CN-121980311-A - Cognitive intelligent communication signal identification method and system

CN121980311ACN 121980311 ACN121980311 ACN 121980311ACN-121980311-A

Abstract

The invention provides a cognitive intelligent communication signal identification method and system, which are used for acquiring a signal data set corresponding to a communication signal to be identified and comprising single signal units under different transmission scenes, constructing a signal characteristic dynamic association network based on characteristic association relations of the single signal units, wherein the signal characteristic dynamic association network comprises characteristic units and dynamic association strength, calling a pre-trained cognitive inference model to conduct hierarchical characteristic inference processing on the signal characteristic dynamic association network to generate hierarchical characteristic inference results of each inference hierarchy, generating a signal identification inference intermediate result comprising a preliminary identification type of each single signal unit according to the hierarchical characteristic inference results, and finally integrating the signal identification inference intermediate results to generate a final signal identification result of the communication signal to be identified. The invention can adapt to complex transmission scenes and effectively improve the accuracy and reliability of signal identification.

Inventors

XU SHENGFA
XING JINGYI
CHEN HONGLIANG
FU XIONGJUN
LI TIANTIAN
GUO HUIPING
LI QIANHUI
LI SHUANGYU

Assignees

北京东方计量测试研究所

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. A cognitive intelligent communication signal recognition method, the method comprising: Acquiring a signal data set corresponding to a communication signal to be identified, wherein the signal data set comprises single signal units acquired under different transmission scenes; constructing a signal characteristic dynamic association network based on the characteristic association relation of single signal units contained in the signal data set, wherein the signal characteristic dynamic association network contains characteristic units and dynamic association strength among the characteristic units; Invoking a pre-trained cognitive reasoning model to conduct hierarchical feature reasoning processing on the signal feature dynamic association network, and generating hierarchical feature reasoning results corresponding to each reasoning hierarchy respectively; generating a signal recognition reasoning intermediate result according to the hierarchical characteristic reasoning result, wherein the signal recognition reasoning intermediate result comprises a preliminary recognition type corresponding to each single signal unit in a signal data set; and integrating the signal recognition reasoning intermediate results corresponding to each single signal unit in the signal data set, and generating a final signal recognition result corresponding to the communication signal to be recognized.
2. The cognitive intelligent communication signal recognition method according to claim 1, wherein the constructing a signal feature dynamic association network based on feature association relations of individual signal units contained in a signal data set, the signal feature dynamic association network containing feature units and feature inter-unit dynamic association strengths comprises: Extracting characteristic units capable of representing essential attributes of signals from single signal units contained in a signal data set, wherein the characteristic units are core elements for supporting signal identification; Performing association relation analysis processing on the extracted feature units, and determining real-time association states among different feature units; calculating a dynamic value of the association strength between the feature units according to the determined real-time association state, and executing adjustment processing along with the change of the transmission scene by the dynamic value of the association strength; Constructing an initial association network structure based on the extracted feature units and the calculated association strength dynamic values, wherein the initial association network structure comprises connection relations of the feature units and association strength labels; And according to the characteristic change of single signal units contained in the signal data set under different transmission scenes, updating the association strength dynamic value in the built initial association network structure to form a signal characteristic dynamic association network.
3. The cognitive intelligent communication signal recognition method according to claim 1, wherein the invoking the pre-trained cognitive inference model performs hierarchical feature inference processing on the signal feature dynamic association network to generate hierarchical feature inference results corresponding to each inference hierarchy respectively, and the method comprises: Inputting the signal characteristic dynamic association network into a network analysis module of a cognitive reasoning model, and executing analysis processing on the hierarchical structure of the signal characteristic dynamic association network and the connection relation of the characteristic units to obtain network structure analysis information; Converting feature units in the network structure analysis information into reasoning feature vectors which can be processed by the cognitive reasoning model through a feature conversion module of the cognitive reasoning model to obtain a reasoning feature vector group; The hierarchical reasoning module of the cognitive reasoning model is utilized to execute the layer-by-layer reasoning processing on the obtained reasoning feature vector group, the reasoning feature vector of the highest reasoning hierarchy in the reasoning feature vector group is executed firstly, and then the reasoning feature vector of the next hierarchy in the reasoning feature vector group is executed in sequence; in the process of each layer of reasoning processing, performing associated reasoning processing on the reasoning feature vector of the current level based on the built-in reasoning rule of the cognitive reasoning model to generate a reasoning confidence sequence corresponding to the current reasoning level; and collecting the inference confidence sequences corresponding to the inference levels respectively, and generating a level characteristic inference result containing the inference information of each inference level.
4. The cognitive intelligent communication signal recognition method according to claim 2, wherein the extracting a feature unit capable of characterizing a signal essential attribute from a single signal unit contained in a signal data set, the feature unit being a core element supporting signal recognition, comprises: performing feature dimension disassembly processing on single signal units contained in the signal data set to determine a core feature dimension capable of reflecting essential attributes of signals; under each determined core feature dimension, acquiring a feature information fragment which can embody the attribute of the core feature dimension in a single signal unit contained in a signal data set; Performing redundancy elimination processing on the acquired characteristic information fragments, and removing the characteristic information fragments which are repeated or can not reflect the essential attribute of the signal; performing unified characterization processing on the feature information fragments subjected to redundancy elimination to form consistent expression forms of the feature information fragments under different core feature dimensions; And screening out characteristic information fragments which can stably reflect the essential attribute of a single signal unit contained in the signal data set and have scene recognition degree from the characteristic information fragments after unified characterization, and determining the screened characteristic information fragments as characteristic units.
5. The method for recognizing cognitive intelligent communication signals according to claim 3, wherein the feature conversion module for converting feature units in the network structure analysis information into inference feature vectors processable by the cognitive inference model to obtain an inference feature vector set comprises: acquiring attribute description information corresponding to each feature unit in the network structure analysis information, wherein the attribute description information comprises types of the feature units, the number of association units and the distribution of association strength dynamic values; inputting the acquired attribute description information into an attribute coding unit of a feature conversion module, and executing coding processing on the input attribute description information to generate an attribute coding sequence; A vector generation unit of the feature conversion module is called to convert the generated attribute coding sequence into an initial reasoning vector which accords with the input requirement of the cognitive reasoning model; performing adaptation adjustment processing on the generated initial reasoning vector, and adjusting the numerical range of the initial reasoning vector to adapt to the calculation requirement of the cognitive reasoning model; and collecting initial inference vectors which are respectively corresponding to all feature units in the network structure analysis information and are adapted and adjusted, and combining all collected initial inference vectors to form an inference feature vector group.
6. The cognitive intelligent communication signal recognition method according to claim 1, wherein the generating a signal recognition inference intermediate result according to the hierarchical feature inference result, the signal recognition inference intermediate result comprising a preliminary recognition type corresponding to each single signal unit in the signal data set, comprises: Analyzing the reasoning confidence coefficient sequences corresponding to the reasoning levels in the level characteristic reasoning results, and determining the reasoning reliability degree corresponding to each reasoning confidence coefficient sequence; screening out inference level results meeting the inference validity requirements according to the determined inference reliability degree, and eliminating inference level results not meeting the inference validity requirements; performing association matching processing on the filtered reasoning hierarchy result and a signal type mapping relation built in the cognitive reasoning model, and determining candidate recognition types respectively corresponding to each single signal unit in the signal data set; Performing priority ranking processing on candidate recognition types respectively corresponding to each single signal unit in the signal data set, and selecting the candidate recognition type with the highest ranking as the primary recognition type corresponding to the corresponding single signal unit; And integrating the preliminary identification type and the filtered reasoning level result corresponding to each single signal unit in the signal data set, and generating a signal identification reasoning intermediate result containing the corresponding relation between the single signal unit identification and the preliminary identification type.
7. The cognitive intelligent communication signal recognition method according to claim 6, wherein the performing a priority ranking process on the candidate recognition types respectively corresponding to each single signal unit in the signal data set, selecting the candidate recognition type with the highest ranking as the preliminary recognition type corresponding to the corresponding single signal unit includes: determining a sequencing basis influencing the priority of the candidate recognition type, wherein the sequencing basis comprises a numerical value corresponding to an inference confidence sequence, the occurrence frequency of the candidate recognition type in the history recognition process and the adaptation condition of the candidate recognition type and a single signal unit transmission scene; setting importance coefficients for each determined sequencing basis, wherein the importance coefficients represent the influence weights of the corresponding sequencing basis in priority sequencing; Generating a priority value corresponding to each candidate recognition type according to the ranking basis value corresponding to each candidate recognition type and the set importance coefficient; performing high-to-low sequencing on priority values of all candidate identification types corresponding to the same single signal unit to form a priority sequence; and selecting the candidate identification type ranked first in the formed priority sequence, and determining the selected candidate identification type as the primary identification type corresponding to the single signal unit.
8. The cognitive intelligent communication signal recognition method according to claim 1, wherein the generating a final signal recognition result corresponding to the communication signal to be recognized by integrating the signal recognition reasoning intermediate results corresponding to each single signal unit in the signal data set respectively includes: Extracting preliminary identification types respectively corresponding to each single signal unit in a signal data set in the signal identification reasoning intermediate result and a filtered reasoning level result; Counting the proportion of the number of single signal units corresponding to the same primary identification type in the total number of single signal units in the signal data set, and generating type proportion information; Combining the generated type duty ratio information and the reasoning confidence sequences in the filtered reasoning hierarchy results corresponding to the single signal units in the signal data set respectively to generate comprehensive reasoning information corresponding to each primary identification type; performing comparison analysis processing on the comprehensive reasoning information corresponding to all the preliminary recognition types respectively, and selecting the preliminary recognition type with the optimal comprehensive reasoning information as the core recognition type corresponding to the communication signal to be recognized; And integrating the selected core recognition type, comprehensive reasoning information corresponding to each primary recognition type and single signal unit distribution situation corresponding to each primary recognition type, and generating a final signal recognition result corresponding to the communication signal to be recognized.
9. The cognitive intelligent communication signal recognition method as set forth in claim 5, wherein the invoking the vector generation unit of the feature conversion module converts the generated attribute code sequence into an initial inference vector that meets the input requirements of the cognitive inference model, comprising: Acquiring the input vector dimension requirement of a cognitive reasoning model, and determining the dimension specification which needs to be met by an initial reasoning vector; Inputting the generated attribute code sequence into a dimension adaptation subunit of the vector generation unit, and executing dimension adjustment processing on the input attribute code sequence to enable the dimension of the attribute code sequence to be close to the determined dimension specification; The feature supplementing subunit of the vector generation unit is used for executing feature supplementing processing on the attribute coding sequence with the dimension adjusted to perfect the missing dimension feature information in the attribute coding sequence; vector optimization processing is carried out on the attribute coding sequence after feature supplementation, and the numerical value of each dimension in the attribute coding sequence is adjusted to improve the characterization effect of the vector on the feature unit; and converting the optimized attribute coding sequence into a vector form meeting the input requirement of the cognitive reasoning model, and determining the converted vector as an initial reasoning vector.
10. A cognitive intelligent communication signal recognition system, characterized in that the cognitive intelligent communication signal recognition system comprises a processor and a memory, wherein the memory is connected with the processor, the memory is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the memory so as to realize the cognitive intelligent communication signal recognition method according to any one of the claims 1-9.

Description

Cognitive intelligent communication signal identification method and system Technical Field The invention relates to the field of artificial intelligence, in particular to a cognitive intelligent communication signal identification method and a cognitive intelligent communication signal identification system. Background In the field of communication, communication signal identification is a key technology and is widely applied to various aspects such as military communication, civil communication management, spectrum monitoring and the like. The traditional communication signal identification method is mainly based on time domain and frequency domain characteristics of signals, and performs signal classification by setting a fixed threshold or template matching. However, with the continuous development of communication technology, the communication environment is increasingly complex, the signal transmission scene presents diversified characteristics, and the modulation mode, the coding mode and the like of the signal become more complex and changeable. The existing communication signal identification method is difficult to accurately capture the essential characteristics of signals when facing complex and changeable transmission scenes, and has poor adaptability to the dynamic changes of the signal characteristics in different scenes. Moreover, the conventional method generally processes single signal features in isolation, and lacks deep mining and utilization of association relations between the signal features, so that accuracy and reliability of signal identification are limited, and high-accuracy and high-adaptability requirements of a modern communication system on signal identification cannot be met. Disclosure of Invention In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a cognitive intelligent communication signal identification method, which includes: Acquiring a signal data set corresponding to a communication signal to be identified, wherein the signal data set comprises single signal units acquired under different transmission scenes; constructing a signal characteristic dynamic association network based on the characteristic association relation of single signal units contained in the signal data set, wherein the signal characteristic dynamic association network contains characteristic units and dynamic association strength among the characteristic units; Invoking a pre-trained cognitive reasoning model to conduct hierarchical feature reasoning processing on the signal feature dynamic association network, and generating hierarchical feature reasoning results corresponding to each reasoning hierarchy respectively; generating a signal recognition reasoning intermediate result according to the hierarchical characteristic reasoning result, wherein the signal recognition reasoning intermediate result comprises a preliminary recognition type corresponding to each single signal unit in a signal data set; and integrating the signal recognition reasoning intermediate results corresponding to each single signal unit in the signal data set, and generating a final signal recognition result corresponding to the communication signal to be recognized. In yet another aspect, an embodiment of the present invention further provides a cognitive intelligent communication signal identifying system, including a processor, and a machine-readable storage medium, where the machine-readable storage medium is connected to the processor, and the machine-readable storage medium is used to store a program, an instruction, or a code, and the processor is used to execute the program, the instruction, or the code in the machine-readable storage medium, so as to implement the method described above. Based on the above aspects, the embodiment of the invention fully considers the dynamic association relation between the signal characteristics under different transmission scenes by constructing the signal characteristic dynamic association network, can more comprehensively and accurately characterize the characteristic information of the signal, and overcomes the limitation of the isolated processing characteristics of the traditional method. And calling a pre-trained cognitive reasoning model to conduct hierarchical feature reasoning processing, and deeply mining inherent logic and association between signal features from different hierarchies by utilizing powerful reasoning capability of the model to generate a reliable hierarchical feature reasoning result. The signal recognition inference intermediate result and the final signal recognition result are generated based on the inference results, so that the method can adapt to complex and changeable transmission scenes, effectively improve the accuracy and reliability of signal recognition, provide powerful support for stable operation and efficient management of a communicati