Complex electromagnetic signal feature marking and correlation analysis method, device and equipment
By extracting features at multiple levels and constructing an electromagnetic signal knowledge graph, combined with intelligent agent collaborative reasoning, the problems of missing signal features and reliance on manual intervention in electromagnetic signal analysis are solved, enabling refined, intelligent, and efficient management of electromagnetic signals.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NAT UNIV OF DEFENSE TECH
- Filing Date
- 2026-02-27
- Publication Date
- 2026-04-17
AI Technical Summary
Existing technologies for electromagnetic signal analysis suffer from limitations such as fixed-dimensional feature extraction leading to missed signal features and increased noise, making it difficult to adapt to the diversity and dynamic changes of the electromagnetic environment. Furthermore, reliance on human experience results in low analysis efficiency and poor scalability.
By employing multi-level feature extraction and adaptive dimension adjustment, an electromagnetic signal knowledge graph is constructed. Combined with the intelligent agent and model context protocol, collaborative reasoning is performed to generate comprehensive electromagnetic signal situation information.
It enables refined, intelligent, and efficient management of electromagnetic signals, improves the accuracy and scalability of signal analysis, and supports real-time response and rapid adaptation to electromagnetic environments.
Smart Images
Figure CN121880731A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electromagnetic signal analysis technology, and in particular to a method, apparatus, and device for feature marking and correlation analysis of complex electromagnetic signals. Background Technology
[0002] With the rapid development of civilian communication technology, drone applications, the Internet of Things industry, and the broadcasting and television industry, the electromagnetic environment in cities and surrounding areas is becoming increasingly complex, and the number of various electromagnetic signals is growing explosively, gradually highlighting the contradiction between the supply and demand of spectrum resources. At the same time, the adverse effects of unlicensed and abnormal signals on normal communication services and public spectrum order are also increasing. How to achieve the rational allocation of spectrum resources, ensure the quality of normal signal transmission, and investigate abnormal signal sources has become a key requirement for maintaining public communication security and supporting the stable development of industries.
[0003] Currently, in the field of electromagnetic signal analysis, traditional techniques mostly employ fixed-dimensional feature extraction strategies, which pre-set fixed feature parameters and extraction dimensions for specific types of signals. However, fixed-dimensional feature extraction can easily lead to the omission of useful signal features or enhanced noise effects, thereby reducing the accuracy of subsequent signal identification and classification, and making it difficult to adapt to the diverse and dynamically changing characteristics of signals in the electromagnetic environment.
[0004] Some technologies attempt to introduce knowledge graphs to improve signal association management capabilities, but existing graphs mostly focus on building static entity relationships and can only store basic information such as signals and devices, geographical regions, etc., making it difficult to characterize the dynamic evolution of signals. At the same time, existing graphs lack the integration of spatiotemporal context information and cannot deeply associate signal features with specific scenarios, making it difficult to quickly uncover potential connections between signals in tasks such as signal tracing and anomaly signal investigation, resulting in low analysis efficiency.
[0005] Furthermore, current electromagnetic signal correlation analysis and decision-making processes still rely heavily on human experience. For example, in areas such as spectrum resource allocation optimization and the development of abnormal signal handling plans, technical personnel need to combine historical data and professional knowledge to make judgments. Faced with massive amounts of signal data, manual analysis is not only inefficient but also prone to inconsistent results due to subjective judgment differences, making it difficult to achieve real-time responses to changes in the electromagnetic environment. At the same time, the high degree of coupling between modules in existing analysis systems means that when adding signal analysis functions, the core system architecture needs to be reconstructed, resulting in poor scalability and an inability to quickly adapt to the new demands brought about by the rapid development of electromagnetic technology. Summary of the Invention
[0006] Therefore, it is necessary to provide a method, apparatus, and equipment for complex electromagnetic signal feature marking and correlation analysis that can meet the requirements of refined, intelligent, and efficient management of electromagnetic signals, addressing the aforementioned technical problems.
[0007] A method for feature labeling and correlation analysis of complex electromagnetic signals, the method comprising:
[0008] Step 1: Collect multi-source heterogeneous electromagnetic signal data and preprocess it to obtain a signal sequence in a unified format; Step 2: Perform multi-level feature extraction on the signal sequence to obtain an initial feature set; perform adaptive dimensionality adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; perform standardization processing on the optimized feature set to generate feature label data; Step 3: Construct an ontology model of the electromagnetic signal knowledge graph to form a graph pattern layer; based on the entity and relationship definitions of the graph pattern layer, extract and fuse the feature-labeled data to obtain standardized knowledge; store the standardized knowledge in a graph database and embed spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph. Step 4: Based on the electromagnetic signal knowledge graph, obtain the association paths between signal entities, and through the collaborative reasoning of the preset intelligent agent and model context protocol, mine the potential associations and behavioral patterns between signals to obtain the collaborative reasoning results. Step 5: Summarize the results of the collaborative reasoning to generate comprehensive electromagnetic signal situation information.
[0009] On the other hand, a device for complex electromagnetic signal feature marking and correlation analysis is also provided, including: The signal acquisition and preprocessing module is used to acquire multi-source heterogeneous electromagnetic signal data and preprocess it to obtain a signal sequence in a unified format. The feature processing module is used to perform multi-level feature extraction on the signal sequence to obtain an initial feature set; perform adaptive dimensionality adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; and perform standardization processing on the optimized feature set to generate feature label data. The knowledge graph construction module is used to construct an ontology model of the electromagnetic signal knowledge graph, forming a graph pattern layer; based on the entity and relation definitions of the graph pattern layer, knowledge extraction and fusion are performed on the feature-labeled data to obtain standardized knowledge; the standardized knowledge is stored in the graph database and embedded with spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph; The collaborative reasoning module is used to obtain the association paths between signal entities based on the electromagnetic signal knowledge graph, and to mine the potential associations and behavioral patterns between signals through collaborative reasoning of the preset intelligent agent and model context protocol, so as to obtain the collaborative reasoning results. The situation information generation module is used to summarize the collaborative reasoning results and generate comprehensive electromagnetic signal situation information.
[0010] On another front, a computer device is also provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned complex electromagnetic signal feature marking and correlation analysis method.
[0011] Compared with existing technologies, the method, apparatus, and device for complex electromagnetic signal feature marking and correlation analysis provided by this invention have the following advantages: 1. By unifying the signal sequence format, heterogeneous interference from multiple sources is eliminated, laying a reliable foundation for subsequent feature extraction and analysis, and improving the coherence and stability of the overall analysis process.
[0012] 2. By obtaining a comprehensive initial feature set through multi-level feature extraction, and combining multiple evaluation indicators for joint decision-making to achieve adaptive dimension adjustment, feature quality is optimized, redundant features are eliminated, and then standardized processing is performed to generate standardized feature label data, which can improve the usability of features and the accuracy of subsequent knowledge extraction.
[0013] 3. By constructing a complete electromagnetic signal knowledge graph, standardizing knowledge definitions through ontology models and graph pattern layers, embedding spatiotemporal context attributes and signal dynamic evolution characteristics, and realizing the systematic integration of signal knowledge, the intrinsic relationships between signal entities are clearly presented, providing comprehensive and reliable knowledge support for subsequent reasoning.
[0014] 4. Based on knowledge graphs, the system obtains associated paths and uses collaborative reasoning between agents and model context protocols to efficiently mine potential associations and behavioral patterns between signals, thereby improving the intelligence and comprehensiveness of reasoning and solving the problem of accurately capturing complex associations between signals.
[0015] 5. By summarizing collaborative reasoning results to generate comprehensive situational information, it realizes closed-loop management of electromagnetic signals from acquisition, processing, analysis to situational presentation, meeting the needs of refined, intelligent and efficient analysis and management of electromagnetic signals. It is highly practical and has a wide range of applications. Attached Figure Description
[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention, and those skilled in the art can obtain other related drawings based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating the method for feature marking and correlation analysis of complex electromagnetic signals in Example 1; Figure 2 This is a structural block diagram of the complex electromagnetic signal feature marking and correlation analysis device in Example 2; Figure 3This is a diagram of the internal structure of the computer device in Example 3.
[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0020] It should be noted that in this invention, the use of terms such as "first," "second," etc., is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0021] It is understood that the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.
[0022] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0023] Example 1 like Figure 1 As shown, this embodiment provides a method for feature marking and correlation analysis of complex electromagnetic signals, including the following steps: Step 1: Collect multi-source heterogeneous electromagnetic signal data and perform preprocessing to obtain a signal sequence in a unified format.
[0024] Step 2: Perform multi-level feature extraction on the signal sequence to obtain an initial feature set; perform adaptive dimensional adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; perform standardization processing on the optimized feature set to generate feature label data.
[0025] Step 3: Construct an ontology model of the electromagnetic signal knowledge graph to form a graph pattern layer; based on the entity and relationship definitions of the graph pattern layer, extract and fuse knowledge from the feature-labeled data to obtain standardized knowledge; store the standardized knowledge in the graph database and embed spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph.
[0026] Step 4: Based on the electromagnetic signal knowledge graph, obtain the association paths between signal entities, and through the pre-set intelligent agent and model context protocol, mine the potential associations and behavioral patterns between signals to obtain the collaborative reasoning results.
[0027] Step 5: Summarize the collaborative reasoning results and generate comprehensive electromagnetic signal situation information.
[0028] In step 1, the multi-source heterogeneous electromagnetic signal data includes electromagnetic signal-related data from different acquisition carriers with varying data formats. Acquisition carriers include spectrum monitoring equipment, satellite reconnaissance equipment, ground stations, and airborne equipment. Data formats encompass structured monitoring data, semi-structured report data, and unstructured electromagnetic domain text data. Preprocessing refers to a series of data processing operations performed on the raw, multi-source heterogeneous electromagnetic signal data to eliminate data noise, standardize data formats, and improve data usability. A standardized signal sequence is an electromagnetic signal data sequence that, after preprocessing, maintains consistent data structure, dimensions, and numerical range, and can be directly used for subsequent feature extraction.
[0029] In the specific implementation of step 1, firstly, multi-source heterogeneous electromagnetic signal data from multiple channels, such as spectrum monitoring equipment, satellite reconnaissance data, ground station records, and airborne equipment acquisition results, are integrated to form the original electromagnetic signal dataset.
[0030] Subsequently, the original dataset was subjected to data cleaning, denoising, and normalization processes in sequence. Data cleaning removed invalid, missing, and abnormal signal data. Denoising used conventional electromagnetic signal denoising algorithms to eliminate the interference of environmental noise and equipment acquisition noise on the effective signal. Normalization mapped signal data with different dimensions and numerical ranges to the same interval.
[0031] Finally, the processed signal data is reorganized according to a preset data structure to obtain a signal sequence in a unified format. Preferably, the preset signal sequence data structure is a one-dimensional time series with the numerical range normalized to the [0,1] interval.
[0032] This step effectively eliminates heterogeneous interference from signal data of different sources and formats by uniformly acquiring and standardizing the preprocessing of multi-source heterogeneous electromagnetic signal data. It solves the problem of difficult data processing in traditional electromagnetic signal analysis, and lays a standardized and reliable data foundation for subsequent multi-level feature extraction, feature dimension adjustment and other operations. At the same time, it improves the coherence of the overall electromagnetic signal analysis process and the stability of data processing.
[0033] In step 2, multi-level feature extraction refers to the method of extracting features from different dimensions and levels of electromagnetic signals, including three levels of feature extraction: low-level physical features, mid-level behavioral features, and high-level semantic features. The initial feature set is a set of features extracted from the signal sequence through multi-level feature extraction, without dimensionality adjustment optimization. Multi-evaluation index joint decision refers to a method of determining the feature dimension adjustment strategy by combining the calculation results of multiple evaluation indices such as short-time energy, spectral entropy, and real-time signal-to-noise ratio, while integrating feature contribution analysis results and computational resource usage. Adaptive dimensionality adjustment refers to the operation of dynamically triggering dimensionality reduction or expansion adjustment modes based on the results of multi-evaluation index joint decision, thereby optimizing the dimensionality of the initial feature set. The optimized feature set is a feature set that, after adaptive dimensionality adjustment, removes redundant features, supplements effective features, and improves feature quality and expression accuracy. The feature-labeled data is labeled data with an "entity-attribute" standardized format generated after standardizing the optimized feature set according to the ontology model specification of the electromagnetic signal knowledge graph.
[0034] In the specific implementation of step 2, a comprehensive extraction of low-level physical features, mid-level behavioral features, and high-level semantic features is first performed on the signal sequence of the unified format. Low-level physical features include carrier frequency, bandwidth, modulation scheme, signal strength, arrival time, and pulse width; mid-level behavioral features are obtained through short-time statistical analysis, including the temporal regularity of signal occurrence, frequency hopping mode, transmission duration, and duty cycle; high-level semantic features are obtained through preliminary semantic annotation based on rule-based models or lightweight learning models, including signal type and potential source category. All extracted features are then integrated to obtain an initial feature set.
[0035] Then, based on the joint decision of multiple evaluation metrics, adaptive dimensionality adjustment is performed on the initial feature set to obtain an optimized feature set, including: Step 201: Calculate the short-time energy, spectral entropy, and real-time signal-to-noise ratio based on the initial feature set, and integrate these three evaluation metrics to form a multi-dimensional evaluation vector. The expression for calculating the short-time energy is as follows: ; The expression for calculating spectral entropy is: ; In the formula, Indicates short-term energy; Indicates the first The signal amplitude at each sampling point; Indicates the length of the short-time analysis window; Represents spectral entropy; Indicates the first Normalized power spectral density at each frequency point; This indicates the total number of frequency points analyzed in the spectrum; Represents a logarithmic function.
[0036] Step 202: Using a dual-threshold hysteresis comparison strategy, the multidimensional evaluation vector is determined by combining the feature contribution analysis results with the computational resource usage.
[0037] Specifically, the feature contribution analysis includes: Step 211: Use the random forest algorithm to score the importance of each feature in the initial feature set and obtain the contribution score of each feature; sort the contribution scores in descending order to obtain the feature contribution ranking result.
[0038] Step 212: Set the core feature proportion threshold and the redundant feature proportion threshold. Based on the feature contribution ranking results, accumulate the proportion of the first contribution of the preceding features. When the proportion of the first contribution reaches the core feature proportion threshold, the corresponding preceding feature is classified as a core feature. Accumulate the proportion of the second contribution of the subsequent features. When the proportion of the second contribution is lower than the redundant feature proportion threshold, the corresponding subsequent feature is classified as a redundant feature.
[0039] Step 213: Calculate the proportion of core features in the initial feature set to obtain core feature proportion data, and calculate the proportion of redundant features in the initial feature set to obtain redundant feature proportion data, thus forming a complete feature contribution analysis result.
[0040] More specifically, a dual-threshold hysteresis comparison strategy is adopted, combining the feature contribution analysis results with computational resource consumption, to determine the multidimensional evaluation vector, including: Step 221: Set upper and lower threshold values for each evaluation index in the multidimensional evaluation vector.
[0041] Step 222: Retrieve the core feature percentage data and redundant feature percentage data from the feature contribution analysis results, and simultaneously obtain the current computing resource occupancy rate and remaining computing power.
[0042] Step 223: If any evaluation index is lower than the lower limit threshold multiple times in a row, and the proportion of redundant features is higher than the proportion of redundant features threshold, or the utilization rate of computing resources is higher than the preset utilization threshold, it is determined that the environment is harsh or the signal is complex.
[0043] Step 224: If any evaluation indicator is higher than the upper limit threshold multiple times in a row, and the proportion of core features is higher than the core feature proportion threshold, and the remaining computing power meets the dimensionality expansion requirements, the environment is judged to be excellent or the features are clear.
[0044] Step 225: If the evaluation index is between the upper and lower thresholds, or if the percentage of redundant features, the percentage of core features, the utilization rate of computing resources, or the remaining computing power does not meet the above judgment conditions, the current feature dimension is maintained and the adjustment mode is not triggered.
[0045] Step 203: If the determination result is a harsh environment or complex signal, trigger the dimension reduction adjustment mode; if the determination result is a favorable environment or clear characteristics, trigger the dimension expansion adjustment mode.
[0046] Step 204: When the dimensionality reduction adjustment mode is triggered, incremental principal component analysis is used to compress the initial feature set online and map it to a low-dimensional subspace to obtain the dimensionality-reduced feature set. Preferably, the cumulative variance contribution rate threshold of the incremental principal component analysis is set to 85%.
[0047] Step 205: When the dimension expansion adjustment mode is triggered, wavelet packet decomposition is used to perform fine time-frequency analysis on the initial feature set, extract sub-band energy features, and merge them with the original feature set to obtain the dimension-expanded feature set. Preferably, the number of decomposition layers of the wavelet packet decomposition is set to 3.
[0048] Step 206: Construct an adjusted feature set based on the dimensionality-reduced and dimensionality-expanded feature sets. Then, calculate the evaluation metrics for the adjusted feature set. The evaluation metrics include reconstruction error and task performance metrics. The task performance metrics include classification accuracy, cluster purity, and silhouette coefficient. The reconstruction error is calculated using the mean squared error formula. If any evaluation metric exceeds a preset threshold, roll back to the previous feature dimension or switch the adjustment strategy to finally obtain the optimized feature set.
[0049] Finally, based on the entity attribute specification defined by the ontology model of the electromagnetic signal knowledge graph, the physical features, behavioral features, and semantic features in the optimized feature set are classified and mapped. The mapped features are then converted into a standardized "entity-attribute-value" format. The standardized features are then deduplicated and disambiguated to finally generate feature-labeled data.
[0050] This step achieves multi-dimensional and comprehensive feature capture of electromagnetic signals through multi-level feature extraction, solving the problem of missing useful features caused by the single feature extraction in traditional methods. The resulting initial feature set can comprehensively represent the characteristic attributes of electromagnetic signals. Based on an adaptive dimension adjustment mechanism with joint decision-making of multiple evaluation indicators, it abandons the traditional fixed-dimensional feature extraction approach. It can dynamically optimize the feature space dimension according to the real-time quality of electromagnetic signals, feature contribution, and computing resource status, effectively eliminating redundant features and supplementing effective features. This not only suppresses noise interference but also improves the accuracy of feature expression while taking into account the efficiency of computing resource utilization. The feature-labeled data generated through standardized processing conforms to the ontology model specifications of knowledge graphs, providing standardized input for subsequent knowledge extraction and fusion, and significantly improving the accuracy and efficiency of subsequent knowledge extraction.
[0051] In step 3, the ontology model refers to the model used to define the core entity types, inter-entity relationship types, and entity attribute specifications in the electromagnetic signal knowledge graph; it is the core skeleton of the knowledge graph. The graph schema layer refers to the hierarchical structure constructed from the ontology model, used to standardize the knowledge organization in the knowledge graph, providing a unified standard for subsequent knowledge extraction, fusion, and storage. Knowledge extraction and fusion refers to the operation of extracting entities and relationships from feature-labeled data and electromagnetic domain text data, aligning and disambiguating multi-source entities, and fusing them into knowledge in a unified format. Standardized knowledge is the unified format knowledge that conforms to the specifications of the graph schema layer after knowledge extraction and fusion. Spatiotemporal context attributes are the time, space, and environmental attributes related to electromagnetic signals, including precise timestamps, environmental noise parameters, location historical activity baselines, acquisition equipment parameters, and monitoring scene types. Signal dynamic evolution characteristics refer to the regular characteristics of the electromagnetic signal's state changing over time and space, with the Markov state transition matrix as the core representation. The electromagnetic signal knowledge graph is a graph that integrates standardized knowledge, spatiotemporal context attributes, and dynamic evolution characteristics of signals. It can systematically present the relationships between electromagnetic signal entities and supports incremental updates and dynamic completion.
[0052] Step 3 is divided into four core stages: ontology model construction, knowledge extraction and fusion, graph construction and attribute embedding, and dynamic graph completion.
[0053] First, a dedicated ontology model for the electromagnetic signal knowledge graph is designed, defining the core entity types and the relationship types between entities. The core entity types include signals, emission sources, geographical regions, time slices, and acquisition devices, while the relationship types between entities include originating from, appearing in, belonging to, cooperating with, and associating with. Based on this ontology model, a standardized graph pattern layer is formed.
[0054] Subsequently, based on the entity and relation definitions at the graph pattern layer, knowledge extraction and fusion are performed on the feature-labeled data to obtain standardized knowledge, including: Step 301: Using a bidirectional long short-term memory network based on an attention mechanism combined with a remote supervision method, entities and relations are extracted from feature-labeled data and electromagnetic domain text data to obtain an initial set of entities and relations.
[0055] Step 302: An unsupervised multi-view hierarchical clustering algorithm is used to align multi-source entities in the initial entity and relation set, eliminating the problems of homonyms and heteronyms, and obtaining the aligned entity and relation set.
[0056] Step 303: Merge the entities and relations in the aligned entity and relation set to form standardized knowledge in a unified format.
[0057] Then, standardized knowledge is stored in a graph database and embedded with spatiotemporal context attributes and associated signal dynamic evolution features to construct an electromagnetic signal knowledge graph, including: Step 311: The standardized knowledge is transformed into triplet data in the format of "entity-relationship-entity". According to the entity and relation definitions of the graph schema layer, the triplet data is imported into the graph database to form a basic electromagnetic signal knowledge graph. Preferably, Neo4j is used as the graph database.
[0058] Step 312 involves adding spatiotemporal environment association parameters to the signal association relationships in the basic electromagnetic signal knowledge graph, such as {signal, associated with, spatial location} and {signal, originating from, signal source}. This adds scene attributes to the signal entities, completing the embedding of spatiotemporal context attributes and resulting in a knowledge graph with spatiotemporal context attributes. The spatiotemporal environment association parameters include precise timestamps, environmental noise parameters, and historical activity baselines; scene attributes include data acquisition device parameters and monitoring scene types.
[0059] Step 313: Divide continuous parameters such as signal power, frequency band, and activity into discrete states. Learn and construct the Markov state transition matrix corresponding to the dynamic evolution characteristics of the signal based on historical signal data. Use the Markov state transition matrix as the dynamic attribute of the signal entity and associate it with the corresponding signal entity node in the knowledge graph with spatiotemporal context attributes to obtain the knowledge graph with fused dynamic features.
[0060] Step 314: Configure an incremental update triggering mechanism. When new electromagnetic signal data is detected and preprocessing, feature extraction and standardization are completed, the knowledge extraction and fusion process is automatically triggered to generate new triplet data. According to the definition of the graph mode layer, the data is updated to the knowledge graph of fused dynamic features to obtain an electromagnetic signal knowledge graph that supports incremental updates.
[0061] Furthermore, after constructing an electromagnetic signal knowledge graph that supports incremental updates, the process also includes: dynamically completing the electromagnetic signal knowledge graph, the specific process of which is as follows: Step 321: Construct a Markov Logic Network (MLN) framework. Transform the entities, relations, attributes in the electromagnetic signal knowledge graph, as well as the state transition probabilities corresponding to the Markov state transition matrix, into predicates and formulas of the Markov Logic Network framework to obtain MLN inference input data.
[0062] Step 322: Based on the Markov logic network framework, perform probabilistic reasoning on the MLN inference input data to obtain the latent relationships that are not explicitly present in the graph and the confidence level corresponding to each latent relationship. The confidence level is used to characterize the probability that the latent relationship actually exists.
[0063] Step 323: Based on the potential relationships, complete the missing relationships in the electromagnetic signal knowledge graph to obtain a set of completed relationships to be verified. Specifically, compare all potential relationships with the existing entity relationships in the electromagnetic signal knowledge graph to locate the missing relationships in the graph. Then, perform preliminary completion of the missing relationships by matching the corresponding potential relationships to obtain a set of completed relationships to be verified.
[0064] Step 324: Set a confidence threshold, filter the set of relationships to be verified and complete, and add potential relationships with confidence levels higher than the confidence threshold to the electromagnetic signal knowledge graph to complete the dynamic completion and improvement of the graph.
[0065] The electromagnetic signal knowledge graph ontology model and graph pattern layer constructed in this step provide a unified standard for the organization and management of electromagnetic signal knowledge, solving the problems of fragmentation and unclear relationships in traditional electromagnetic signal knowledge. By combining a bidirectional long short-term memory network based on an attention mechanism with a remotely supervised knowledge extraction method and an entity alignment method based on unsupervised multi-view hierarchical clustering, efficient knowledge extraction and fusion from multi-source data are achieved, improving the completeness and consistency of knowledge. Embedding spatiotemporal context attributes and associating them with the dynamic evolution characteristics of signals during graph construction breaks through the limitation of traditional knowledge graphs that only store static entity relationships, enabling accurate characterization of the spatiotemporal characteristics and dynamic evolution patterns of electromagnetic signals. Incremental update mechanisms and probabilistic reasoning completion using Markov logic networks realize the dynamic evolution and improvement of the electromagnetic signal knowledge graph, allowing the graph to adapt to the dynamic changes in the electromagnetic environment in real time, providing comprehensive, reliable, and dynamic knowledge support for subsequent associative reasoning.
[0066] In step 4, the association path between signal entities refers to the entity-relationship link connecting different signal entities, obtained through multi-dimensional association queries based on the topological structure of the electromagnetic signal knowledge graph. It is an intuitive representation of the association relationship between signal entities. An agent refers to a pre-defined functional module with specialized electromagnetic signal analysis and reasoning capabilities. Each agent collaborates and is responsible for different dimensions of signal analysis. The Model Context Protocol (MCP) is a standardized protocol used to regulate the communication rules and service call methods between agents and the core server, enabling collaborative work among agents. Collaborative reasoning is a reasoning method where the MCP server schedules various specialized agents to form a temporary collaborative reasoning alliance. Each agent conducts specialized reasoning based on the association path between signal entities, and the reasoning results are fused and verified. The collaborative reasoning result is a complete analysis result containing potential associations between signals, collaborative behavior patterns, and behavioral development laws, formed after specialized reasoning by each agent, fusion and deduplication by the MCP server, and validity verification.
[0067] In the specific implementation of step 4, firstly, based on the completed and dynamically supplemented electromagnetic signal knowledge graph, a flexible multi-hop association query is performed. According to the actual needs of electromagnetic signal analysis, the corresponding signal entity nodes are extracted from the graph, the entity-relationship links between each signal entity are mined, and the association paths between signal entities with attribute annotations are obtained.
[0068] Then, through collaborative reasoning between the pre-defined agent and the model context protocol, the potential correlations and behavioral patterns between signals are mined to obtain the collaborative reasoning results, including: Step 401: Pre-define multiple types of intelligent agents, including a signal tracing agent, a behavior pattern recognition agent, an intent prediction agent, and a knowledge management agent. Configure each agent with an inference algorithm adapted to electromagnetic signal analysis to mine the correlations and patterns in different dimensions of the signal. Specifically, configure an electromagnetic propagation model for the signal tracing agent, a graph community discovery algorithm for the behavior pattern recognition agent, a sequence prediction model combined with Markov state transition probabilities for the intent prediction agent, and a domain knowledge verification algorithm for the knowledge management agent.
[0069] Step 402: Construct a Model Context Protocol (MCP) server to encapsulate the system's core reasoning capabilities into standardized services, establish communication links between each agent and the electromagnetic signal knowledge graph, and enable each agent to synchronously obtain the association paths between signal entities and the information parameters in the electromagnetic signal knowledge graph through the MCP server.
[0070] Step 403: Based on the analysis requirements of the association path, the model context protocol server schedules corresponding agents to form a temporary collaborative reasoning alliance and initiates collaborative reasoning. Specifically, the signal tracing agent, using an electromagnetic propagation model, retrieves information parameters such as acquisition device parameters and electromagnetic propagation characteristics of geographical areas from the electromagnetic signal knowledge graph, and mines source associations between signal entities based on the association path to obtain signal source association results. The behavior pattern recognition agent, using a graph community discovery algorithm, retrieves information parameters such as signal spatiotemporal context and baseline collaborative behavior of similar signals from the electromagnetic signal knowledge graph, and mines collaborative behavior patterns between signal entities based on the association path to obtain collaborative behavior pattern results. The intent prediction agent, using Markov state transition probabilities, retrieves information parameters such as signal dynamic evolution characteristics and historical behavior patterns of signal sources from the electromagnetic signal knowledge graph, and infers the behavioral development patterns of signal entities based on the association path to obtain behavioral evolution pattern results. The knowledge management agent retrieves information parameters such as electromagnetic domain ontology specifications and verified signal association rules from the electromagnetic signal knowledge graph to verify the reasoning basis of each agent, ensuring that the reasoning process conforms to the characteristics of the electromagnetic signal domain, and obtains trend inference results.
[0071] In step 404, each agent feeds back the signal source association results, collaborative behavior pattern results, behavior evolution law results, and trend inference results to the model context protocol server. The model context protocol server merges, deduplicates, and verifies the validity of all inference results, removes invalid inference information, and integrates them to form collaborative inference results.
[0072] This step utilizes multi-hop association queries based on the electromagnetic signal knowledge graph to obtain the association paths between signal entities, enabling rapid discovery of surface-level relationships between signal entities and providing clear analytical clues for subsequent deep reasoning. Through a pre-defined collaborative reasoning mechanism between specialized agents and model context protocols, it overcomes the limitations of traditional single-reasoning methods, such as low efficiency and incomplete discovery of potential associations. It also solves the problems of traditional electromagnetic signal analysis relying on human experience and having low levels of intelligence. The division of labor and mutual verification among agents not only improves the comprehensiveness and accuracy of the reasoning results but also enables in-depth mining of potential associations, collaborative behavior patterns, and behavioral development laws between signals. The standardized service encapsulation of the model context protocol decouples the agents from the core system, improving the system's modularity and scalability.
[0073] In step 5, the comprehensive electromagnetic signal situation information refers to the information formed by integrating, analyzing, and visualizing the results of collaborative reasoning. It can comprehensively and intuitively reflect the overall state of the electromagnetic environment in a specific area and time period, including multi-dimensional content such as signal source distribution, inter-signal correlation, signal collaborative behavior patterns, and signal behavior intent analysis.
[0074] In the specific implementation of step 5, the collaborative reasoning results are first integrated in multiple dimensions. The signal source association results, collaborative behavior pattern results, behavior evolution law results, and trend inference results are classified and summarized according to the monitoring area, monitoring time, signal type, and other dimensions of the electromagnetic signal. Duplicate and contradictory reasoning information is eliminated, and valid and high-confidence reasoning results are retained.
[0075] The integrated reasoning results were then structured and analyzed, and a systematic analytical conclusion was formed by combining the spatiotemporal context attributes and dynamic evolution characteristics of the electromagnetic signal knowledge graph.
[0076] Finally, the systematic analysis conclusions are visualized. Preferably, they are presented in the form of a situation map, which includes information such as the geographical distribution of signal sources, the trajectory of signal cooperative behavior, and the confidence level of signal intent judgment. Ultimately, comprehensive electromagnetic signal situation information that can fully characterize the state of the regional electromagnetic environment is generated.
[0077] This step integrates scattered inference results into a systematic electromagnetic environment analysis conclusion by summarizing and structuring the collaborative inference results from multiple dimensions. This solves the problem of scattered electromagnetic signal analysis results and the inability to form a comprehensive control basis in traditional methods. The generated comprehensive electromagnetic signal situation information can comprehensively and intuitively reflect the overall state of the complex electromagnetic environment, realizing closed-loop management of electromagnetic signals from acquisition, preprocessing, feature extraction, knowledge construction to intelligent inference. This situation information provides direct and effective data support for spectrum resource allocation, abnormal signal investigation, electromagnetic situation awareness, and comprehensive decision-making, meeting the core requirements for refined, intelligent, and efficient management of electromagnetic signals.
[0078] It should be understood that, although this embodiment Figure 1 The steps are shown sequentially as indicated by the arrows, but they are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order in which these steps are performed; they can be executed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
[0079] Example 2 Based on the complex electromagnetic signal feature labeling and correlation analysis method in Example 1, this example discloses a complex electromagnetic signal feature labeling and correlation analysis device, such as... Figure 2As shown, the complex electromagnetic signal feature labeling and correlation analysis device includes: a signal acquisition and preprocessing module 601, a feature processing module 602, a knowledge graph construction module 603, a collaborative reasoning module 604, and a situation information generation module 605, wherein: The signal acquisition and preprocessing module 601 is used to acquire multi-source heterogeneous electromagnetic signal data and perform preprocessing to obtain a signal sequence in a unified format.
[0080] The feature processing module 602 is used to perform multi-level feature extraction on the signal sequence to obtain an initial feature set; to perform adaptive dimensional adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; and to perform standardization processing on the optimized feature set to generate feature label data.
[0081] The knowledge graph construction module 603 is used to construct the ontology model of the electromagnetic signal knowledge graph, forming a graph pattern layer; based on the entity and relationship definitions of the graph pattern layer, knowledge is extracted and fused from the feature-labeled data to obtain standardized knowledge; the standardized knowledge is stored in the graph database and embedded with spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph.
[0082] The collaborative reasoning module 604 is used to obtain the association paths between signal entities based on the electromagnetic signal knowledge graph. Through collaborative reasoning of the preset intelligent agent and model context protocol, it can explore the potential associations and behavioral patterns between signals and obtain collaborative reasoning results.
[0083] The situation information generation module 605 is used to summarize the collaborative reasoning results and generate comprehensive electromagnetic signal situation information.
[0084] In this embodiment, the specific working processes and principles of the signal acquisition and preprocessing module 601, feature processing module 602, knowledge graph construction module 603, collaborative reasoning module 604, and situational information generation module 605 are the same as those in Embodiment 1, and therefore will not be described again in this embodiment. Each unit module can be implemented entirely or partially through software, hardware, or a combination thereof. Each unit module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each of the above unit modules.
[0085] Example 3 like Figure 3 The diagram illustrates a terminal device disclosed in this embodiment, comprising a transmitter, a receiver, a memory, and a processor. The transmitter transmits instructions and data, the receiver receives instructions and data, the memory stores computer-executed instructions, and the processor executes the computer-executed instructions stored in the memory to implement the method described in Embodiment 1 above.
[0086] It is important to note that the aforementioned memory can be either standalone or integrated with the processor. When the memory is set up independently, the terminal device also includes a bus for connecting the memory and the processor.
[0087] Example 4 This embodiment discloses a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the method in Embodiment 1 above.
[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0089] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0090] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.
Claims
1. A method for feature marking and correlation analysis of complex electromagnetic signals, characterized in that, The method includes: Step 1: Collect multi-source heterogeneous electromagnetic signal data and preprocess it to obtain a signal sequence in a unified format; Step 2: Perform multi-level feature extraction on the signal sequence to obtain an initial feature set; perform adaptive dimensionality adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; perform standardization processing on the optimized feature set to generate feature label data; Step 3: Construct an ontology model of the electromagnetic signal knowledge graph to form a graph pattern layer; based on the entity and relationship definitions of the graph pattern layer, extract and fuse the feature-labeled data to obtain standardized knowledge; store the standardized knowledge in a graph database and embed spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph. Step 4: Based on the electromagnetic signal knowledge graph, obtain the association paths between signal entities, and through the collaborative reasoning of the preset intelligent agent and model context protocol, mine the potential associations and behavioral patterns between signals to obtain the collaborative reasoning results. Step 5: Summarize the results of the collaborative reasoning to generate comprehensive electromagnetic signal situation information.
2. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 1, characterized in that, In step 2, an adaptive dimensionality adjustment is performed on the initial feature set based on a joint decision using multiple evaluation metrics to obtain an optimized feature set, including: Step 201: Calculate the short-time energy, spectral entropy, and real-time signal-to-noise ratio based on the initial feature set to form a multi-dimensional evaluation vector; Step 202: Using a dual-threshold hysteresis comparison strategy, the multidimensional evaluation vector is determined by combining the feature contribution analysis results with the computational resource usage. Step 203: If the determination result is a harsh environment or complex signal, trigger the dimension reduction adjustment mode; if the determination result is a favorable environment or clear characteristics, trigger the dimension expansion adjustment mode. Step 204: When the dimensionality reduction adjustment mode is triggered, incremental principal component analysis is used to compress the initial feature set online and map it to a low-dimensional subspace to obtain the dimensionality-reduced feature set. Step 205: When the dimension expansion adjustment mode is triggered, wavelet packet decomposition is used to perform fine time-frequency analysis on the initial feature set, extract sub-band energy features and merge them with the original feature set to obtain the dimension-expanded feature set; Step 206: Construct an adjusted feature set based on the reduced-dimensional feature set and the expanded-dimensional feature set, and then calculate the evaluation index of the adjusted feature set; if any evaluation index exceeds a preset threshold, roll back to the previous feature dimension or switch the adjustment strategy to finally obtain the optimized feature set.
3. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 2, characterized in that, In step 202, the feature contribution analysis includes: Step 211: Use the random forest algorithm to score the importance of each feature in the initial feature set to obtain the contribution score of each feature; sort the contribution scores to obtain the feature contribution ranking result. Step 212: Set the core feature proportion threshold and the redundant feature proportion threshold. Based on the feature contribution ranking result, accumulate the proportion of the first contribution of the preceding features. When the proportion of the first contribution reaches the core feature proportion threshold, the corresponding preceding feature is classified as a core feature. Accumulate the proportion of the second contribution of the subsequent features. When the proportion of the second contribution is lower than the redundant feature proportion threshold, the corresponding subsequent feature is classified as a redundant feature. Step 213: Calculate the proportion of core features in the initial feature set to obtain core feature proportion data, and calculate the proportion of redundant features in the initial feature set to obtain redundant feature proportion data, thus forming a complete feature contribution analysis result.
4. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 3, characterized in that, In step 202, a dual-threshold hysteresis comparison strategy is adopted, combining the feature contribution analysis results with the computational resource usage, to determine the multidimensional evaluation vector, including: Step 221: Set an upper limit threshold and a lower limit threshold for each evaluation index in the multidimensional evaluation vector; Step 222: Retrieve the core feature percentage data and redundant feature percentage data from the feature contribution analysis results, and simultaneously obtain the current computing resource occupancy rate and remaining computing power; Step 223: If any evaluation index is lower than the lower limit threshold multiple times in a row, and the proportion of redundant features is higher than the proportion of redundant features threshold, or the resource utilization rate is higher than the preset utilization threshold, it is determined that the environment is harsh or the signal is complex. Step 224: If any evaluation index is higher than the upper limit threshold multiple times in a row, and the proportion of core features is higher than the proportion of core features threshold, and the remaining computing power of computing resources meets the dimensionality expansion requirements, it is determined that the environment is excellent or the features are clear. Step 225: If the evaluation index is between the upper limit threshold and the lower limit threshold, or if the redundant feature ratio data, core feature ratio data, computing resource occupancy rate or remaining computing power do not meet the above judgment conditions, maintain the current feature dimension and do not trigger the adjustment mode.
5. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 1, characterized in that, In step 3, based on the entity and relation definitions of the graph pattern layer, knowledge extraction and fusion are performed on the feature-labeled data to obtain standardized knowledge, including: Step 301: Using a bidirectional long short-term memory network based on an attention mechanism combined with a remote supervision method, entities and relations are extracted from the feature-labeled data and electromagnetic domain text data to obtain an initial set of entities and relations. Step 302: An unsupervised multi-view hierarchical clustering algorithm is used to align the multi-source entities in the initial entity and relation set, eliminating the problems of homonyms and heteronyms, and obtaining the aligned entity and relation set. Step 303: Merge the entities and relations in the aligned entity and relation set to form standardized knowledge in a unified format.
6. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 5, characterized in that, In step 3, the standardized knowledge is stored in a graph database and embedded with spatiotemporal context attributes and associated signal dynamic evolution features to construct an electromagnetic signal knowledge graph, including: Step 311: The standardized knowledge is converted into triplet data. According to the entity and relation definitions of the graph pattern layer, the triplet data is imported into the graph database to form a basic electromagnetic signal knowledge graph. Step 312: Add spatiotemporal environment association parameters to the signal association class relationships in the basic electromagnetic signal knowledge graph, and associate scene attributes with signal entities to complete the embedding of spatiotemporal context attributes, thereby obtaining a knowledge graph with spatiotemporal context attributes; Step 313: Retrieve the Markov state transition matrix corresponding to the dynamic evolution features of the signal, use the Markov state transition matrix as the dynamic attribute of the signal entity, and associate it with the corresponding signal entity node in the knowledge graph with spatiotemporal context attributes to obtain the knowledge graph with fused dynamic features. Step 314: Configure an incremental update triggering mechanism, generate new triplet data based on the detected new electromagnetic signals, update the knowledge graph of the fused dynamic features according to the definition of the graph mode layer, and finally construct an electromagnetic signal knowledge graph that supports incremental updates.
7. The method for feature marking and correlation analysis of complex electromagnetic signals according to claim 6, characterized in that, Step 3, after constructing the electromagnetic signal knowledge graph that supports incremental updates, also includes: dynamically completing the electromagnetic signal knowledge graph; Dynamic completion of the electromagnetic signal knowledge graph includes: Step 321: Construct a Markov logic network framework, and transform the entities, relations, attributes in the electromagnetic signal knowledge graph, as well as the state transition probabilities corresponding to the Markov state transition matrix, into predicates and formulas of the Markov logic network framework to obtain MLN inference input data. Step 322: Based on the Markov logic network framework, perform probabilistic reasoning on the MLN inference input data to obtain potential relationships and corresponding confidence levels; Step 323: Based on the potential relationships, complete the missing relationships in the electromagnetic signal knowledge graph to obtain a set of complete relationships to be verified; Step 324: Set a confidence threshold, filter the set of relationships to be verified and complete, and add potential relationships with a confidence level higher than the confidence threshold to the electromagnetic signal knowledge graph.
8. The method for feature marking and correlation analysis of complex electromagnetic signals according to any one of claims 1 to 6, characterized in that, In step 4, through collaborative reasoning using a pre-defined intelligent agent and model context protocol, potential correlations and behavioral patterns between signals are mined to obtain collaborative reasoning results, including: Step 401: Preset multiple types of intelligent agents, including signal tracing intelligent agents, behavior pattern recognition intelligent agents, intent prediction intelligent agents and knowledge management intelligent agents, and configure inference algorithms adapted to electromagnetic signal analysis for each intelligent agent to mine the correlation and patterns of signals in different dimensions. Step 402: Construct a model context protocol server, encapsulate the core reasoning capabilities of the system into a standardized service, establish a communication link between each intelligent agent and the electromagnetic signal knowledge graph, and synchronously obtain the association paths between signal entities and the information parameters in the electromagnetic signal knowledge graph. Step 403: Based on the analysis requirements of the associated paths, the model context protocol server schedules corresponding agents to form a temporary collaborative reasoning alliance and initiates collaborative reasoning. Specifically, the signal tracing agent, using an electromagnetic propagation model, mines source associations between signal entities based on the information parameters and the associated paths to obtain signal source association results; the behavior pattern recognition agent, using a graph community discovery algorithm, mines collaborative behavior patterns between signal entities based on the information parameters and the associated paths to obtain collaborative behavior pattern results; the intent prediction agent, using Markov state transition probabilities, infers the behavioral development patterns of signal entities based on the information parameters and the associated paths to obtain behavioral evolution pattern results; and the knowledge management agent verifies the reasoning basis of each agent to obtain trend inference results. Step 404: Each agent feeds back the signal source association results, collaborative behavior pattern results, behavior evolution law results, and trend inference results to the model context protocol server. The model context protocol server fuses, deduplicates, and verifies the validity of all inference results, removes invalid inference information, and integrates them to form collaborative inference results.
9. A device for feature marking and correlation analysis of complex electromagnetic signals, characterized in that, The device includes: The signal acquisition and preprocessing module is used to acquire multi-source heterogeneous electromagnetic signal data and preprocess it to obtain a signal sequence in a unified format. The feature processing module is used to perform multi-level feature extraction on the signal sequence to obtain an initial feature set; perform adaptive dimensionality adjustment on the initial feature set based on joint decision of multiple evaluation indicators to obtain an optimized feature set; and perform standardization processing on the optimized feature set to generate feature label data. The knowledge graph construction module is used to construct an ontology model of the electromagnetic signal knowledge graph, forming a graph pattern layer; based on the entity and relation definitions of the graph pattern layer, knowledge extraction and fusion are performed on the feature-labeled data to obtain standardized knowledge; the standardized knowledge is stored in the graph database and embedded with spatiotemporal context attributes and associated signal dynamic evolution features to construct the electromagnetic signal knowledge graph; The collaborative reasoning module is used to obtain the association paths between signal entities based on the electromagnetic signal knowledge graph, and to mine the potential associations and behavioral patterns between signals through collaborative reasoning of the preset intelligent agent and model context protocol, so as to obtain the collaborative reasoning results. The situation information generation module is used to summarize the collaborative reasoning results and generate comprehensive electromagnetic signal situation information.
10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the complex electromagnetic signal feature marking and correlation analysis method according to any one of claims 1 to 7.