A method and apparatus for pulse data augmentation based on modeling, simulation and inversion
By setting the radiation source pulse parameter set and performing simulation inversion processing, the initial pulse data set is generated and optimized, solving the problem of scarce radiation source signal samples. This enables accurate simulation and rich data in complex electromagnetic environments, improving the model's transfer recognition capability and analysis reliability.
Patent Information
- Application Number
- CN202610714103.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-22
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2046-05-22
AI Technical Summary
In existing technologies, the complex and variable electromagnetic environment leads to a scarcity of radiation source signal samples and a limited number of model training samples, resulting in model overfitting, insufficient robustness and transfer recognition capabilities. Traditional data augmentation methods cannot perform targeted augmentation for specific scenarios.
By setting the set of radiation source pulse parameters, an initial pulse data set is generated, and simulation inversion processing is performed, including simulation beat, random generation of pulse characteristic parameters, classification processing, data balance evaluation, and parameter updating, to construct a simulated pulse data set.
It achieves accurate simulation of radiation source signals in complex electromagnetic environments, enriches the diversity and complexity of data, improves the transferability and recognition capability of the model and the reliability of analysis, and provides sufficient data resources to support subsequent research.
Smart Images

Figure CN122452163B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of simulation modeling data processing and data augmentation technology, specifically to a pulse data augmentation method and apparatus based on modeling simulation inversion. Background Technology
[0002] In recent years, deep learning has been widely used due to its powerful feature extraction capabilities. However, obtaining a large number of labeled real radiation source signal samples faces two major difficulties. On the one hand, the electromagnetic environment is complex and variable, and signals are easily contaminated, resulting in a scarcity of effective samples. On the other hand, the number of samples available for training is extremely limited when facing unknown and novel radiation targets. Directly training complex models using a small number of samples leads to overfitting, where the model only memorizes specific details of the training samples rather than learning generalized feature patterns. This causes a sharp decline in robustness in real complex electromagnetic environments and the ability to transfer and identify the location and type of radiation sources. Furthermore, traditional data augmentation or general generation methods are "flood-like" and cannot provide targeted augmentation for specific scenarios, categories, or conditions that affect the model's performance. Summary of the Invention
[0003] This invention primarily addresses the problems of weak model transfer recognition capabilities due to insufficient sample radiation source data and the difficulty in accurately using resources to compensate for the weakest link in the model due to blind data generation. This invention discloses a pulse data enhancement method and device based on modeling simulation inversion.
[0004] In a first aspect, this invention discloses a pulse data augmentation method based on modeling and simulation inversion, comprising: S1, Set the set of radiation source pulse parameters; S2, Based on the set of radiation source pulse parameters, generate an initial pulse data set; S3, perform simulation inversion processing on the initial pulse data set to obtain a simulated pulse data set.
[0005] The set of radiation source pulse parameters includes: S11, The simulation beat and total time are acquired; the simulation beat is used to determine the time interval of signal change, and the total simulation time represents the total duration of the simulation. S12, during the time of each simulation cycle, a corresponding set of pulse characteristic parameter information is randomly generated; the set of pulse characteristic parameter information includes pulse characteristic parameters; the pulse characteristic parameters include pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth; S13. Using the simulated beat and total time, as well as the set of pulse characteristic parameters for each simulated beat, a set of radiation source pulse parameters is constructed.
[0006] The step of generating an initial pulse data set based on the radiation source pulse parameter set includes: S21, within each simulation cycle, several initial pulse data are generated based on the pulse characteristic parameter information set of the simulation cycle; the parameters of each initial pulse data are determined by the pulse characteristic parameters in the pulse characteristic parameter information set. S22, using the initial pulse data of all simulated beats, construct the initial pulse data set.
[0007] The process of performing simulation inversion on the initial pulse data set to obtain a simulated pulse data set includes: S31, classify the initial pulse data of each simulation cycle to obtain a set of category information; S32, perform a data balance evaluation on the category information set to obtain an evaluation result set; S33, Based on the evaluation result set, update the pulse characteristic parameter information set of the simulated beat to obtain an updated pulse characteristic parameter information set; S34, within each simulation cycle, a corresponding subset of pulse data is generated based on the updated set of pulse characteristic parameter information; S35 uses a subset of pulse data from all simulated beats to construct a set of simulated pulse data.
[0008] The initial pulse data for each simulation cycle is classified to obtain a set of category information, including: S311, using the preset set of typical feature parameters for each pulse category, perform similarity calculation with each initial pulse data of the simulation beat to obtain a similarity value for each pulse category; S312, determine the pulse category with the minimum similarity value, which is the category information of the initial pulse data of the simulation beat; S313, S311 to S312 are executed for each initial pulse data of the simulation beat to obtain the category information of each initial pulse data; S314: Using the category information of all initial pulse data, a category information set is constructed.
[0009] The data balance evaluation of the category information set, to obtain an evaluation result set, includes: S321, For each simulation cycle, the total number of initial pulse data contained in each category of information in the corresponding category information set is obtained; S322, Perform discreteness calculation on the total number of initial pulse data contained in the category information to obtain the total value of data balance; The expression for calculating the discreteness is: , in, For the total value of data balance, The result is the total number of initial pulse data contained in the i-th category information divided by the total number of initial pulse data of the simulation beat, where N is the total number of categories; S323, Based on the total value of data balance, calculate the unbalanced value of each category of information; The expression for calculating the imbalance value of the i-th category information is: , in, The unbalanced value of the i-th category information; S324, using the unbalanced values of all categories of information, construct the evaluation result set of the simulation beat.
[0010] The step of updating the pulse characteristic parameter information set of the simulated beat based on the evaluation result set to obtain an updated pulse characteristic parameter information set includes: S331, for each simulation cycle, based on the descending order of the total number of initial pulse data contained in all category information, number all category information to obtain the sequence number information of each category information; the category information with the largest total number of initial pulse data contained is numbered 1, and the category information with the smallest total number of initial pulse data contained is numbered M, where M is an even number; S332, update the pulse feature parameters corresponding to the initial pulse data contained in the category information with sequence number information from 1 to M / 2, and obtain the updated pulse feature parameters of the category information; S333, using the updated pulse feature parameters of all categories of the simulated beat information, an updated pulse feature parameter information set is constructed.
[0011] A second aspect of the present invention discloses a pulse data enhancement device based on modeling and simulation inversion, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the pulse data enhancement method based on modeling simulation inversion.
[0012] In a third aspect of this invention, a computer-readable storage medium is disclosed, the computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the pulse data enhancement method based on modeling simulation inversion.
[0013] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the pulse data enhancement method based on modeling simulation inversion.
[0014] The beneficial effects of this invention are as follows: This invention, by setting the simulation cycle and total time, can precisely control the time interval of simulated signal changes and the duration of the entire simulation process, providing a foundation for simulating radiation source signals at different time scales. Within each simulation cycle, a set of pulse characteristic parameters is randomly generated, covering key parameters such as pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth. This greatly enriches the diversity and complexity of the simulated signal, enabling a more realistic reflection of the random variations of these parameters in actual radiation source signals. Using these generated parameters to construct a radiation source pulse parameter set provides a comprehensive and reliable parameter basis for the subsequent generation of initial pulse data. This allows the generated initial pulse data to simulate realistic and complex radiation source signal scenarios, laying a solid foundation for further processing and analysis of the pulse data.
[0015] Within each simulation cycle, this invention generates several initial pulse data points based on a pre-set set of pulse characteristic parameter information. The parameters of these initial pulse data points are precisely determined by the pulse characteristic parameters in the pulse characteristic parameter information set. By constructing an initial pulse data set using the initial pulse data from all simulation cycles, this method comprehensively covers pulse data at different time points, resulting in a rich and diverse set capable of simulating complex and varied pulse signal scenarios in real-world environments. This provides ample data resources for subsequent in-depth analysis and processing of the pulse data, acting as a solid data foundation for future research. This allows subsequent research to proceed on this rich and diverse data basis, greatly improving the reliability and accuracy of the research results. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation
[0017] To better understand the content of this invention, an embodiment is provided here.
[0018] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.
[0019] In a first aspect, this invention discloses a pulse data augmentation method based on modeling and simulation inversion, comprising: S1, Set the set of radiation source pulse parameters; S2, Based on the set of radiation source pulse parameters, generate an initial pulse data set; S3, perform simulation inversion processing on the initial pulse data set to obtain a simulated pulse data set.
[0020] The set of radiation source pulse parameters includes: S11, The simulation beat and total time are acquired; the simulation beat is used to determine the time interval of signal change, and the total simulation time represents the total duration of the simulation. S12, during the time of each simulation cycle, a corresponding set of pulse characteristic parameter information is randomly generated; the set of pulse characteristic parameter information includes pulse characteristic parameters; the pulse characteristic parameters include pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth; S13. Using the simulated beat and total time, as well as the set of pulse characteristic parameters for each simulated beat, a set of radiation source pulse parameters is constructed.
[0021] The acquisition of simulated beats and total time allows for the analysis of the duration of the actually acquired electromagnetic pulses to obtain the total time, and the analysis of the variation period of the actually acquired electromagnetic pulses to obtain the simulated beats.
[0022] The step of generating an initial pulse data set based on the radiation source pulse parameter set includes: S21, within each simulation cycle, several initial pulse data are generated based on the pulse characteristic parameter information set of the simulation cycle; the parameters of each initial pulse data are determined by the pulse characteristic parameters in the pulse characteristic parameter information set. S22, using the initial pulse data of all simulated beats, construct the initial pulse data set.
[0023] The process of performing simulation inversion on the initial pulse data set to obtain a simulated pulse data set includes: S31, classify the initial pulse data of each simulation cycle to obtain a set of category information; S32, perform a data balance evaluation on the category information set to obtain an evaluation result set; S33, Based on the evaluation result set, update the pulse characteristic parameter information set of the simulated beat to obtain an updated pulse characteristic parameter information set; S34, within each simulation cycle, a corresponding subset of pulse data is generated based on the updated set of pulse characteristic parameter information; S35 uses a subset of pulse data from all simulated beats to construct a set of simulated pulse data.
[0024] The initial pulse data for each simulation cycle is classified to obtain a set of category information, including: S311, using the preset set of typical feature parameters for each pulse category, perform similarity calculation with each initial pulse data of the simulation beat to obtain a similarity value for each pulse category; S312, determine the pulse category with the minimum similarity value, which is the category information of the initial pulse data of the simulation beat; S313, S311 to S312 are executed for each initial pulse data of the simulation beat to obtain the category information of each initial pulse data; S314: Using the category information of all initial pulse data, a category information set is constructed.
[0025] The set of typical characteristic parameters includes typical values for pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth.
[0026] The expression for calculating the similarity is: , in, A vector consisting of the set of typical feature parameters of a pulse category. This is a vector composed of pulse characteristic parameters corresponding to the initial pulse data of the simulated beat. Indicates calculation and cosine similarity, This represents the similarity value between pulse categories. and They represent the calculated results. of Distance and Frobenius norm.
[0027] The similarity calculation expression of this invention comprehensively considers the cosine similarity between vectors, Distance and Frobenius norm. This comprehensive consideration allows for a thorough and in-depth measurement of the similarity between the initial impulse data and the set of typical feature parameters. Cosine similarity reflects the similarity of vector directions. Distance reflects the absolute difference in distance between vectors in space, while the Frobenius norm measures the difference between vectors from a matrix perspective. The combination of these three factors provides a highly accurate basis for determining the category of initial impulse data, significantly improving classification accuracy and making the classification results more reliable.
[0028] The initial pulse data of each simulation beat is classified to obtain a set of category information. This can be done by using the PCA clustering algorithm to process the initial pulse data of the simulation beat based on a preset set of typical feature parameters for each pulse category, obtaining the category information for each initial pulse data, and then using the category information of all initial pulse data to obtain a set of category information.
[0029] The data balance evaluation of the category information set, to obtain an evaluation result set, includes: S321, For each simulation cycle, the total number of initial pulse data contained in each category of information in the corresponding category information set is obtained; S322, perform discreteness calculation on the total number of initial pulse data contained in each category of information to obtain the total value of data balance; The expression for calculating the discreteness is: , in, For the total value of data balance, The result is the total number of initial pulse data contained in the i-th category information divided by the total number of initial pulse data of the simulation beat, where N is the total number of categories; S323, Based on the total value of data balance, calculate the unbalanced value of each category of information; The expression for calculating the imbalance value of the i-th category information is: , in, The unbalanced value of the i-th category information; S324, using the unbalanced values of all categories of information, construct the evaluation result set of the simulation beat.
[0030] The expression for calculating dispersion quantifies the data distribution balance using the entropy principle, providing clear guidance for data augmentation optimization. It transforms the proportion of data in each category into directly comparable values. When the number of data in all categories is similar, the calculated total balance value is higher, intuitively reflecting the uniformity of the data distribution; when the number of data in a certain category is extremely small or extremely large, the total value will decrease significantly, clearly exposing the data imbalance problem. This avoids the subjectivity of manual judgment of data balance. Through specific numerical values, the degree of imbalance in the current dataset can be clearly identified without relying on experience, ensuring that subsequent parameter adjustments have clear goals and measurement standards.
[0031] The expression for calculating the imbalance value, based on the total balance value, further identifies specific imbalance categories, achieving precise optimization "from the whole to the part." By calculating the correlation between a single category and the overall balance, the category contributing most to data imbalance can be directly identified. For example, the larger the imbalance value of a category, the more significant the impact of the deviation in the number of data in that category on the overall balance, allowing for quick identification of categories requiring focused adjustment. Categories with high imbalance values can be prioritized for parameter updates, specifically increasing or decreasing the data generation volume for that category to avoid indiscriminate adjustments and improve the efficiency of data balancing. The imbalance value of each category can serve as a metric for measuring the adjustment effect. After updating the pulse characteristic parameters, the imbalance value can be recalculated, and the effectiveness of the adjustment can be intuitively judged through numerical changes, ensuring that the data balance optimization process is controllable and verifiable.
[0032] The step of updating the pulse characteristic parameter information set of the simulated beat based on the evaluation result set to obtain an updated pulse characteristic parameter information set includes: S331, for each simulation cycle, based on the descending order of the total number of initial pulse data contained in all category information, number all category information to obtain the sequence number information of each category information; the category information with the largest total number of initial pulse data contained is numbered 1, and the category information with the smallest total number of initial pulse data contained is numbered M, where M is an even number; S332, update the pulse feature parameters corresponding to the initial pulse data contained in the category information with sequence number information from 1 to M / 2, and obtain the updated pulse feature parameters of the category information; S333, using the updated pulse feature parameters of all categories of the simulated beat information, an updated pulse feature parameter information set is constructed.
[0033] The expression for calculating the updated pulse feature parameters for category information with sequence number i is: , in, For the category information with sequence number i, update the feature parameters of the j-th pulse. This refers to the j-th pulse feature parameter corresponding to the initial pulse data contained in the category information with sequence number i. For the unbalanced values of category information with index M+1-i, The typical value of the j-th pulse feature parameter in the typical feature parameter set of category information with sequence number M+1-i; The updated pulse characteristic parameter calculation expression cleverly utilizes the imbalanced values of category information and typical characteristic parameters to update the pulse characteristic parameters. In this process, the imbalanced values serve as the basis for adjustment, while the typical characteristic parameters act as a reference standard; the two work together to precisely adjust the pulse characteristic parameters. This update calculation makes the data characteristics more aligned with actual needs, more representative, and more reasonable, thereby enhancing the effectiveness of the data in subsequent analysis and providing a higher-quality data foundation for various analyses and applications based on this data.
[0034] The pulse characteristic parameters corresponding to j=1 to 5 are pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth, respectively.
[0035] In all embodiments of the present invention, the values of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.
[0036] In the calculation expressions of this invention, the variables involved have all been dimensionless before calculation.
[0037] A second aspect of the present invention discloses a pulse data enhancement device based on modeling and simulation inversion, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the pulse data enhancement method based on modeling simulation inversion.
[0038] In a third aspect of this invention, a computer-readable storage medium is disclosed, the computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the pulse data enhancement method based on modeling simulation inversion.
[0039] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the pulse data enhancement method based on modeling simulation inversion.
[0040] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A pulse data augmentation method based on modeling, simulation, and inversion, characterized in that, include: S1, Set the set of radiation source pulse parameters, including: S11, The simulation beat and total time are acquired; the simulation beat is used to determine the time interval of signal changes, and the total time represents the total duration of the simulation. S12, during the time of each simulation cycle, a corresponding set of pulse characteristic parameter information is randomly generated; the set of pulse characteristic parameter information includes pulse characteristic parameters; the pulse characteristic parameters include pulse amplitude, carrier frequency, pulse width, pulse period, and pulse bandwidth; S13, using the simulation beat and total time, as well as the pulse characteristic parameter information set of each simulation beat, a radiation source pulse parameter set is constructed; S2, Based on the set of radiation source pulse parameters, generate an initial pulse data set; S3, perform simulation inversion processing on the initial pulse data set to obtain a simulated pulse data set, including: S31, classify the initial pulse data for each simulation cycle to obtain a set of category information, including: S311, using the preset set of typical feature parameters for each pulse category, perform similarity calculation with each initial pulse data of the simulation beat to obtain a similarity value for each pulse category; S312, determine the pulse category with the minimum similarity value, which is the category information of the initial pulse data of the simulation beat; S313, S311 to S312 are executed for each initial pulse data of the simulation beat to obtain the category information of each initial pulse data; S314, using the category information of all initial pulse data, constructs a category information set; S32, perform a data balance evaluation on the category information set to obtain an evaluation result set; S33, Based on the evaluation result set, update the pulse characteristic parameter information set of the simulated beat to obtain an updated pulse characteristic parameter information set; S34, within each simulation cycle, a corresponding subset of pulse data is generated based on the updated set of pulse characteristic parameter information; S35 uses a subset of pulse data from all simulated beats to construct a set of simulated pulse data.
2. The pulse data enhancement method based on modeling and simulation inversion as described in claim 1, characterized in that, The generation of an initial pulse data set based on the radiation source pulse parameter set includes: S21, within each simulation cycle, several initial pulse data are generated based on the pulse characteristic parameter information set of the simulation cycle; the parameters of each initial pulse data are determined by the pulse characteristic parameters in the pulse characteristic parameter information set. S22, using the initial pulse data of all simulated beats, construct the initial pulse data set.
3. The pulse data augmentation method based on modeling and simulation inversion as described in claim 1, characterized in that, The data balance evaluation of the category information set yields an evaluation result set, including: S321, For each simulation cycle, the total number of initial pulse data contained in each category of information in the corresponding category information set is obtained; S322, Perform discreteness calculation on the total number of initial pulse data contained in the category information to obtain the total value of data balance; The expression for calculating the discreteness is: , in, For the total value of data balance, The result is the total number of initial pulse data contained in the i-th category information divided by the total number of initial pulse data of the simulation beat, where N is the total number of categories; S323, Based on the total value of data balance, calculate the unbalanced value of each category of information; The expression for calculating the imbalance value of the i-th category information is: in, The unbalanced value of the i-th category information; S324, using the unbalanced values of all categories of information, construct the evaluation result set of the simulation beat.
4. The pulse data augmentation method based on modeling and simulation inversion as described in claim 1, characterized in that, The step of updating the pulse characteristic parameter information set of the simulated beat based on the evaluation result set to obtain an updated pulse characteristic parameter information set includes: S331, for each simulation cycle, based on the descending order of the total number of initial pulse data contained in all category information, number all category information to obtain the sequence number information of each category information; the category information with the largest total number of initial pulse data contained is numbered 1, and the category information with the smallest total number of initial pulse data contained is numbered M, where M is an even number; S332, update the pulse feature parameters corresponding to the initial pulse data contained in the category information with sequence number information from 1 to M / 2, and obtain the updated pulse feature parameters of the category information; S333, using the updated pulse feature parameters of all categories of the simulated beat information, an updated pulse feature parameter information set is constructed.
5. A pulse data enhancement device based on modeling, simulation, and inversion, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the pulse data augmentation method based on modeling simulation inversion as described in any one of claims 1 to 4.
6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when invoked by a computer, are used to execute the pulse data augmentation method based on modeling simulation inversion as described in any one of claims 1 to 4.
7. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the pulse data enhancement method based on modeling simulation inversion as described in any one of claims 1 to 4.
Citation Information
Patent Citations
Radiation source identification method based on deep Bayesian clustering and category association
CN119513723A
Rotating shaft simulation data enhancement method and system combined with generative adversarial network
CN121389804A
