delay spread parameter. The formula is:

[0045] in: This represents the coherence bandwidth parameter. This represents the calculated root mean square time delay spread parameter. It is a proportionality constant related to the channel fading characteristics. For a typical Rayleigh fading channel, The value is typically 5. It can be understood that the ideal reference signal waveform can be generated locally based on the standard modulation template of the specific ship communication protocol. In some embodiments, referring to Table 1, the distribution of multipath delay spread can be specifically represented by a discrete power delay spectrum table.

[0046] Table 1: Example Multipath Component Power Delay Distribution

[0047] Optionally, waveform fidelity loss can be calculated across the frequency band to focus on signal quality assessment for the core communication frequency band. It is understood that in piecewise autocorrelation calculations, the choice of piece length needs to strike a balance between time-domain resolution and the stability of the autocorrelation function estimation.

[0048] See Figure 4This is a bar chart comparing the quality indicators of ship communication signal reconstruction interpolation strategies. It shows the performance of different interpolation strategies on two core indicators: "waveform fidelity loss" and "root mean square delay spread." The lower the values ​​of these two indicators, the better the signal reconstruction effect of the interpolation strategy. This chart is a quantitative tool for selecting ship communication signal reconstruction strategies, intuitively presenting the "quality-efficiency" trade-off between different interpolation strategies, avoiding experience-based selection. It supports technical personnel in quickly matching the optimal strategy based on the actual needs of ship communication (such as signal accuracy and computing resources). It reflects the business logic that "interpolation algorithm is the core influencing factor of signal reconstruction quality," providing data basis for the parameter optimization of communication systems.

[0049] In one embodiment of the present invention, a first weighting coefficient is set for the waveform fidelity loss, a second weighting coefficient is set for the root mean square delay spread parameter, and a third weighting coefficient is set for the coherence bandwidth parameter. The waveform fidelity loss is weighted using the first weighting coefficient to obtain a fidelity score. The root mean square delay spread parameter is weighted using the second weighting coefficient to obtain a delay score. The coherence bandwidth parameter is weighted using the third weighting coefficient to obtain a bandwidth score. The fidelity score, delay score, and bandwidth score are added together to obtain a comprehensive signal quality score. The comprehensive signal quality score, waveform fidelity loss, root mean square delay spread parameter, coherence bandwidth parameter, and identification information of the original communication signal are integrated into a structured data record. The structured data record is output as a quantization quality report for the original communication signal.

[0050] In practice, generating a quantitative quality report by integrating signal integrity measurement results and channel distortion analysis results is a systematic data integration and evaluation process. Signal integrity measurement results include waveform fidelity loss, while channel distortion analysis results include root mean square delay spread (RMS) and coherence bandwidth parameters. A first weighting coefficient is assigned to the waveform fidelity loss, a second weighting coefficient to the RMS delay spread parameter, and a third weighting coefficient to the coherence bandwidth parameter. The values ​​of these weighting coefficients are determined based on the sensitivity priorities of different dimensions of signal quality for different ship communication services. For example, in VHF voice communication scenarios, waveform fidelity can be given more attention, while in data link communication scenarios, both delay spread and bandwidth characteristics need to be balanced.

[0051] The waveform fidelity loss is weighted using a first weighting coefficient to obtain a fidelity score. Before weighting, the waveform fidelity loss is typically normalized to eliminate the influence of dimensions. The root mean square delay spread parameter is weighted using a second weighting coefficient to obtain a delay score. The root mean square delay spread parameter is also normalized before weighting. The coherence bandwidth parameter is weighted using a third weighting coefficient to obtain a bandwidth score. The fidelity score, delay score, and bandwidth score are added together to obtain a comprehensive signal quality score, which provides a single numerical quality evaluation from 0 to 100 or a similar range. In some embodiments, the comprehensive signal quality score... The calculation uses the following linear weighted model:

[0052] in: This represents the overall signal quality score. This indicates the amount of waveform fidelity loss. This represents the root mean square delay spread parameter. This represents the coherence bandwidth parameter. It is a normalization function that maps the original parameters to a 0-1 rating interval. This represents the first weighting coefficient. This represents the second weighting coefficient. Denotes the third weight coefficient, and satisfies .

[0053] The signal quality comprehensive score, waveform fidelity loss, root mean square delay spread parameter, coherence bandwidth parameter, and the identification information of the original communication signal are integrated into a structured data record. The identification information of the original communication signal includes timestamp, geographical location, signal frequency, and signal standard. The structured data record is output as a quantitative quality report for the original communication signal. The report can be presented in JSON, XML, or a specific binary format to facilitate automatic parsing. It is understood that the specific values ​​of the first, second, and third weighting coefficients need to be determined according to the communication system design specifications or quality requirements conforming to maritime communication standards. In some embodiments, the normalization function... This can involve linear or non-linear scaling of the original parameters, such as setting an ideal upper limit and an unacceptable lower limit for waveform fidelity loss, and performing a linear mapping within this range. Optionally, the quantification quality report may also include a qualitative evaluation level, such as "Excellent," "Good," "Average," or "Poor," based on a preset range in which the overall signal quality score falls.

[0054] See Figure 5This is a bar chart showing the weighted allocation of signal quality indicators under different communication scenarios on ships. It illustrates the priority given to three core indicators: waveform fidelity, root mean square delay spread, and coherence bandwidth, across different communication scenarios. This chart serves as a scenario-based tool for assessing ship communication signal quality, clearly defining the priority of indicators for different communication scenarios and avoiding a "one-size-fits-all" approach to quality assessment. It supports the signal detection system in dynamically adjusting the weights of the assessment algorithm based on the current communication scenario, improving the accuracy of the assessment results. It reflects the logic that "the business needs of the communication scenario determine the focus of signal quality assessment," providing a basis for the intelligent optimization of ship communication systems.

[0055] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for automatically detecting the quality of ship communication signals, characterized in that, include: The original communication signal is captured by at least one signal processing antenna on the ship, generating a raw signal sequence containing multiple dimensions; The original signal sequence is subjected to hierarchical decomposition to generate discrete signal data unit groups within the intersection dimension of a preset frequency domain sub-band and a preset time domain window; A topological mapping based on vector features is performed on the signal data unit group to construct a spatial distribution map of the signal data; In the spatial distribution map, the core region of the main signal is identified based on the density gradient and vector field streamlines between the signal data units; Using the core region of the main signal as the root node, trace the diffusion path of the signal energy along the vector field streamline and locate the discontinuous nodes that appear on the energy diffusion path; The signal data units of the discontinuous nodes and their neighborhoods are aggregated to form a cluster of suspected interference signals to be determined; A morphological consistency check is performed on the suspected interference signal clusters, and the interference component data and the effective signal component data are separated from the spatial distribution map based on the check results; The effective signal component data are recombined to reconstruct a continuous ship communication signal waveform; Based on the reconstructed ship communication signal waveform, signal integrity measurement and channel distortion analysis are performed. By combining the signal integrity measurement results with the channel distortion analysis results, a quantitative quality report for the original communication signal is generated.

2. The method for automatically detecting the quality of ship communication signals according to claim 1, characterized in that, The step of performing hierarchical decomposition on the original signal sequence to generate a discrete group of signal data units within the intersection dimension of a preset frequency domain sub-band and a preset time domain window includes: Set a set of preset frequency domain sub-bands covering the target communication frequency band, and set a set of preset time domain windows with overlapping areas; The original signal sequence is passed sequentially through the filter group corresponding to the preset frequency domain sub-band to obtain a set of sub-band signal sequences; Each sub-band signal sequence is input into the corresponding preset time-domain window for truncation, generating multiple time-frequency two-dimensional signal segments; The mean energy, zero-crossing rate statistic, and spectral centroid coordinates of each signal segment are extracted as the basic feature vector of the signal data unit. Each signal segment and its corresponding basic feature vector are bound together and encapsulated into an independent signal data unit; All the encapsulated signal data units are indexed and arranged according to their time-frequency coordinates to form the discrete signal data unit group.

3. The method for automatically detecting the quality of ship communication signals according to claim 2, characterized in that, The step of performing topological mapping based on vector features on the group of signal data units to construct a spatial distribution map of the signal data includes: The basic feature vector of the signal data unit is used as the coordinate point in the multidimensional space; Calculate the geodesic distance between each pair of coordinate points in the multidimensional space, and establish a nearest neighbor network between the coordinate points based on the geodesic distance; Each edge in the nearest neighbor network is assigned a direction, which points from a coordinate point with a lower energy value to a coordinate point with a higher energy value, forming a directed edge; By summing all the coordinate points and all the directed edges, a topological network graph with vector directions is generated; The topological network diagram is projected onto a three-dimensional space with signal strength, frequency offset, and time delay spread as axes to form a visualized spatial distribution map of the signal data.

4. The method for automatically detecting the quality of ship communication signals according to claim 1, characterized in that, The process of identifying the core region of the main signal in the spatial distribution map based on the density gradient and vector field streamlines between signal data units includes: In the three-dimensional space of the spatial distribution map, the cell density value in the neighborhood of the spatial location of each signal data unit is calculated; Based on the cell density values ​​of all signal data units, draw the density isosurface in the three-dimensional space; Identify the high-density region that is closed and has the largest enclosed volume in the density isosurface; Extract the direction data of all directed edges within the high-density region, and generate the vector field streamlines describing the mainstream direction of signal energy through integration. The spatial sub-region with the highest convergence of vector field streamlines within the high-density region is designated as the main signal core region.

5. The method for automatically detecting the quality of ship communication signals according to claim 4, characterized in that, The step of tracing the diffusion path of signal energy along the vector field streamlines, with the main signal core region as the root node, and locating discontinuous nodes appearing on the energy diffusion path includes: Starting from the geometric center point of the main signal core region, trace the vector field streamlines in both the positive and negative directions; Record the spatial coordinates of each signal data unit through which the vector field streamline passes, forming multiple energy diffusion paths; Calculate the characteristic distance between adjacent signal data units on each of the energy diffusion paths; When multiple intervals with a feature distance greater than a preset jump threshold appear consecutively on a certain energy diffusion path, it is determined that there is a path break in this segment of the energy diffusion path; Mark the signal data units on both sides of the path break as path endpoints; The path endpoints that are not directly connected to any other path endpoints on the energy diffusion path are selected and defined as the non-continuous nodes.

6. The method for automatically detecting the quality of ship communication signals according to claim 5, characterized in that, The aggregation of signal data units from the discontinuous nodes and their neighborhoods to form a cluster of suspected interference signals to be determined includes: A spherical neighborhood is defined in the spatial distribution map, centered on each of the non-continuous nodes and bounded by a preset clustering radius. Collect all signal data units that fall within each of the spherical neighborhoods; For each collected signal data unit, calculate its similarity to the central discontinuous node in the feature vector space; Retain signal data units with a similarity greater than a preset similarity threshold, and discard signal data units with a similarity less than or equal to the preset similarity threshold; Each of the non-contiguous nodes is merged with all the signal data units reserved for it, forming an independent candidate set; A union operation is performed on multiple candidate sets whose spatial distance is less than a preset merging threshold. The merged set is the whole set of suspected interference signals to be determined.

7. The method for automatically detecting the quality of ship communication signals according to claim 1, characterized in that, The step of performing a morphological consistency check on the suspected interference signal cluster, and separating the interference component data and the valid signal component data from the spatial distribution map based on the check result, includes: For each of the suspected interference signal clusters, extract the spatial coordinates of all signal data units on its boundary contour; Based on the extracted spatial coordinates, the ratio of the convex hull volume to the surface area of ​​the suspected interference signal cluster is calculated as its morphological compactness parameter. Analyze the directional distribution of the directed edges within the suspected interference signal cluster and calculate its directional entropy value; Set a binary decision threshold for the morphological compactness parameter and the directional entropy value; The morphological compactness parameter and the directional entropy value of the suspected interference signal cluster are input into the binary decision threshold for comparison; If the comparison result meets the interference feature criteria, the suspected interference signal cluster is determined to be an interference component, and all signal data units contained therein are marked as interference component data. If the comparison result does not meet the interference feature criterion, the suspected interference signal cluster is determined to be a distorted part of the valid signal, and all signal data units contained therein are marked as valid signal component data. The signal data units marked as interference components are removed from the spatial distribution map.

8. The method for automatically detecting the quality of ship communication signals according to claim 7, characterized in that, The reassembly of the effective signal component data to reconstruct a continuous ship communication signal waveform includes: In the spatial distribution map after removing the interference component data, only the signal data units marked as valid signal component data are retained; Based on the time-frequency coordinates of the retained effective signal component data, it is restored to the corresponding time-frequency two-dimensional signal segment; All the restored time-frequency two-dimensional signal segments are spliced ​​together according to their time-frequency coordinates to fill the gaps that may be caused by the removal of interference components. The gaps are filled by bilinear interpolation of adjacent segments. An inverse transformation is performed on the spliced ​​and filled complete time-frequency matrix to synthesize the continuous ship communication signal waveform in the time domain.

9. The method for automatically detecting the quality of ship communication signals according to claim 1, characterized in that, The process of performing signal integrity measurement and channel distortion analysis based on the reconstructed ship communication signal waveform includes: Align the reconstructed ship communication signal waveform with the preset ideal reference signal waveform in the time domain; The sum of squared point-by-point errors between the reconstructed ship communication signal waveform after alignment and the ideal reference signal waveform is calculated as the waveform fidelity loss. The reconstructed ship communication signal waveform is subjected to segmented autocorrelation calculation to obtain the distribution of its multipath delay spread; Based on the distribution of the multipath delay spread, calculate the root mean square delay spread parameter and coherence bandwidth parameter of the channel; The waveform fidelity loss, the root mean square delay spread parameter, and the coherence bandwidth parameter together constitute the signal integrity measurement result and the channel distortion analysis result.

10. The method for automatically detecting the quality of ship communication signals according to claim 1, characterized in that, The process of combining the signal integrity metric results and the channel distortion analysis results to generate a quantitative quality report for the original communication signal includes: A first weighting coefficient is set for the waveform fidelity loss, a second weighting coefficient is set for the root mean square delay spread parameter, and a third weighting coefficient is set for the coherence bandwidth parameter. The waveform fidelity loss is weighted using the first weighting coefficient to obtain a fidelity score item; The root mean square delay spread parameter is weighted using the second weighting coefficient, and a delay scoring term is used. The coherent bandwidth parameters are weighted using the third weighting coefficient to obtain the bandwidth score item; The fidelity score, the delay score, and the bandwidth score are added together to obtain the overall signal quality score. The signal quality comprehensive score, the waveform fidelity loss, the root mean square delay spread parameter, the coherence bandwidth parameter, and the identification information of the original communication signal are integrated into a structured data record. The structured data record is output as a quantization quality report for the original communication signal.

Images