A method and system for signal modulation recognition under interference conditions based on neural networks

Abstract

This invention provides a method and system for signal modulation identification in interference environments based on neural networks. The method involves acquiring the interference environment signal to be identified, inputting it into an interference adaptation module to perform correlation modeling between interference features and signal features, obtaining interference adaptation features; then, inputting the interference adaptation features into the interference suppression layer of the signal modulation identification neural network for interference component separation, obtaining signal purification features; next, inputting the signal purification features into a modulation classification layer for modulation mode matching, obtaining preliminary modulation type identification results; then, correlating and verifying the preliminary modulation type identification results with the interference type information in the interference adaptation features, obtaining the final modulation type identification result; finally, generating a signal identification report containing modulation parameter descriptions based on the final result and sending it to the target signal analysis terminal. This invention can effectively cope with complex interference environments, improve the accuracy and reliability of signal modulation identification, and provide strong support for the optimization of communication systems.

Description

Technical Field

[0001] This invention relates to the field of modern communications, and more specifically, to a method and system for signal modulation recognition under interference environments based on neural networks. Background Technology

[0002] In the field of modern communications, signal modulation identification is a crucial step in ensuring the stable operation of communication systems and achieving efficient information transmission. However, with the increasing complexity of communication environments and the growing number of interference factors, signal modulation identification under interference conditions faces significant challenges.

[0003] Traditional signal modulation recognition methods primarily rely on the extraction of time-domain and frequency-domain features of the signal, combined with simple classification algorithms. However, in interference environments, the interfering signal and the target modulation signal intertwine, severely interfering with the features extracted by traditional methods and leading to a significant drop in recognition accuracy. For example, in the presence of strong noise interference, the random nature of the noise can mask the key features of the modulation signal, making it difficult for methods based on fixed feature extraction and classification to accurately distinguish different modulation types.

[0004] Furthermore, while existing neural network-based signal modulation identification methods have improved recognition performance to some extent, most have not fully considered the complex relationship between interference and signal characteristics, nor the varying impacts of interference on different modulation types. In terms of interference suppression, the lack of effective interference component separation mechanisms makes it impossible to accurately remove interference components, thus affecting the accuracy of subsequent modulation classification. Therefore, there is an urgent need for a method capable of accurately identifying signal modulation types in interference environments to meet the high signal processing requirements of modern communication systems. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a signal modulation identification method based on neural networks under interference conditions, the method comprising:

[0006] The interference environment signal to be identified is acquired, and the interference environment signal to be identified is input into a preset interference adaptation module. The interference adaptation module performs correlation modeling processing on the interference environment signal to be identified and the signal features to obtain interference adaptation features.

[0007] The interference adaptation features are input into the interference suppression layer of a pre-trained signal modulation recognition neural network. The interference suppression layer separates the interference components from the interference adaptation features to obtain signal purification features.

[0008] The signal cleansing features are input into the modulation classification layer of the signal modulation recognition neural network. The modulation classification layer performs modulation pattern matching processing on the signal cleansing features to obtain a preliminary modulation type recognition result.

[0009] The preliminary modulation type identification result is correlated and verified with the interference type information in the interference adaptation feature to obtain the final modulation type identification result.

[0010] Based on the final modulation type identification result, a signal identification report containing a description of modulation parameters is generated, and the signal identification report is sent to the target signal analysis terminal.

[0011] In another aspect, embodiments of the present invention also provide a signal modulation recognition system based on neural networks under interference conditions, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0012] Based on the above, this invention, through the construction of an interference adaptation module, a signal modulation recognition neural network and its functional layers (interference suppression layer, modulation classification layer), and in conjunction with key components such as an interference and modulation association rule base and a modulation mode feature base, achieves high-precision and robust identification of signal modulation types in complex interference environments. This neural network-based signal modulation recognition method in interference environments first utilizes the interference adaptation module to model the association between interference features and signal features, effectively extracting interference adaptation features. Subsequently, the interference suppression layer separates interference components from the interference adaptation features to obtain signal purification features, significantly reducing the impact of interference on signal recognition. Next, the modulation classification layer performs modulation mode matching on the signal purification features to obtain preliminary modulation type identification results. Furthermore, by associating and verifying the preliminary identification results with the interference type information in the interference adaptation features, the identification deviation caused by interference is effectively corrected, improving the accuracy of the final modulation type identification. Finally, based on the final identification results, a signal identification report containing modulation parameter descriptions is generated and sent to the target signal analysis terminal, providing comprehensive and accurate data support for signal analysis. Overall, this invention achieves accurate identification of signal modulation types in interference environments through the collaborative work of multiple levels and components, demonstrating significant innovation and practicality. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the execution flow of the signal modulation recognition method based on neural networks under interference environment provided in the embodiments of the present invention.

[0014] Figure 2This is a schematic diagram of exemplary hardware and software components of a signal modulation recognition system based on neural networks under interference conditions provided in an embodiment of the present invention. Detailed Implementation

[0015] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a signal modulation recognition method based on a neural network under interference conditions, according to an embodiment of the present invention. The following is a detailed description of this signal modulation recognition method based on a neural network under interference conditions.

[0016] Step S110: Obtain the interference environment signal to be identified, input the interference environment signal to be identified into the preset interference adaptation module, and perform correlation modeling processing of interference features and signal features on the interference environment signal to be identified through the interference adaptation module to obtain interference adaptation features.

[0017] In the field of wireless communication, signal transmission is often subject to various interferences, such as electromagnetic interference and co-channel interference, placing the received signal in an interfering environment. To accurately identify the modulation type of the signal, it is first necessary to acquire the interference environment signal to be identified. This signal can be a mixed signal received through a communication antenna from different transmitting sources, containing both the useful signal and various interfering signals.

[0018] Subsequently, the acquired interference environment signal to be identified is input into a pre-set interference adaptation module. This interference adaptation module is pre-designed and trained to perform correlation modeling of interference features and signal features. The correlation modeling process involves in-depth analysis of the intrinsic relationship between interference and signal, identifying how interference affects signal features, and how signal features change in the presence of interference.

[0019] For example, in a complex wireless communication scenario, there may be multiple interference signals of different frequencies and intensities. These interference signals will affect the amplitude, phase, and frequency characteristics of the useful signal to varying degrees. The interference adaptation module will conduct a detailed analysis of these effects and establish a correlation model between interference characteristics and signal characteristics by learning from and analyzing a large amount of historical data.

[0020] Through this correlation modeling process, the interference adaptation feature is finally obtained. This interference adaptation feature integrates relevant information from both the interference and the signal, and can more accurately reflect the true characteristics of the signal under interference conditions.

[0021] Step S111: Perform signal segmentation processing on the interference environmental signal to be identified to obtain multiple continuous signal segment units.

[0022] After acquiring the interference environment signal to be identified, it is usually a continuous and complex signal stream, making direct processing of the entire signal quite difficult. Therefore, it is necessary to perform signal segmentation.

[0023] Signal segmentation is the process of dividing a continuous signal stream into multiple consecutive signal segments according to certain rules. The segmentation rules can be determined based on specific application scenarios and requirements; a common method is to divide based on time intervals. For example, a long signal stream can be divided into multiple consecutive signal segments of fixed duration, such as each time interval being a unit.

[0024] Each signal segment unit contains signal information for that time period, which may include characteristics such as signal amplitude, phase, and frequency. By dividing the signal into multiple segment units, it is easier to analyze and process each segment independently, and it is also possible to better capture the signal's changing characteristics over different time periods.

[0025] Step S112: Extract the time-domain fluctuation characteristics and frequency-domain distribution characteristics of each signal segment unit. The time-domain fluctuation characteristics reflect the amplitude variation law of the signal segment unit in the time dimension, and the frequency-domain distribution characteristics reflect the energy distribution law of the signal segment unit in the frequency dimension.

[0026] For each segmented signal unit, it is necessary to extract its key features, namely time-domain fluctuation features and frequency-domain distribution features.

[0027] Time-domain fluctuation characteristics primarily reflect the amplitude variation pattern of a signal over time. Within a signal segment, the signal amplitude changes over time; this change may be periodic or random. By analyzing the amplitude values ​​of the signal at different time points, the time-domain fluctuation characteristics of the signal can be obtained. For example, parameters such as the maximum, minimum, and average amplitude of the signal over a period of time, as well as the rate of change of amplitude, can be calculated. These parameters describe the fluctuation of the signal over time.

[0028] Frequency domain distribution characteristics reflect the energy distribution pattern of a signal along the frequency dimension. Any signal can be decomposed into combinations of different frequency components, and signals with different frequency components have different energies. By performing spectral analysis on a signal, the energy distribution of the signal at different frequencies can be obtained. For example, methods such as Fourier transform can be used to convert the signal from the time domain to the frequency domain, thus obtaining the signal's spectrum. The spectrum shows the energy intensity of the signal at different frequencies; by analyzing the spectrum, the main frequency components, bandwidth, and other characteristics of the signal can be determined.

[0029] These two features are crucial for understanding the characteristics of a signal: time-domain fluctuation features help us understand the dynamic changes of a signal over time, while frequency-domain distribution features allow us to understand the frequency composition and energy distribution of the signal.

[0030] Step S113: Call the pre-built interference type database. The interference type database stores standard interference features corresponding to various standard interference types. Each standard interference feature includes the time-domain fluctuation reference law, frequency-domain energy reference distribution, and interference intensity reference range of the corresponding interference type. Compare the time-domain fluctuation features and frequency-domain distribution features with the standard interference features in the interference type database to determine the interference type information corresponding to each signal segment unit.

[0031] To determine the type of interference contained in each signal segment unit, a pre-built interference type database needs to be accessed. This database was established through extensive experiments and data analysis, and it stores standard interference characteristics corresponding to various standard interference types.

[0032] Each type of standard interference has its own unique time-domain fluctuation reference pattern, frequency-domain energy reference distribution, and interference intensity reference range. For example, a specific type of electromagnetic interference may exhibit periodic amplitude fluctuations in the time domain, be concentrated in certain specific frequency ranges in the frequency domain, and have a certain range of interference intensity.

[0033] The time-domain fluctuation characteristics and frequency-domain distribution characteristics of each extracted signal segment unit are compared with standard interference characteristics in the interference type database. The comparison process involves assessing the similarity between the two in terms of time-domain fluctuation patterns, frequency-domain energy distribution, and interference intensity. For example, similarity indices, such as correlation coefficients, can be calculated. Higher similarity indicates that the interference type contained in the signal segment unit is more likely to match a certain standard interference type in the database.

[0034] Through this comparison process, the interference type information corresponding to each signal segment unit is finally determined. This information includes the interference type name, interference characteristic parameters, etc.

[0035] Step S114: Adjust the correlation modeling parameters based on the interference type information, and perform fusion modeling processing on the time domain fluctuation characteristics, frequency domain distribution characteristics and interference type information through the adjusted correlation modeling parameters to generate a feature correlation matrix.

[0036] After determining the interference type information corresponding to each signal segment unit, the correlation modeling parameters need to be adjusted based on this information. Different interference types affect signal characteristics in different ways and to varying degrees. Therefore, the correlation modeling parameters need to be adjusted according to the specific interference type to ensure that an accurate correlation model between interference features and signal features can be established.

[0037] For example, if a signal segment contains a strong electromagnetic interference, this interference may have a significant impact on the amplitude and frequency of the signal. In this case, the parameters related to amplitude and frequency need to be adjusted accordingly during correlation modeling to better reflect the impact of this interference.

[0038] The adjusted correlation modeling parameters are used to perform fusion modeling of time-domain fluctuation characteristics, frequency-domain distribution characteristics, and interference type information. The fusion modeling process integrates these different types of information and establishes the correlations between them through specific algorithms and models. For example, machine learning algorithms, such as neural networks, can be used, taking time-domain fluctuation characteristics, frequency-domain distribution characteristics, and interference type information as input, and training the model to learn the correlation patterns between them.

[0039] Finally, a feature correlation matrix is ​​generated through fusion modeling. This feature correlation matrix contains correlation information between interference features and signal features. It is a multi-dimensional matrix, where each element represents the degree of correlation between different features. The feature correlation matrix can more comprehensively and accurately reflect the relationship between interference and signal.

[0040] Step S115: Perform dimension normalization on the feature correlation matrix so that the normalized feature correlation matrix meets the preset feature dimension requirements, and use the normalized feature correlation matrix as the interference adaptation feature.

[0041] After generating the feature correlation matrix, since different signal segment units may have different feature dimensions, the feature correlation matrix needs to be normalized for easier subsequent processing and analysis.

[0042] Dimension normalization adjusts the dimensions of the feature association matrix to meet preset feature dimension requirements. These preset requirements are determined based on the needs of subsequent processing modules; for example, a subsequent interference suppression layer may have specific requirements for the dimensions of the input features.

[0043] In the process of dimensionality normalization, various methods can be employed, such as feature selection and feature extraction. Feature selection involves choosing the most representative features from the original feature correlation matrix and removing redundant features, thereby reducing the feature dimensionality. Feature extraction, on the other hand, transforms and combines the original features to generate new features, achieving the goal of reducing dimensionality.

[0044] After dimension normalization, a normalized feature correlation matrix is ​​obtained. This feature correlation matrix meets the preset requirements in terms of dimension and retains the key correlation information between interference features and signal features. The normalized feature correlation matrix is ​​used as the interference adaptation feature.

[0045] Step S120: Input the interference adaptation features into the interference suppression layer of the pre-trained signal modulation recognition neural network to separate the interference components and obtain the signal purification features.

[0046] After obtaining the interference adaptation features, they are input into the interference suppression layer of a pre-trained signal modulation recognition neural network. The main function of this interference suppression layer is to perform interference component separation processing on the interference adaptation features to remove interference components from the signal and obtain pure signal features.

[0047] Signal modulation recognition neural networks are trained on a large amount of data and possess powerful feature processing and classification capabilities. The interference suppression layer is an important component of this network. It uses specific algorithms and models to analyze and process the interference features of the input signal, identify the interference components, and separate them from the signal.

[0048] Step S121: Input the interference adaptation features into the feature decomposition sub-layer of the interference suppression layer. The feature decomposition sub-layer uses an adaptive decomposition algorithm to split the interference adaptation features into feature components, resulting in multiple feature component units.

[0049] After the interference adaptation features are input into the feature decomposition sub-layer of the interference suppression layer, the feature decomposition sub-layer uses an adaptive decomposition algorithm to split its feature components.

[0050] Adaptive decomposition algorithms can automatically adjust the decomposition method and parameters according to the specific characteristics and structure of the interference adaptation features to achieve the best feature component splitting effect. Different interference adaptation features may have different complexities and feature structures, and adaptive decomposition algorithms can flexibly adapt to these changes, splitting the interference adaptation features into multiple feature component units.

[0051] For example, a complex interference adaptation feature may contain multiple different types of interference components and signal components. An adaptive decomposition algorithm can break it down into different feature component units, each corresponding to a specific interference component or signal component.

[0052] Step S122: Perform interference contribution calculation on each feature component unit, and determine the interference component label and signal component label corresponding to each feature component unit based on the calculated interference contribution.

[0053] For each feature component unit obtained from the splitting, its interference contribution needs to be calculated to determine whether the unit belongs to the interference component or the signal component.

[0054] The calculation of interference contribution is based on the similarity between the feature component unit and the interference feature. In the previous steps, an interference type database has been established, containing feature information for various standard interference types. By comparing each feature component unit with a standard interference feature, the similarity between that unit and the interference feature can be obtained.

[0055] The higher the similarity, the more likely the feature component is to be an interference component, and the higher its interference contribution; the lower the similarity, the more likely the component is to be a signal component, and the lower its interference contribution.

[0056] Based on the calculated interference contribution, a corresponding interference component label and signal component label are determined for each feature component unit. If the interference contribution of a feature component unit exceeds a certain preset threshold, it is labeled as an interference component; if the interference contribution is below the threshold, it is labeled as a signal component.

[0057] Step S1221: Extract standard interference feature templates from the interference type database that match the current interference type information. The standard interference feature templates include the fluctuation periodicity of the interference in the time domain, the frequency band pattern of the interference energy concentration in the frequency domain, the maximum amplitude range of the interference, the peak energy range of the interference, the attenuation coefficient range of the interference on the signal amplitude, and the offset range of the interference on the signal frequency.

[0058] To calculate the interference contribution of the feature component unit, it is necessary to extract the standard interference feature template that matches the current interference type information from the interference type database.

[0059] In the previous steps, the interference type information corresponding to each signal segment unit has been determined. Based on this information, a standard interference feature template that matches the signal segment unit is searched from the interference type database.

[0060] The standard interference feature template contains multi-faceted information about the interference, such as the fluctuation periodicity of the interference in the time domain, which describes the periodic changes of the interference signal over time; the energy concentration frequency band pattern of the interference in the frequency domain, which indicates the main energy distribution frequency band of the interference signal in the frequency dimension; the maximum amplitude range and peak energy range of the interference, which can help understand the intensity of the interference; and the range of the attenuation coefficient of the interference on the signal amplitude and the range of the frequency offset of the interference on the signal, which describe the degree of influence of the interference on the signal characteristics.

[0061] By extracting the standard interference feature template, each feature component unit can be compared with the template in detail, thereby accurately calculating its interference contribution.

[0062] Step S1222: Calculate the feature similarity between each feature component unit and the standard interference feature template. The feature similarity reflects the degree of agreement between the feature component unit and the standard interference feature template in terms of interference time-domain fluctuation period, interference frequency-domain energy concentration frequency band, interference amplitude maximum range, interference energy peak range, interference-to-signal amplitude attenuation coefficient range, and interference-to-signal frequency offset range.

[0063] After extracting the standard interference feature template, it is necessary to calculate the feature similarity between each feature component unit and the template.

[0064] Feature similarity is calculated by comparing the features of the feature component units and the standard interference feature template in multiple aspects. Specifically, it compares their degree of similarity in terms of interference time-domain fluctuation periodicity, interference frequency-domain energy concentration band pattern, maximum interference amplitude range, interference energy peak range, interference amplitude attenuation coefficient range, and interference frequency offset range.

[0065] For example, regarding the time-domain fluctuation periodicity, we can compare whether the fluctuation period of the characteristic component unit is consistent with that of the standard interference characteristic template; for the frequency domain energy concentration band pattern, we can compare the similarity between the two in terms of main frequency components and frequency bandwidth. By comprehensively comparing these aspects, we can obtain a comprehensive feature similarity index, which can reflect the degree of similarity between the characteristic component unit and the standard interference characteristic template.

[0066] Step S1223: Based on feature similarity, set the interference contribution calculation formula, substitute the feature similarity into the interference contribution calculation formula, and obtain the interference contribution value of each feature component unit.

[0067] Based on the calculated feature similarity, a formula for calculating the interference contribution is established. This formula is determined by the relationship between feature similarity and interference contribution, and can typically employ a linear or non-linear functional relationship.

[0068] For example, feature similarity can be used as a weighting factor, multiplied by a preset coefficient to obtain the interference contribution value. Alternatively, a more complex nonlinear function can be used to calculate different interference contribution values ​​based on different ranges of feature similarity.

[0069] By substituting the feature similarity of each feature component unit into the interference contribution calculation formula, the interference contribution value of that unit can be obtained. This value can intuitively reflect the degree of contribution of that feature component unit to the interference.

[0070] Step S1224: Set the interference contribution threshold and compare the interference contribution value of each feature component unit with the interference contribution threshold.

[0071] To determine whether each feature component is an interference component or a signal component, an interference contribution threshold needs to be set. The interference contribution threshold is determined based on the actual application scenario and requirements; it serves as a boundary to distinguish between interference components and signal components.

[0072] The interference contribution value of each feature component unit is compared with the interference contribution threshold. If the interference contribution value is greater than or equal to the interference contribution threshold, it indicates that the feature component unit is more likely to be an interference component; if the interference contribution value is less than the interference contribution threshold, it indicates that the unit is more likely to be a signal component.

[0073] Step S1225: If the interference contribution value is greater than or equal to the interference contribution threshold, then label the feature component unit with an interference component label; if the interference contribution value is less than the interference contribution threshold, then label the feature component unit with a signal component label.

[0074] Based on the comparison results, each feature component unit is labeled accordingly. If the interference contribution value of a feature component unit is greater than or equal to the interference contribution threshold, it is labeled as an interference component, indicating that the unit is an interference component; if the interference contribution value is less than the interference contribution threshold, it is labeled as a signal component, indicating that the unit is a signal component.

[0075] In this way, all characteristic component units can be divided into two categories: interference components and signal components.

[0076] Step S123: Based on the interference component label and the signal component label, the component filtering sub-layer of the interference suppression layer retains the feature component units with the signal component label and removes the feature component units with the interference component label to obtain the signal feature component set.

[0077] After obtaining the label of each feature component unit, it is filtered through the component filtering sublayer of the interference suppression layer.

[0078] The component filtering sublayer retains feature component units labeled with the signal component tag and discards feature component units labeled with the interference component tag, based on the interference component tag and the signal component tag. This effectively removes interference components from the signal and retains only the pure signal components.

[0079] After filtering, a set of signal feature components is obtained. This set of signal feature components contains only feature component units related to the signal, and these units can more accurately reflect the true characteristics of the signal.

[0080] Step S124: Input the set of signal feature components into the component recombination sublayer of the interference suppression layer, and perform ordered recombination processing on the set of signal feature components through the component recombination sublayer to generate a recombined feature matrix.

[0081] After obtaining the set of signal feature components, it is input into the component reconstruction sublayer of the interference suppression layer. The main function of this component reconstruction sublayer is to perform ordered reconstruction of the set of signal feature components to restore the original structure and features of the signal.

[0082] In the preceding eigenvalue decomposition sublayer, the signal is broken down into multiple feature component units, the order of which may be shuffled. The component recombination sublayer reassembles these feature component units according to certain rules and algorithms, forming an ordered whole.

[0083] Ordered recombination can be performed using various methods, such as those based on time order or feature correlation. Through ordered recombination, a reconstructed feature matrix is ​​generated. This reconstructed feature matrix contains the pure signal features and restores the original structure and feature relationships of the signal.

[0084] Step S125: Perform noise filtering on the reconstructed feature matrix to eliminate residual interference noise components in the reconstructed feature matrix, and use the filtered reconstructed feature matrix as the signal purification feature.

[0085] After generating the reconstructed feature matrix, although most of the interference components have been removed, some residual interference noise may still remain. To further improve the purity of the signal, noise filtering processing is required on the reconstructed feature matrix.

[0086] Noise filtering is a process that uses specific filters or algorithms to suppress and eliminate noise components in the reconstructed feature matrix. Common noise filtering methods include low-pass filtering, high-pass filtering, and band-pass filtering. These filters can be selected based on the frequency characteristics of the noise.

[0087] After noise filtering, residual interference noise components in the reconstructed feature matrix are eliminated, resulting in the filtered reconstructed feature matrix. This matrix is ​​then used as the signal cleansing feature.

[0088] Step S130: Input the signal purification features into the modulation classification layer of the signal modulation recognition neural network and perform modulation pattern matching processing to obtain preliminary modulation type recognition results.

[0089] After obtaining the signal sanitization features, they are input into the modulation classification layer of the signal modulation recognition neural network. The main function of this modulation classification layer is to perform modulation pattern matching processing on the signal sanitization features to determine the modulation type of the signal.

[0090] The modulation classification layer is a key part of the signal modulation recognition neural network. It analyzes and compares the sanitized features of the input signal and matches them with the standard feature vectors in the pre-trained modulation mode feature library to find the best matching modulation type.

[0091] Step S131: Input the signal purification features into the feature mapping sub-layer of the modulation classification layer, and map the signal purification features to the preset modulation feature space through the feature mapping sub-layer to obtain the modulation space feature vector.

[0092] After the signal purification features are input into the feature mapping sublayer of the modulation classification layer, the feature mapping sublayer maps them into a preset modulation feature space.

[0093] The preset modulation feature space is a multi-dimensional space, which is predefined based on the characteristics of different modulation types. The feature mapping sublayer uses a specific mapping algorithm to convert the signal cleansing features into a vector in the modulation space, namely the modulation space feature vector.

[0094] Mapping algorithms can employ linear or nonlinear mapping, selecting the appropriate method based on the specific characteristics of the signal sanitization features and the structure of the modulation feature space. Through mapping, the signal sanitization features are transformed from the original feature space to the modulation feature space, resulting in different positions and distributions of signals with different modulation types within the modulation feature space.

[0095] Step S132: Call the pre-trained modulation mode feature library. The modulation mode feature library stores standard feature vectors and modulation type identifiers corresponding to various standard modulation types. Each standard feature vector contains phase change law features, amplitude modulation law features, and frequency offset law features of the corresponding modulation type. The modulation type identifier contains the name code and parameter configuration code of the corresponding modulation type.

[0096] To perform modulation mode matching, a pre-trained modulation mode feature library needs to be accessed. This feature library stores standard feature vectors and modulation type identifiers corresponding to various standard modulation types.

[0097] Each standard modulation type has its unique phase change pattern, amplitude modulation pattern, and frequency offset pattern, which constitute the standard feature vector of that modulation type. The standard feature vector is obtained by feature extraction and analysis of a large number of standard signal samples, and it can represent the typical characteristics of that modulation type.

[0098] The modulation type identifier includes the name code and parameter configuration code of the corresponding modulation type. The name code is used to uniquely identify different modulation types, while the parameter configuration code contains specific parameter information of the modulation type, such as modulation depth and carrier frequency.

[0099] By calling the modulation mode feature library, feature information of various standard modulation types can be obtained.

[0100] Step S1321: Collect standard signal samples of various known modulation types. Each modulation type corresponds to multiple standard signal samples under different parameter configurations. The different parameter configurations include different modulation depth configurations, different carrier frequency configurations, and different symbol rate configurations.

[0101] To construct a modulation mode feature library, it is first necessary to collect standard signal samples of various known modulation types. For each modulation type, it is necessary to collect standard signal samples under multiple parameter configurations.

[0102] Different parameter configurations can affect the characteristics of a signal. For example, different modulation depth configurations can lead to different amplitude modulation degrees; different carrier frequency configurations can cause changes in the center frequency of the signal; and different symbol rate configurations can affect the transmission rate and bandwidth of the signal.

[0103] By collecting standard signal samples under various parameter configurations, a more comprehensive understanding of the characteristic variation range of each modulation type can be obtained.

[0104] Step S1322: Perform feature extraction processing on each standard signal sample to obtain the sample feature vector corresponding to each standard signal sample. The sample feature vector contains the phase change law feature, amplitude modulation law feature and frequency offset law feature of the standard signal sample, and the dimension of the sample feature vector is consistent with the dimension of the modulation space feature vector.

[0105] After collecting standard signal samples, feature extraction processing needs to be performed on each sample to obtain the corresponding sample feature vector.

[0106] The sample feature vector mainly contains the phase change pattern, amplitude modulation pattern, and frequency offset pattern of the standard signal sample. The extraction methods for these features are similar to those for the extraction of time-domain fluctuation features and frequency-domain distribution features mentioned earlier. These features can be obtained by analyzing and calculating parameters such as the phase, amplitude, and frequency of the signal.

[0107] To facilitate subsequent comparison and matching, the dimension of the sample feature vector needs to be consistent with the dimension of the modulation space feature vector. This can be achieved by selecting appropriate feature parameters and feature dimensions during the feature extraction process.

[0108] Step S1323: Perform statistical analysis on the multiple sample feature vectors corresponding to each modulation type, and calculate the mean vector and variance matrix of the sample feature vectors for each modulation type. The mean vector reflects the average level of the sample feature vectors of the modulation type in each dimension, and the variance matrix reflects the degree of dispersion of the sample feature vectors of the modulation type in each dimension.

[0109] After obtaining multiple sample feature vectors corresponding to each modulation type, these vectors need to be statistically analyzed.

[0110] Statistical analysis mainly involves calculating the mean vector and variance matrix of the feature vectors for each modulation type. The mean vector is the average value of all feature vectors for that modulation type across all dimensions, reflecting the average level of the feature vectors across all dimensions. The variance matrix is ​​the matrix obtained by calculating the variance of all feature vectors for that modulation type across all dimensions, reflecting the degree of dispersion of the feature vectors across all dimensions.

[0111] By calculating the mean vector and variance matrix, we can gain a more comprehensive understanding of the characteristic distribution of each modulation type.

[0112] Step S1324: Use the mean vector as the standard feature vector corresponding to the modulation type, and use the name and encoding information of the modulation type as the modulation type identifier, wherein the encoding information includes the category code and parameter configuration code of the modulation type.

[0113] Based on the mean vector and variance matrix obtained from statistical analysis, the mean vector is used as the standard feature vector corresponding to this modulation type. The mean vector can represent the typical characteristics of this modulation type, and as a standard feature vector, modulation mode matching can be performed more accurately.

[0114] Simultaneously, the name and encoding information of the modulation type are used as the modulation type identifier. The encoding information includes the category code and parameter configuration code of the modulation type. The category code is used to uniquely identify different modulation types, while the parameter configuration code contains the specific parameter information of the modulation type.

[0115] In this way, a standard feature vector and modulation type identifier are determined for each modulation type, and this information is stored in the modulation mode feature library.

[0116] Step S1325: After organizing the standard feature vectors and modulation type identifiers corresponding to all modulation types according to the preset storage format, store them in the database to form a pre-trained modulation mode feature library. The preset storage format includes the numerical values ​​of each dimension of the standard feature vector, the modulation type name, and the modulation type encoding information for each record.

[0117] After organizing the standard feature vectors and modulation type identifiers corresponding to all modulation types according to the preset storage format, they are stored in the database to form a pre-trained modulation mode feature library.

[0118] The default storage format is designed to facilitate data management and retrieval. Each record contains the numerical values ​​of each dimension of the standard feature vector, the modulation type name, and the modulation type encoding information. This storage method allows for convenient querying and updating of the modulation mode feature library.

[0119] Step S133: Calculate the vector distance between the modulation space feature vector and each standard feature vector in the modulation mode feature library. The vector distance reflects the degree of difference between the two feature vectors in terms of phase change law features, amplitude modulation law features, and frequency offset law features.

[0120] After calling the modulation mode feature library, it is necessary to calculate the vector distance between the modulation space feature vector and each standard feature vector in the library.

[0121] Vector distance is an indicator that measures the degree of difference between two eigenvectors. It reflects the differences between the two eigenvectors in terms of phase change characteristics, amplitude modulation characteristics, and frequency offset characteristics. Common methods for calculating vector distance include Euclidean distance and Manhattan distance.

[0122] By calculating the vector distance, the degree of difference between the modulation space feature vector and each standard feature vector can be quantified.

[0123] Step S134: Select the standard feature vector with the smallest vector distance and obtain the modulation type identifier corresponding to the standard feature vector.

[0124] After calculating the vector distance between the modulation space feature vector and each standard feature vector, the standard feature vector with the smallest vector distance is selected.

[0125] The smallest vector distance indicates that the modulation space feature vector is most similar to the standard feature vector, meaning that the modulation type corresponding to the standard feature vector is most likely the modulation type of the signal. The modulation type identifier corresponding to the standard feature vector is obtained, which includes the modulation type name encoding and parameter configuration encoding.

[0126] Step S135: Generate a preliminary modulation type identification result containing the modulation type name and matching confidence based on the modulation type identifier. The matching confidence is negatively correlated with the vector distance, that is, the smaller the vector distance, the higher the matching confidence.

[0127] Based on the obtained modulation type identifier, a preliminary modulation type identification result containing the modulation type name and matching confidence level is generated.

[0128] The modulation type name can be retrieved through the name code in the modulation type identifier, which clarifies the initially identified modulation type. The matching confidence is determined based on the vector distance. Since the vector distance reflects the degree of difference between the modulation space feature vector and the standard feature vector, the smaller the vector distance, the higher the accuracy of the match. Therefore, the matching confidence is negatively correlated with the vector distance.

[0129] In this way, the generated preliminary modulation type identification results include both the modulation type name and the confidence level of the match.

[0130] Step S140: Perform correlation verification processing between the preliminary modulation type identification result and the interference type information in the interference adaptation features to obtain the final modulation type identification result.

[0131] After obtaining the preliminary modulation type identification result, it is necessary to perform correlation and verification processing with the interference type information in the interference adaptation features to verify the accuracy of the preliminary identification result and obtain the final modulation type identification result.

[0132] Different types of interference can affect the identification of signal modulation types, and some interference may lead to false identification. By performing correlation verification, the impact of interference factors on the identification results can be taken into account, thereby improving the accuracy of identification.

[0133] Step S141: Extract the modulation type name and matching confidence from the preliminary modulation type identification result, and at the same time extract the interference type information from the interference adaptation features. The interference type information includes the interference category name, the interference characteristic parameters, and a description of the degree of interference's impact on the signal.

[0134] First, the modulation type name and matching confidence score are extracted from the preliminary modulation type identification results. The modulation type name clarifies the initially identified modulation type, while the matching confidence score reflects the reliability of the identification results.

[0135] Simultaneously, interference type information is extracted from the interference adaptation features. This information includes the interference category name, characteristic parameters of the interference, and a description of the degree of impact of the interference on the signal. This information is crucial for determining the reasonableness of the preliminary identification results in the current interference environment.

[0136] Step S142: Call the pre-built interference and modulation association rule library. The interference and modulation association rule library stores the correspondence between different interference types and the corresponding modulation types. Each correspondence includes the category name of the interference type, the name list of the corresponding modulation type, and a description of the applicable scenario.

[0137] To perform correlation verification, a pre-built interference and modulation correlation rule library needs to be invoked. This modulation correlation rule library stores the correspondence between different interference types and adapted modulation types.

[0138] Each correspondence includes the category name of the interference type, a list of names of the adapted modulation types, and a description of the applicable scenarios for that correspondence. These correspondences were obtained through extensive experiments and data analysis, and they reflect which modulation types can better guarantee the transmission stability and accuracy of signals under different interference types.

[0139] By calling the interference and modulation association rule library, we can obtain the appropriate modulation type information for different interference types.

[0140] Step S1421: Collect signal transmission test data for various modulation types under different interference types. The signal transmission test data includes the category name of the interference type, the characteristic parameters of the interference, the name of the modulation type, the bit error rate data and the signal transmission success rate data during the signal transmission process.

[0141] To build a rule base for the association between interference and modulation, it is first necessary to collect signal transmission test data of various modulation types under different interference types.

[0142] Transmission tests were conducted on signals with various modulation types under different interference environments. The interference type name, characteristic parameters of the interference, modulation type name, bit error rate (BER) data, and signal transmission success rate data were recorded. This data reflects the signal transmission performance of each modulation type under different interference conditions.

[0143] By collecting a large amount of test data, we can gain a more comprehensive understanding of the relationship between different interference types and modulation types.

[0144] Step S1422: Analyze and process the signal transmission test data. For each type of interference, calculate the average signal transmission success rate of each modulation type under that interference type. The average signal transmission success rate is the ratio of the number of successful signal transmissions in multiple tests to the total number of tests for that modulation type.

[0145] After collecting the signal transmission test data, it needs to be analyzed and processed. For each type of interference, the average signal transmission success rate of each modulation type under that interference type is calculated.

[0146] The average signal transmission success rate is calculated as the ratio of the number of successful signal transmissions in multiple tests to the total number of tests for that modulation type. This average value reflects the stability and reliability of signal transmission for that modulation type under that type of interference.

[0147] By calculating the average signal transmission success rate, the performance of different modulation types under the same interference type can be compared.

[0148] Step S1423: Set a signal transmission success rate threshold. The modulation type whose average signal transmission success rate under each type of interference is greater than or equal to the signal transmission success rate threshold is determined as the appropriate modulation type corresponding to that interference type. The signal transmission success rate threshold is set according to the reliability requirements of the signal transmission scenario.

[0149] Setting a signal transmission success rate threshold is to determine the appropriate modulation type for each type of interference. The signal transmission success rate threshold is set based on the reliability requirements of the signal transmission scenario; different signal transmission scenarios have different reliability requirements, therefore the threshold will also vary.

[0150] Modulation types whose average signal transmission success rate is greater than or equal to the signal transmission success rate threshold under each interference type are identified as the suitable modulation types for that interference type. These modulation types can guarantee a high signal transmission success rate under that interference type, and are therefore considered suitable modulation types.

[0151] Step S1424: Record each type of interference and its corresponding adapted modulation type in the form of association rules. Each association rule includes an interference type identifier, an adapted modulation type list, and a rule confidence score. The interference type identifier includes the interference category name and feature parameter encoding. The adapted modulation type list includes the names of all adapted modulation types under the interference type. The rule confidence score is calculated based on the amount of test data corresponding to the association rule. The larger the amount of test data, the higher the rule confidence score.

[0152] Each interference type and its corresponding adaptive modulation type are recorded as association rules. Each association rule includes an interference type identifier, a list of adaptive modulation types, and a rule confidence level.

[0153] The interference type identifier includes the interference category name and characteristic parameter encoding, used to uniquely identify different interference types. The adapted modulation type list contains the names of all adapted modulation types under this interference type. The rule confidence is calculated based on the amount of test data corresponding to the association rule. The larger the amount of test data, the higher the reliability of the association rule, and therefore the higher the rule confidence.

[0154] In this way, the correspondence between interference types and adaptive modulation types is recorded in the form of association rules.

[0155] Step S1425: Classify and store all association rules according to interference type to form a pre-built interference and modulation association rule library, and regularly update and maintain the association rule library according to new test data. The update and maintenance includes adding new association rules, adjusting the rule confidence of existing association rules, and deleting outdated association rules.

[0156] All association rules are categorized and stored according to interference type, forming a pre-built interference and modulation association rule base. Categorized storage facilitates querying and management of the rule base, improving its utilization efficiency.

[0157] Meanwhile, since the signal transmission environment and interference types may change, the association rule base needs to be updated and maintained regularly based on new test data. This updating and maintenance includes adding new association rules to adapt to emerging interference and modulation types; adjusting the rule confidence of existing association rules and reassessing their reliability based on new test data; and deleting outdated association rules to ensure the accuracy and effectiveness of the rule base.

[0158] Step S143: Query the interference and modulation association rule base, and determine the set of suitable modulation types corresponding to the current interference type information based on the interference category name in the current interference type information. The set of suitable modulation types includes all modulation type names that can maintain signal transmission stability under the current interference type.

[0159] After calling the interference and modulation association rule base, the rule base is queried based on the interference category name in the current interference type information to determine the set of suitable modulation types corresponding to the current interference type information.

[0160] The set of compatible modulation types contains the names of all modulation types that can maintain signal transmission stability under the current interference type. Information about these compatible modulation types can be obtained by querying the rule base.

[0161] Step S144: Determine whether the modulation type name in the preliminary modulation type identification result belongs to the set of compatible modulation types.

[0162] The modulation type name in the preliminary modulation type identification result is compared with the set of suitable modulation types to determine whether the modulation type name belongs to the set of suitable modulation types.

[0163] If the modulation type belongs to the set of suitable modulation types, it means that the initially identified modulation type is reasonable under the current interference environment and has high signal transmission stability; if it does not belong to the set of suitable modulation types, it means that the initial identification result may have errors and further adjustments are needed.

[0164] Step S145: If it belongs to the set of suitable modulation types, the preliminary modulation type identification result is directly used as the final modulation type identification result; if it does not belong to the set of suitable modulation types, the modulation space feature vector output by the feature mapping sublayer of the modulation classification layer is obtained first, and then the vector distance between the modulation space feature vector and each standard feature vector in the set of suitable modulation types is recalculated. The modulation type identifier corresponding to the standard feature vector with the smallest vector distance is selected, and the final modulation type identification result is generated by combining the new matching confidence. The new matching confidence is also negatively correlated with the recalculated vector distance.

[0165] If the modulation type name in the preliminary modulation type identification result belongs to the set of compatible modulation types, it means that the preliminary identification result is reliable, and it can be directly used as the final modulation type identification result.

[0166] If the modulation type does not belong to the set of suitable modulation types, rematching is required. First, obtain the modulation space feature vector output by the feature mapping sublayer of the modulation classification layer. Then, recalculate the vector distance between this vector and each standard feature vector in the set of suitable modulation types. Select the modulation type identifier corresponding to the standard feature vector with the smallest vector distance, and combine it with the new matching confidence to generate the final modulation type identification result. The new matching confidence is also negatively correlated with the recalculated vector distance; the smaller the vector distance, the higher the matching confidence.

[0167] In this way, after correlation verification, the final modulation type identification result is obtained, which improves the accuracy and reliability of the identification.

[0168] Step S150: Generate a signal identification report containing modulation parameter descriptions based on the final modulation type identification result and send it to the target signal analysis terminal.

[0169] After obtaining the final modulation type identification result, a signal identification report containing a description of the modulation parameters needs to be generated based on the result, and the report is sent to the target signal analysis terminal.

[0170] Signal identification reports provide signal analysts with detailed signal information, including modulation type and modulation parameters, which helps them to further analyze and process the signals.

[0171] Step S151: Extract the modulation type name, matching confidence score and corresponding modulation type identifier from the final modulation type identification result. The modulation type identifier includes the modulation type name code and parameter configuration code.

[0172] The modulation type name, matching confidence score, and corresponding modulation type identifier are extracted from the final modulation type identification result. The modulation type name clarifies the final identified modulation type, the matching confidence score reflects the reliability of the identification result, and the modulation type identifier contains the modulation type name code and parameter configuration code.

[0173] Step S152: Call the pre-stored modulation parameter database. The modulation parameter database stores typical modulation parameters corresponding to various modulation types. Each typical modulation parameter includes the modulation depth range, carrier frequency range, symbol rate range, and phase offset range of the modulation type. Query the typical modulation parameters corresponding to the modulation type according to the parameter configuration code in the modulation type identifier.

[0174] The system accesses a pre-stored modulation parameter database, which contains typical modulation parameters for various modulation types. Each typical modulation parameter includes information such as the modulation depth range, carrier frequency range, symbol rate range, and phase offset range for that modulation type.

[0175] Based on the parameter configuration code in the modulation type identifier, query the database for the typical modulation parameters corresponding to that modulation type.

[0176] Step S153: Combine the original signal information of the interference environment signal to be identified. The original signal information includes the signal acquisition time, acquisition location, initial amplitude and initial frequency of the signal. Adjust the typical modulation parameters to obtain the actual modulation parameter description that matches the current signal. The adjustment is based on the degree of fit between the initial state of the signal in the original signal information and the typical modulation parameters.

[0177] The original signal information of the interference environment signal to be identified includes the signal acquisition time, acquisition location, initial amplitude, and initial frequency. This information reflects the initial state and characteristics of the signal.

[0178] Typical modulation parameters are adjusted to obtain an actual modulation parameter description that matches the current signal. This adjustment is based on the degree of fit between the initial state of the signal in the original signal information and the typical modulation parameters. For example, if the initial amplitude and frequency of the original signal differ from the range in the typical modulation parameters, these differences can be used to appropriately adjust the typical modulation parameters to better match the actual situation of the current signal.

[0179] Step S154: Integrate the modulation type name, matching confidence, actual modulation parameter description, and interference type information according to the preset report template to generate a signal identification report. The report template contains fixed information columns and format requirements. The information columns include basic signal information, modulation identification results, interference information, and parameter description. The format requirements include font type, font size, and paragraph spacing.

[0180] The modulation type name, matching confidence level, actual modulation parameter description, and interference type information are integrated according to the preset report template to generate a signal identification report.

[0181] The pre-defined report template includes fixed information sections and formatting requirements. Information sections include: a basic signal information section (recording signal acquisition time, acquisition location, etc.); a modulation identification result section (recording modulation type name and matching confidence level); an interference information section (recording interference type information); and a parameter description section (recording actual modulation parameters). Formatting requirements include font type, font size, and paragraph spacing, ensuring the report's standardization and readability.

[0182] By integrating the data according to the report template, the generated signal identification report is complete in content and format, making it easy for signal analysts to view and analyze.

[0183] Step S155: Establish a communication connection with the target signal analysis terminal, and transmit the signal identification report to the target signal analysis terminal through a preset communication protocol. The preset communication protocol includes the encoding method, verification method and transmission rate control method of data transmission to ensure that the target signal analysis terminal can correctly receive and parse the signal identification report.

[0184] After generating the signal identification report, it needs to be sent to the target signal analysis terminal. First, a communication connection needs to be established with the target signal analysis terminal. This connection can be established via wired or wireless means, the specific method chosen based on the actual situation.

[0185] The signal identification report is transmitted to the target signal analysis terminal via a preset communication protocol. This preset protocol includes data transmission encoding, verification, and transmission rate control methods. The encoding method converts the signal identification report into a suitable transmission format; the verification method ensures the accuracy of data transmission, for example, by using parity checking or cyclic redundancy check; and the transmission rate control method controls the data transmission speed to ensure stability and reliability.

[0186] By using a pre-defined communication protocol, the target signal analysis terminal can correctly receive and parse the signal identification report, providing accurate signal information for signal analysts.

[0187] A signal modulation recognition system based on neural networks under interference conditions.

[0188] This embodiment also provides a signal modulation recognition system based on neural networks under interference conditions. The signal modulation recognition system based on neural networks under interference conditions includes a processor and a memory, and the memory and processor are connected.

[0189] The memory stores programs, instructions, or code that implement the steps of the neural network-based signal modulation recognition method under interference conditions. The processor executes the programs, instructions, or code in the memory to identify the modulation type of signals under interference conditions.

[0190] During system operation, the processor first acquires the interference environment signal to be identified from the receiver, and then processes it sequentially according to the steps of the above method, including interference adaptation, interference suppression, modulation classification, and correlation verification. Finally, it generates a signal identification report and sends the report to the target signal analysis terminal through the transmitter.

[0191] This system enables accurate identification of signal modulation types in interference environments, providing strong support for signal processing and analysis in fields such as wireless communication and radar detection. Simultaneously, the system's design and implementation fully consider data legality, compliance, and privacy protection, ensuring its security and reliability. The construction and training process of the neural network model is documented in detail, including the model's modules, layers, and connections, as well as the specific training steps and parameters, enabling those skilled in the art to implement the invention's solution without difficulty based on the specifications.

[0192] Figure 2 The illustration shows exemplary hardware and software components of a neural network-based signal modulation recognition system 100 for interference environments, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the neural network-based signal modulation recognition system 100 for interference environments and to perform the functions described in this application.

[0193] The neural network-based signal modulation recognition system 100 under interference conditions can be a general-purpose server or a special-purpose server; both can be used to implement the neural network-based signal modulation recognition method under interference conditions of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0194] For example, the neural network-based signal modulation identification system 100 in an interference environment may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the neural network-based signal modulation identification system 100 in an interference environment may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The neural network-based signal modulation identification system 100 in an interference environment also includes an I / O interface 150 between the computer and other input / output devices.

[0195] For ease of explanation, only one processor is described in the signal modulation recognition system 100 under interference environment based on neural network. However, it should be noted that the signal modulation recognition system 100 under interference environment based on neural network in this application may also include multiple processors. Therefore, the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the signal modulation recognition system 100 under interference environment based on neural network executes steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0196] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned signal modulation recognition method based on neural network under interference environment is implemented.

[0197] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A signal modulation recognition method based on neural networks under interference conditions, characterized in that, The method includes: The interference environment signal to be identified is acquired, and the signal is input into a preset interference adaptation module. The interference adaptation module performs correlation modeling processing on the interference environment signal to be identified, combining interference features with signal features, to obtain interference adaptation features, specifically including: The interference environment signal to be identified is segmented into multiple continuous signal segment units. Extract the time-domain fluctuation characteristics and frequency-domain distribution characteristics of each signal segment unit. The time-domain fluctuation characteristics reflect the amplitude variation law of the signal segment unit in the time dimension, and the frequency-domain distribution characteristics reflect the energy distribution law of the signal segment unit in the frequency dimension. A pre-built interference type database is invoked. The interference type database stores standard interference features corresponding to various standard interference types. Each standard interference feature includes the time-domain fluctuation reference law, frequency-domain energy reference distribution, and interference intensity reference range of the corresponding interference type. The time-domain fluctuation feature and the frequency-domain distribution feature are compared with the standard interference features in the interference type database to determine the interference type information corresponding to each signal segment unit. Based on the interference type information, the correlation modeling parameters are adjusted, and the time-domain fluctuation characteristics, the frequency-domain distribution characteristics, and the interference type information are fused and modeled using the adjusted correlation modeling parameters to generate a feature correlation matrix. The feature association matrix is ​​dimensionally normalized to meet the preset feature dimension requirements, and the normalized feature association matrix is ​​used as the interference adaptation feature. The interference adaptation features are input into the interference suppression layer of a pre-trained signal modulation recognition neural network. The interference suppression layer separates the interference components from the interference adaptation features to obtain signal purification features. The signal cleansing features are input into the modulation classification layer of the signal modulation recognition neural network. The modulation classification layer performs modulation pattern matching processing on the signal cleansing features to obtain a preliminary modulation type recognition result. The preliminary modulation type identification result is correlated and verified with the interference type information in the interference adaptation feature to obtain the final modulation type identification result. Based on the final modulation type identification result, a signal identification report containing a description of modulation parameters is generated, and the signal identification report is sent to the target signal analysis terminal.

2. The signal modulation recognition method based on neural networks under interference environment according to claim 1, characterized in that, The process of inputting the interference adaptation features into the interference suppression layer of a pre-trained signal modulation recognition neural network, and performing interference component separation processing on the interference adaptation features through the interference suppression layer to obtain signal cleansing features includes: The interference adaptation features are input into the feature decomposition sub-layer of the interference suppression layer. The feature decomposition sub-layer uses an adaptive decomposition algorithm to split the interference adaptation features into feature components, resulting in multiple feature component units. For each of the feature component units, the interference contribution is calculated, and the interference component label and signal component label corresponding to each of the feature component units are determined based on the calculated interference contribution. Based on the interference component label and the signal component label, the component filtering sublayer of the interference suppression layer retains the feature component units with the signal component label and removes the feature component units with the interference component label to obtain the signal feature component set. The set of signal feature components is input into the component recombination sublayer of the interference suppression layer, and the set of signal feature components is recombined in an orderly manner through the component recombination sublayer to generate a recombined feature matrix. The reconstructed feature matrix is ​​subjected to noise filtering to eliminate residual interference noise components, and the filtered reconstructed feature matrix is ​​used as the signal purification feature.

3. The signal modulation recognition method based on neural networks under interference environment according to claim 2, characterized in that, The step of calculating the interference contribution of each feature component unit and determining the interference component label and signal component label corresponding to each feature component unit based on the calculated interference contribution includes: Extract standard interference feature templates from the interference type database that match the current interference type information. The standard interference feature templates include the fluctuation periodicity of the interference in the time domain, the energy concentration frequency band of the interference in the frequency domain, the maximum amplitude range of the interference, the peak energy range of the interference, the attenuation coefficient range of the interference on the signal amplitude, and the offset range of the interference on the signal frequency. Calculate the feature similarity between each of the feature component units and the standard interference feature template. The feature similarity reflects the degree of agreement between the feature component units and the standard interference feature template in terms of interference time-domain fluctuation period, interference frequency-domain energy concentration band, maximum interference amplitude range, peak interference energy range, interference-to-signal amplitude attenuation coefficient range, and interference-to-signal frequency offset range. Based on the feature similarity, an interference contribution calculation formula is set, and the feature similarity is substituted into the interference contribution calculation formula to obtain the interference contribution value of each feature component unit. An interference contribution threshold is set, and the interference contribution value of each feature component unit is compared with the interference contribution threshold. If the interference contribution value is greater than or equal to the interference contribution threshold, then the feature component unit is labeled with an interference component label; if the interference contribution value is less than the interference contribution threshold, then the feature component unit is labeled with a signal component label.

4. The signal modulation recognition method based on neural networks under interference environment according to claim 1, characterized in that, The step of inputting the signal sanitization features into the modulation classification layer of the signal modulation recognition neural network, and performing modulation pattern matching processing on the signal sanitization features through the modulation classification layer to obtain a preliminary modulation type recognition result includes: The signal purification features are input into the feature mapping sublayer of the modulation classification layer, and the signal purification features are mapped to a preset modulation feature space through the feature mapping sublayer to obtain a modulation space feature vector. The pre-trained modulation mode feature library is invoked. The modulation mode feature library stores standard feature vectors and modulation type identifiers corresponding to various standard modulation types. Each standard feature vector contains phase change law features, amplitude modulation law features and frequency offset law features of the corresponding modulation type. The modulation type identifier contains the name code and parameter configuration code of the corresponding modulation type. Calculate the vector distance between the modulation space feature vector and each standard feature vector in the modulation mode feature library. The vector distance reflects the degree of difference between the two feature vectors in terms of phase change law characteristics, amplitude modulation law characteristics, and frequency offset law characteristics. Filter out the standard feature vector with the smallest vector distance, and obtain the modulation type identifier corresponding to the standard feature vector; Based on the modulation type identifier, a preliminary modulation type identification result containing the modulation type name and matching confidence is generated. The matching confidence is negatively correlated with the vector distance, that is, the smaller the vector distance, the higher the matching confidence.

5. The signal modulation recognition method based on neural networks under interference environment according to claim 4, characterized in that, The process involves calling a pre-trained modulation mode feature library, which stores standard feature vectors and modulation type identifiers corresponding to various standard modulation types, including: Collect standard signal samples of various known modulation types. Each modulation type corresponds to multiple standard signal samples under different parameter configurations. The different parameter configurations include different modulation depth configurations, different carrier frequency configurations, and different symbol rate configurations. For each of the standard signal samples, feature extraction processing is performed to obtain a sample feature vector corresponding to each standard signal sample. The sample feature vector includes the phase change law feature, amplitude modulation law feature, and frequency offset law feature of the standard signal sample, and the dimension of the sample feature vector is consistent with the dimension of the modulation space feature vector. Statistical analysis is performed on multiple sample feature vectors corresponding to each modulation type to calculate the mean vector and variance matrix of the sample feature vectors for each modulation type. The mean vector reflects the average level of the sample feature vectors of the modulation type in each dimension, and the variance matrix reflects the degree of dispersion of the sample feature vectors of the modulation type in each dimension. The mean vector is used as the standard feature vector corresponding to the modulation type, and the name and encoding information of the modulation type are used as the modulation type identifier, wherein the encoding information includes the category code and parameter configuration code of the modulation type. After organizing the standard feature vectors and modulation type identifiers corresponding to all modulation types according to the preset storage format, they are stored in the database to form a pre-trained modulation mode feature library. The preset storage format includes the numerical values ​​of each dimension of the standard feature vector, the modulation type name, and the modulation type encoding information for each record.

6. The signal modulation recognition method based on neural networks under interference environment according to claim 1, characterized in that, The step of associating and verifying the preliminary modulation type identification result with the interference type information in the interference adaptation features to obtain the final modulation type identification result includes: Extract the modulation type name and matching confidence from the preliminary modulation type identification result, and simultaneously extract the interference type information from the interference adaptation feature. The interference type information includes the interference category name, the interference characteristic parameters, and a description of the degree of interference's impact on the signal. The pre-built interference and modulation association rule library is invoked. The interference and modulation association rule library stores the correspondence between different interference types and the appropriate modulation types. Each correspondence includes the category name of the interference type, the name list of the appropriate modulation type, and a description of the applicable scenario of the correspondence. The interference and modulation association rule base is queried, and the set of suitable modulation types corresponding to the current interference type information is determined based on the interference category name in the current interference type information. The set of suitable modulation types includes all modulation type names that can maintain signal transmission stability under the current interference type. Determine whether the modulation type name in the preliminary modulation type identification result belongs to the set of compatible modulation types; If the modulation type belongs to the set of adaptive modulation types, the preliminary modulation type identification result is directly used as the final modulation type identification result. If the modulation type does not belong to the set of adaptive modulation types, the modulation space feature vector output by the feature mapping sublayer of the modulation classification layer is obtained first, and then the vector distance between the modulation space feature vector and each standard feature vector in the set of adaptive modulation types is recalculated. The modulation type identifier corresponding to the standard feature vector with the smallest vector distance is selected, and the final modulation type identification result is generated by combining the new matching confidence. The new matching confidence is also negatively correlated with the recalculated vector distance.

7. The signal modulation recognition method based on neural networks under interference environment according to claim 6, characterized in that, The invocation of a pre-built interference and modulation association rule library, which stores the correspondence between different interference types and adapted modulation types, includes: Collect signal transmission test data for various modulation types under different interference types. The signal transmission test data includes the category name of the interference type, the characteristic parameters of the interference, the name of the modulation type, the bit error rate data and the signal transmission success rate data during the signal transmission process. The signal transmission test data is analyzed and processed. For each type of interference, the average signal transmission success rate of each modulation type under that interference type is calculated. The average signal transmission success rate is the ratio of the number of successful signal transmissions in multiple tests to the total number of tests for that modulation type. Set a signal transmission success rate threshold, and determine the modulation type whose average signal transmission success rate under each type of interference is greater than or equal to the signal transmission success rate threshold as the corresponding adaptive modulation type. The signal transmission success rate threshold is set according to the reliability requirements of the signal transmission scenario. Each type of interference and its corresponding adaptive modulation type are recorded in the form of association rules. Each association rule includes an interference type identifier, an adaptive modulation type list, and a rule confidence score. The interference type identifier includes the interference category name and feature parameter code. The adaptive modulation type list includes the names of all adaptive modulation types under this interference type. The rule confidence score is calculated based on the amount of test data corresponding to the association rule. The larger the amount of test data, the higher the rule confidence score. All association rules are categorized and stored according to interference type to form a pre-built interference and modulation association rule library. The association rule library is regularly updated and maintained based on new test data. The update and maintenance include adding new association rules, adjusting the rule confidence of existing association rules, and deleting outdated association rules.

8. The signal modulation recognition method based on neural networks under interference environment according to claim 1, characterized in that, The step of generating a signal identification report containing modulation parameter descriptions based on the final modulation type identification result and sending the signal identification report to the target signal analysis terminal includes: Extract the modulation type name, matching confidence score, and corresponding modulation type identifier from the final modulation type identification result. The modulation type identifier includes the name code and parameter configuration code of the modulation type. The pre-stored modulation parameter database is invoked. The modulation parameter database stores typical modulation parameters corresponding to various modulation types. Each typical modulation parameter includes the modulation depth range, carrier frequency range, symbol rate range, and phase offset range of the modulation type. The typical modulation parameters corresponding to the modulation type are queried according to the parameter configuration code in the modulation type identifier. Combining the original signal information of the interference environment signal to be identified, which includes the signal acquisition time, acquisition location, initial amplitude, and initial frequency, the typical modulation parameters are adjusted to obtain an actual modulation parameter description that matches the current signal. The adjustment is based on the degree of fit between the initial state of the signal in the original signal information and the typical modulation parameters. The modulation type name, matching confidence, actual modulation parameter description and interference type information are integrated according to the preset report template to generate a signal identification report. The report template contains fixed information columns and format requirements. The information columns include basic signal information column, modulation identification result column, interference information column and parameter description column. The format requirements include font type, font size and paragraph spacing. A communication connection is established with the target signal analysis terminal, and the signal identification report is transmitted to the target signal analysis terminal through a preset communication protocol. The preset communication protocol includes the data transmission encoding method, verification method and transmission rate control method to ensure that the target signal analysis terminal can correctly receive and parse the signal identification report.

9. A signal modulation recognition system based on neural networks under interference conditions, characterized in that, The neural network-based signal modulation recognition system under interference conditions includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the neural network-based signal modulation recognition method under interference conditions as described in any one of claims 1-8.

Images