Method, system and device for identifying radio frequency data signal

Abstract

The present invention relates to the field of radio frequency data identification technology, and in particular to a method, system and device for identifying radio frequency data signals. The method comprises the following steps: acquiring radio frequency data, performing spectrum analysis and fractal clustering to obtain spectrum fractal data. Then, spectrum reconstruction and optimization conversion are performed to obtain low-loss conversion data. Radio frequency signal labels are defined by analyzing spectrum fractal dimensions and characteristic indices. For strong signals, frequency mode analysis and phase correlation analysis are performed to identify strong signals. Simulation completion and auxiliary identification are performed on weak signals to obtain weak signal data; the present invention effectively processes radio frequency data by combining signal processing, feature extraction and classification and identification technologies, extracts useful information therein, and realizes accurate identification and classification of signals, thereby providing important support for the optimization and performance improvement of radio frequency communication systems.

Description

Technical Field

[0001] The present invention relates to the technical field of radio frequency data identification, and in particular to a method, system and device for identifying radio frequency data signals. Background Art

[0002] In the early stages of development, the method of identifying RF data signals mainly relied on pattern matching. This method achieves signal recognition by building a template library in advance and matching the RF signal with the template. However, the pattern matching method has a low tolerance for complex signals and noise, and requires a large number of templates to be built in advance, which makes it difficult to identify new signals. With the development of machine learning and pattern recognition, the method of identifying RF data signals gradually adopts feature extraction and classification. This method extracts the features of RF signals, such as spectral features, time domain features, and statistical features, and then uses classification algorithms to identify signals. This method improves the accuracy and robustness of recognition to a certain extent, but there are still challenges in processing complex signal structures and high-dimensional features. With the rise of deep learning, the method of identifying RF data signals gradually adopts deep neural networks for modeling and training. Deep learning methods can automatically learn the abstract representation and feature representation of signals through multi-level neural network structures, thereby achieving high-accuracy signal recognition. This method has achieved remarkable results in many fields, but the performance of deep learning methods is limited for RF signal data with small sample data volume and difficult labeling. In addition, the processing ability of complex signals is limited, and complex signals cannot be accurately classified and identified. Summary of the invention

[0003] Based on this, it is necessary to provide a method, system and device for identifying radio frequency data signals to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for identifying a radio frequency data signal comprises the following steps:

[0005] Step S1: acquiring radio frequency data; performing spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; performing fractal clustering on the scale spectrum graph to obtain spectrum fractal data;

[0006] Step S2: reconstruct the spectrum of the scale spectrum clustering cluster to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data;

[0007] Step S3: performing spectral fractal analysis on the low-loss conversion data to generate a power spectrum fractal dimension; performing peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain a peak-bandwidth characteristic index; defining a label for the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label;

[0008] Step S4: When the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency characteristic pattern analysis is performed on the strong RF signal tag to generate a strong tag frequency change pattern; a phase characteristic correlation analysis is performed on the strong RF signal tag to obtain signal phase correlation data; based on the strong tag frequency change pattern and the signal phase correlation data, a strong signal identification is performed on the strong RF signal tag to obtain strong signal specific data;

[0009] Step S5: When the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and the weak RF signal tag is simulated and completed based on the preset historical RF data signal to generate a simulated enhanced tag; the weak RF signal tag is assisted in identification based on the simulated enhanced tag to obtain specific data of the weak signal.

[0010] The present invention can convert radio frequency data into scale spectrum diagram and spectrum fractal data through spectrum hierarchical analysis and fractal clustering. Such processing can extract the spectrum characteristics of the signal, and cluster the signal according to the spectrum similarity, providing a basis for subsequent processing and analysis. By performing spectrum reconstruction and spectrum optimization progressive conversion on the scale spectrum clustering cluster, sub-signals of each scale and low-loss conversion data can be extracted. Such processing can decompose and represent the signal according to different spectrum scales, while reducing data loss, which is helpful for subsequent feature analysis and signal processing. Through spectrum fractal analysis and peak-bandwidth characteristic index analysis, the spectrum fractal dimension and peak-bandwidth characteristic index of the signal can be calculated. These indicators can describe the spectrum characteristics and frequency information of the signal, and provide a basis for subsequent signal classification and identification. Through the label definition based on the spectrum fractal dimension and characteristic index, the signal can be classified and marked. Such processing can divide the signal into strong radio frequency signals and weak radio frequency signals according to a preset threshold, providing a basis for subsequent strong signal identification and weak signal complementation, and the specific data and characteristics of the strong signal can be identified by performing frequency characteristic pattern analysis and phase characteristic correlation analysis on the strong radio frequency signal. As for weak RF signals, the recognition and understanding capabilities of weak signals can be enhanced through simulation completion based on historical RF data, providing more complete signal data. Through multiple stages and processing steps, the RF data is analyzed and processed from different angles and features, the spectrum, features and patterns of the signal are extracted, and the classification, recognition and completion of the RF signal are achieved. This can help in signal analysis, communication system optimization, and interference detection applications in the RF field. Therefore, the present invention effectively processes RF data, extracts useful information therein, and realizes accurate recognition and classification of signals by combining signal processing, feature extraction and classification and recognition technologies, thereby providing important support for the optimization and performance improvement of RF communication systems.

[0011] In this specification, a system for identifying a radio frequency data signal is provided, which is used to perform the method for identifying a radio frequency data signal as described above. The system for identifying a radio frequency data signal includes:

[0012] The spectrum fractal module is used to obtain radio frequency data; perform spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; perform fractal clustering on the scale spectrum graph to obtain spectrum fractal data;

[0013] The progressive conversion module is used to reconstruct the spectrum of the scale spectrum diagram to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data;

[0014] The label definition module is used to perform spectral fractal analysis on the low-loss conversion data to generate the power spectrum fractal dimension; perform peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain the peak-bandwidth characteristic index; perform label definition on the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain the radio frequency signal label;

[0015] A strong signal identification module is used to define the RF signal tag as a strong RF signal tag when the power spectrum fractal dimension is greater than a preset fractal dimension threshold and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, and to perform frequency characteristic pattern analysis on the strong RF signal tag to generate a strong tag frequency change pattern; to perform phase characteristic correlation analysis on the strong RF signal tag to obtain signal phase correlation data; to perform strong signal identification on the strong RF signal tag based on the strong tag frequency change pattern and the signal phase correlation data to obtain specific data of the strong signal;

[0016] The weak signal identification module is used to define the RF signal tag as a weak RF signal tag when the power spectrum fractal dimension is less than a preset fractal dimension threshold and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, and simulate and complete the weak RF signal tag based on the preset historical RF data signal to generate a simulated enhanced tag; and perform auxiliary identification on the weak RF signal tag based on the simulated enhanced tag to obtain specific data of the weak signal.

[0017] The benefit of the present invention lies in that the original radio frequency data can be effectively converted into a more structured and characteristic scale spectrum diagram and spectrum fractal data through spectrum hierarchical analysis and fractal clustering, so that the signal characteristics are clearer and distinguishable. By performing spectrum optimization and progressive conversion on the scale sub-signal and spectrum fractal data, it is helpful to reduce the loss in the signal transmission process and improve the quality and reliability of the signal. According to the power spectrum fractal dimension and the peak-bandwidth characteristic index, the low-loss conversion data is labeled and defined, which can effectively distinguish strong radio frequency signals from weak radio frequency signals, and perform corresponding identification and classification. For strong signals, frequency characteristic pattern analysis and phase characteristic correlation analysis can be further performed to obtain specific data of strong signals, which helps to deeply understand and utilize the information of strong signals; while for weak signals, by simulating enhanced labels and auxiliary identification, the recognition ability can be enhanced and the integrity and accuracy of data acquisition can be improved. Different processing is performed for strong signals and weak signals respectively. For strong signals, by frequency characteristic pattern analysis and phase characteristic correlation analysis, the characteristics of the signal can be further understood and specific data information can be obtained therefrom. For weak signals, by simulating enhanced labels and auxiliary identification, the recognition ability of the signal can be improved, so as to obtain more complete data information. Therefore, the present invention effectively processes radio frequency data, extracts useful information therein, and realizes accurate identification and classification of signals by combining signal processing, feature extraction and classification recognition technologies, thereby providing important support for the optimization and performance improvement of radio frequency communication systems.

[0018] The present invention also provides a device for identifying radio frequency data signals, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, for use in the method for identifying radio frequency data signals as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of the steps of a method for identifying a radio frequency data signal;

[0020] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0021] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0022] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0023] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0025] It should be understood that, although the terms "first" and "second" are used here to describe each unit, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiment, the first unit can be referred to as the second unit, and similarly the second unit can be referred to as the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed associated items.

[0026] To achieve this, please refer to Figures 1 to 3 , a method for identifying a radio frequency data signal, comprising the following steps:

[0027] Step S1: acquiring radio frequency data; performing spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; performing fractal clustering on the scale spectrum graph to obtain spectrum fractal data;

[0028] Step S2: reconstruct the spectrum of the scale spectrum clustering cluster to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data;

[0029] Step S3: performing spectral fractal analysis on the low-loss conversion data to generate a power spectrum fractal dimension; performing peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain a peak-bandwidth characteristic index; defining a label for the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label;

[0030] Step S4: When the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency characteristic pattern analysis is performed on the strong RF signal tag to generate a strong tag frequency change pattern; a phase characteristic correlation analysis is performed on the strong RF signal tag to obtain signal phase correlation data; based on the strong tag frequency change pattern and the signal phase correlation data, a strong signal identification is performed on the strong RF signal tag to obtain strong signal specific data;

[0031] Step S5: When the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and the weak RF signal tag is simulated and completed based on the preset historical RF data signal to generate a simulated enhanced tag; the weak RF signal tag is assisted in identification based on the simulated enhanced tag to obtain specific data of the weak signal.

[0032] The present invention can convert radio frequency data into scale spectrum diagram and spectrum fractal data through spectrum hierarchical analysis and fractal clustering. Such processing can extract the spectrum characteristics of the signal, and cluster the signal according to the spectrum similarity, providing a basis for subsequent processing and analysis. By performing spectrum reconstruction and spectrum optimization progressive conversion on the scale spectrum clustering cluster, sub-signals of each scale and low-loss conversion data can be extracted. Such processing can decompose and represent the signal according to different spectrum scales, while reducing data loss, which is helpful for subsequent feature analysis and signal processing. Through spectrum fractal analysis and peak-bandwidth characteristic index analysis, the spectrum fractal dimension and peak-bandwidth characteristic index of the signal can be calculated. These indicators can describe the spectrum characteristics and frequency information of the signal, and provide a basis for subsequent signal classification and identification. Through the label definition based on the spectrum fractal dimension and characteristic index, the signal can be classified and marked. Such processing can divide the signal into strong radio frequency signals and weak radio frequency signals according to a preset threshold, providing a basis for subsequent strong signal identification and weak signal complementation, and the specific data and characteristics of the strong signal can be identified by performing frequency characteristic pattern analysis and phase characteristic correlation analysis on the strong radio frequency signal. As for weak RF signals, the recognition and understanding capabilities of weak signals can be enhanced through simulation completion based on historical RF data, providing more complete signal data. Through multiple stages and processing steps, the RF data is analyzed and processed from different angles and features, the spectrum, features and patterns of the signal are extracted, and the classification, recognition and completion of the RF signal are achieved. This can help in signal analysis, communication system optimization, and interference detection applications in the RF field. Therefore, the present invention effectively processes RF data, extracts useful information therein, and realizes accurate recognition and classification of signals by combining signal processing, feature extraction and classification and recognition technologies, thereby providing important support for the optimization and performance improvement of RF communication systems.

[0033] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of the method for identifying a radio frequency data signal of the present invention. In this example, the method for identifying a radio frequency data signal includes the following steps:

[0034] Step S1: acquiring radio frequency data; performing spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; performing fractal clustering on the scale spectrum graph to obtain spectrum fractal data;

[0035] In the embodiment of the present invention, the target signal source is collected by a high-performance radio frequency acquisition device, the sampling rate is not less than 10MHz, and the acquisition time is not less than 60 seconds. The collected original radio frequency signal contains power spectrum density, phase, and delay parameters. The collected 120 seconds of radio frequency data is subjected to short-time Fourier transform to obtain a power spectrum density matrix in the time-frequency domain, the matrix size is 2048x2048, and the power spectrum density matrix is ​​divided into 128 sub-bands in the time domain and the frequency domain to construct a 128x128 time-frequency matrix, that is, a scale spectrum diagram. A hierarchical clustering algorithm based on Higuchi fractal dimension, Petrosian fractal dimension, and Katz fractal dimension is used to perform cluster analysis on the constructed 128x128 scale spectrum diagram, and three fractal dimension features are calculated for each 128x128 sub-region to obtain a 384-dimensional fractal feature vector; then the Ward clustering algorithm is used to perform hierarchical clustering according to these fractal features, and finally 6 cluster centers are determined to obtain spectrum fractal data.

[0036] Step S2: reconstruct the spectrum of the scale spectrum clustering cluster to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data;

[0037] In the embodiment of the present invention, spectrum reconstruction is performed on six clusters. For each 128x128 cluster, the corresponding sub-signal waveform is reconstructed by inverse Fourier transform to obtain six time-domain sub-signals, and the six sub-signals and spectrum fractal data are progressively converted by a spectrum optimization algorithm based on Hilbert-Huang transform, including: calculating the Hilbert-Huang spectrum for the six sub-signals respectively to obtain the energy time-frequency distribution, fusing the Hilbert-Huang spectrum with the spectrum fractal data, optimizing the time-frequency characteristics of the sub-signals, and reconstructing the optimized time-frequency distribution by using an adaptive filter to obtain six low-distortion optimized sub-signals.

[0038] Step S3: performing spectral fractal analysis on the low-loss conversion data to generate a power spectrum fractal dimension; performing peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain a peak-bandwidth characteristic index; defining a label for the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label;

[0039] In an embodiment of the present invention, the Higuchi fractal dimension and the Petrosian fractal dimension are calculated for the 6 low-loss conversion sub-signals, respectively. It includes: calculating the power spectrum density for each sub-signal and using the Higuchi algorithm and the Petrosian algorithm to calculate the fractal dimension of the power spectrum, respectively, to obtain a 12-dimensional fractal feature vector. For the 6 low-loss conversion sub-signals, the peak power spectrum density and -10dB bandwidth are calculated, respectively, including detecting the peak point of the power spectrum density curve of each sub-signal, recording the peak amplitude and calculating the bandwidth of each sub-signal at -10dB. The peak amplitude and bandwidth are combined into a 6-dimensional peak-bandwidth feature vector. The 12-dimensional fractal feature vector and the 6-dimensional peak-bandwidth feature vector are normalized, and then concatenated into an 18-dimensional feature vector, and the K-means algorithm is used to perform unsupervised clustering on the 18-dimensional feature vector, and 4 cluster centers are determined as signal labels.

[0040] Step S4: When the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency characteristic pattern analysis is performed on the strong RF signal tag to generate a strong tag frequency change pattern; a phase characteristic correlation analysis is performed on the strong RF signal tag to obtain signal phase correlation data; based on the strong tag frequency change pattern and the signal phase correlation data, a strong signal identification is performed on the strong RF signal tag to obtain strong signal specific data;

[0041] In the embodiment of the present invention, if the power spectrum fractal dimension of the signal is greater than the preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than the preset characteristic index threshold, the radio frequency signal tag is defined as a strong radio frequency signal tag, and the frequency characteristic pattern analysis is performed on the strong radio frequency signal tag, using a continuous wavelet transform method, using a high-resolution wavelet basis function, a window length of 100 sampling points, extracting the frequency change pattern, performing phase characteristic correlation analysis on the strong radio frequency signal tag, calculating the phase difference of adjacent sampling points, selecting a sample pair with a phase difference greater than 0.2 radians, and calculating their phase correlation coefficients, and performing strong signal identification based on the strong tag frequency change pattern and the signal phase correlation data. Using a support vector machine (SVM) classifier, the strong radio frequency signal tag is classified and identified according to the frequency change pattern and the phase correlation data as features, and the classification accuracy requirement is set to 90%.

[0042] Step S5: When the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and the weak RF signal tag is simulated and completed based on the preset historical RF data signal to generate a simulated enhanced tag; the weak RF signal tag is assisted in identification based on the simulated enhanced tag to obtain specific data of the weak signal.

[0043] In an embodiment of the present invention, when the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal label is defined as a weak RF signal label, and a historical RF signal database is established. The database contains the time domain waveform and frequency domain characteristic data of various RF signals, and several historical signals that are most similar to the current weak signal label feature are selected from the database, and linear superposition of the time domain and frequency domain is performed to generate a simulated enhanced weak signal label. The weak signal label and the simulated enhanced label are input as features into a support vector machine (SVM) classifier for model training, and new weak RF signals are effectively identified, and their frequency domain and time domain features are output.

[0044] Preferably, step S1 comprises the following steps:

[0045] Step S11: acquiring the highest frequency component of the collected signal; collecting radio frequency signals based on the highest frequency component of the collected signal to obtain radio frequency data;

[0046] Step S12: performing spectrum filtering on the radio frequency data to obtain a filtered radio frequency data signal; performing multi-frequency analysis on the filtered radio frequency data signal to generate radio frequency signal components of various scales;

[0047] Step S13: Performing rapid spectrum analysis on each scale radio frequency signal component to obtain a scale spectrum diagram;

[0048] Step S14: performing cluster analysis on the scale spectrum graph to obtain scale spectrum clusters;

[0049] Step S15: Estimating the fractal dimension of the scaled frequency spectrum clusters to obtain frequency spectrum fractal data.

[0050] The invention is beneficial in that it can ensure that the bandwidth of the acquisition system is sufficient to cover the entire spectrum range of the signal, avoid missing high-frequency components, and thus obtain a complete radio frequency data signal. By removing irrelevant noise through spectrum filtering and then performing multi-frequency analysis, radio frequency signal components at different scales can be obtained. This is conducive to analyzing the spectrum characteristics of the signal from both the overall and local levels. By performing rapid spectrum analysis on radio frequency signal components of each scale and clustering the spectrum graph, the energy concentration characteristics of the signal in different frequency bands can be found. This helps to identify the spectrum distribution characteristics of the signal. By estimating the fractal dimension of the spectrum graph clustering cluster, the nonlinear dynamic characteristic index of the signal can be obtained. This provides important characteristic indicators for subsequent identification, helps to accurately distinguish different types of radio frequency signals, fully explores the multi-scale characteristics of radio frequency signals in the frequency domain and time-frequency domain, provides rich feature analysis basis for subsequent signal identification, and can effectively improve the accuracy and robustness of identification. This method based on signal spectrum and fractal analysis provides important technical support for radio frequency monitoring and intelligent identification applications in complex electromagnetic environments.

[0051] In the embodiment of the present invention, the acquisition system uses a digital-to-analog converter with a sampling rate of 200MHz to sample the input radio frequency signal. The sampled data is analyzed by fast Fourier transform (FFT), and the highest frequency component of the signal is 95MHz. The radio frequency data is filtered using a bandpass filter to filter out frequency components below 5MHz and above 100MHz to obtain a filtered radio frequency data signal, and the filtered signal is subjected to wavelet transform to extract radio frequency signal components in five frequency bands of 5-10MHz, 10-20MHz, 20-40MHz, 40-80MHz and 80-100MHz. Fast Fourier transform (FFT) is performed on the RF signal components of the frequency band to calculate the power spectral density of each frequency band. The scaled spectrum diagram is input into the K-means clustering algorithm for clustering analysis. According to the clustering results, the spectrum diagram is divided into 3-5 different clusters. Each cluster represents a spectrum distribution pattern with similar characteristics. The fractal dimension of the power spectrum of the spectrum cluster cluster is calculated respectively. The power spectrum is plotted as a double logarithmic coordinate graph and its straight line slope is calculated to obtain the fractal dimension characteristics of each cluster.

[0052] Preferably, step S2 comprises the following steps:

[0053] Step S21: performing inverse Fourier transform on the scaled spectrum to obtain reconstructed time-domain sub-signal components; performing multi-layer decomposition on the reconstructed time-domain sub-signal components to obtain sub-signals of various scales;

[0054] Step S22: extracting fractal features from the frequency spectrum fractal data to obtain frequency spectrum fractal feature data;

[0055] Step S23: using the sub-signals of each scale and the spectrum fractal feature data to adaptively reconstruct the RF data to generate an optimized reconstructed spectrum;

[0056] Step S24: performing low-loss conversion on the radio frequency data according to the optimized reconstructed spectrum to obtain low-loss converted data.

[0057] The benefit of the present invention lies in that the time domain sub-signal components of different frequency bands can be obtained by performing an inverse Fourier transform on the scale spectrum. In this way, the characteristics of the signal in the two dimensions of time domain and frequency domain can be fully explored, and rich time-frequency characteristic information can be provided for subsequent adaptive spectrum reconstruction. The fractal dimension and lacunarity index describing the nonlinear dynamic characteristics of the signal can be obtained by extracting the characteristics of the spectrum fractal data. These fractal characteristics can reflect the complexity and irregularity of the signal, and are an important basis for distinguishing different types of signals. The original radio frequency data can be reconstructed by an adaptive method using the time domain sub-signal and spectrum fractal characteristics to generate an optimized reconstructed spectrum. This reconstruction process can effectively reduce noise, highlight the essential characteristics of the signal, and lay a good foundation for subsequent low-loss conversion. Based on the optimized reconstructed spectrum, the radio frequency data is converted into a low-loss signal. This conversion method can retain the spectrum and time domain characteristics of the signal to the greatest extent, reduce the information loss caused by the quantization and encoding process, and ensure the accuracy of signal recognition. The multi-scale characteristics of the radio frequency signal are deeply explored from the three dimensions of time domain, frequency domain and fractal, and high-quality signal data is finally generated through adaptive spectrum reconstruction and low-loss conversion. This provides a reliable foundation for subsequent signal recognition and classification tasks, greatly improving the accuracy and robustness of recognition.

[0058] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0059] Step S21: performing inverse Fourier transform on the scaled spectrum to obtain reconstructed time-domain sub-signal components; performing multi-layer decomposition on the reconstructed time-domain sub-signal components to obtain sub-signals of various scales;

[0060] In an embodiment of the present invention, an inverse Fourier transform (IFT) is performed on the scale spectrum diagram to convert the frequency domain data into time domain data. For example, each frequency component of the scale spectrum diagram is restored to a time domain sub-signal component by an inverse Fourier transform. The reconstructed time domain sub-signal components are multi-layered decomposed to obtain sub-signals of various scales. For example, the time domain sub-signal components are hierarchically decomposed using a wavelet transform or a multi-scale analysis method to obtain sub-signals at different scales. By iteratively applying a wavelet transform or a multi-scale analysis method, the time domain sub-signal components are further decomposed into more detailed sub-signals to obtain higher-scale sub-signals.

[0061] Step S22: extracting fractal features from the frequency spectrum fractal data to obtain frequency spectrum fractal feature data;

[0062] In the embodiment of the present invention, the spectrum data is converted into a box counting method fractal representation, and the spectrum fractal data is subjected to fractal feature extraction to obtain key features describing the spectrum fractal structure, including fractal dimension, fractal spectrum index, and fractal intersection. : For each spectrum fractal data, the corresponding characteristic value is calculated according to the selected fractal feature extraction method. For example, for the fractal dimension, the box counting method or the multifractal analysis method is used to calculate the fractal dimension value, and the extracted spectrum fractal feature data is organized into a unified format.

[0063] Step S23: using the sub-signals of each scale and the spectrum fractal feature data to adaptively reconstruct the RF data to generate an optimized reconstructed spectrum;

[0064] In an embodiment of the present invention, each scale sub-signal is used as input data, and spectrum analysis is performed on each scale sub-signal, including power spectrum density estimation, autocorrelation function calculation, extracting the spectrum characteristics of each scale sub-signal, integrating the spectrum characteristics of each scale sub-signal, forming multimodal sub-spectral feature data, and each sub-spectral feature contains spectrum information at different scales. The multimodal sub-spectral feature data is constructed into a feature vector, each feature vector represents the multimodal sub-spectral feature of a sample, and the feature vector is calculated using a generalized fractal dimension algorithm, such as the Haar fractal dimension or the box counting method, to obtain a generalized fractal dimension. The generalized fractal dimension and the spectrum fractal feature data are used as input, and a dynamic conversion model is established using a machine learning method, such as a neural network or a support vector machine, and the dynamic conversion model is trained using labeled training data, and the model parameters are optimized to maximize the prediction accuracy and generalization ability. The received RF data is input into the dynamic conversion model, and the nonlinear dynamic transformation is simulated to map the data from the original space to the optimized reconstructed spectrum space. The transformation result generated by the dynamic conversion model is used to obtain an optimized reconstructed spectrum, which includes the nonlinear optimization reconstruction of the original RF data.

[0065] Step S24: performing low-loss conversion on the radio frequency data according to the optimized reconstructed spectrum to obtain low-loss converted data.

[0066] In the embodiment of the present invention, the optimized reconstructed spectrum is used as input data. The result of the optimized reconstructed spectrum is converted with low loss, and a method based on compressed sensing theory is used to map high-dimensional radio frequency data to low-dimensional space, retaining the main information while reducing the data dimension to reduce data transmission and storage costs. Through the inverse conversion operation, the low-dimensional conversion data is reconstructed into an approximate representation of the original radio frequency data, retaining the main features and information of the data, and the reconstructed data is verified and adjusted to ensure that no important information is lost during the conversion process.

[0067] Preferably, step S23 includes the following steps:

[0068] Step S231: performing high-dimensional spectrum analysis on each scale sub-signal to obtain multi-modal sub-spectrum feature data;

[0069] Step S232: performing generalized fractal dimension calculation on the multimodal sub-spectrum feature data to obtain the generalized fractal dimension;

[0070] Step S233: dynamically transform and model the generalized fractal dimension and the frequency spectrum fractal feature data to generate a dynamic transformation model;

[0071] Step S234: using the dynamic conversion model to perform nonlinear dynamic transformation simulation on the radio frequency data to obtain an optimized reconstructed spectrum.

[0072] The benefit of the present invention is that by performing high-dimensional spectrum analysis on sub-signals of each scale, characteristic information of different frequency components can be extracted, which helps to more accurately describe the spectral characteristics of the RF signal. The calculation of the generalized fractal dimension can provide a comprehensive measure of multimodal sub-spectral characteristic data, which helps to further understand the complexity and diversity of RF signals. Through dynamic transformation modeling, combined with the generalized fractal dimension and spectral fractal characteristic data, a more dynamic model can be established to better capture the correlation and time-varying nature between RF signal characteristics. The dynamic transformation model is used to simulate the nonlinear dynamic transformation of RF data, which can optimize the reconstructed spectrum and improve the accuracy and fidelity of signal reconstruction. This helps to more accurately restore the spectral characteristics of the original RF signal and provide more reliable signal analysis and recognition results. Combining a variety of analysis methods and modeling techniques, feature extraction and reconstruction of RF signals from multiple angles improves the accuracy and reliability of signal analysis and helps to more deeply understand and identify RF data signals.

[0073] In an embodiment of the present invention, each scale sub-signal is used as input data, and spectrum analysis is performed on each scale sub-signal, including power spectrum density estimation, autocorrelation function calculation, extracting the spectrum characteristics of each scale sub-signal, integrating the spectrum characteristics of each scale sub-signal, forming multimodal sub-spectral feature data, and each sub-spectral feature contains spectrum information at different scales. The multimodal sub-spectral feature data is constructed into a feature vector, each feature vector represents the multimodal sub-spectral feature of a sample, and the feature vector is calculated using a generalized fractal dimension algorithm, such as the Haar fractal dimension or the box counting method, to obtain a generalized fractal dimension. The generalized fractal dimension and the spectrum fractal feature data are used as input, and a dynamic conversion model is established using a machine learning method, such as a neural network or a support vector machine, and the dynamic conversion model is trained using labeled training data, and the model parameters are optimized to maximize the prediction accuracy and generalization ability. The received RF data is input into the dynamic conversion model, and the nonlinear dynamic transformation is simulated to map the data from the original space to the optimized reconstructed spectrum space. The transformation result generated by the dynamic conversion model is used to obtain an optimized reconstructed spectrum, which includes the nonlinear optimization reconstruction of the original RF data.

[0074] Preferably, step S3 comprises the following steps:

[0075] Step S31: performing power spectrum density analysis on the low-loss conversion data to obtain signal power spectrum density data;

[0076] Step S32: Calculate the box dimension of the signal power spectrum density data to generate the power spectrum fractal dimension;

[0077] Step S33: performing independent signal power spectrum peak analysis on the low-loss conversion data to obtain a signal power spectrum peak;

[0078] Step S34: calculating the peak-bandwidth correlation characteristic index of the signal power spectrum peak to obtain the peak-bandwidth characteristic index;

[0079] Step S35: labeling the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label.

[0080] The benefit of the present invention is that through power spectrum density analysis, the spectrum energy distribution of the radio frequency signal can be obtained, which is helpful to understand the frequency characteristics and energy distribution of the signal. Through box dimension calculation, the fractal characteristics of the signal power spectrum density data can be measured, providing a measure of signal complexity and irregularity. The independent signal power spectrum peak analysis can determine the peak position and energy peak size in the signal power spectrum, providing a description of the signal spectrum peak characteristics. The calculation of the peak-bandwidth correlation characteristic index can quantify the degree of correlation between the signal power spectrum peak and its corresponding bandwidth, providing further analysis of the signal spectrum characteristics. Based on the label definition of the power spectrum fractal dimension and the peak-bandwidth characteristic index, the radio frequency signal can be classified and identified according to these characteristic indicators. Through multiple analysis steps and characteristic indicators, the radio frequency signal is comprehensively extracted and analyzed, which improves the accuracy and reliability of signal identification and classification. It can help understand and describe the frequency characteristics, energy distribution, complexity and correlation of radio frequency signals, thereby providing useful information for radio frequency signal processing and application.

[0081] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0082] Step S31: performing power spectrum density analysis on the low-loss conversion data to obtain signal power spectrum density data;

[0083] In the embodiment of the present invention, the acquired low-loss conversion data is preprocessed, including filtering to remove noise and signal compensation, and the power spectrum density of the preprocessed data is calculated using Fourier transform or autocorrelation function method, and the signal is converted from the time domain to the frequency domain to obtain the energy distribution of the signal at different frequencies. Key information such as the main frequency components and power peaks are extracted from the calculated power spectrum density diagram.

[0084] Step S32: Calculate the box dimension of the signal power spectrum density data to generate the power spectrum fractal dimension;

[0085] In an embodiment of the present invention, the parameters used to calculate the box dimension are determined, including the box count and the box size, and the power spectral density data is divided into a corresponding number of small boxes according to the box count and the box size. Each small box can be a square, a cube or other shapes. For each small box, the coverage of the data points in the power spectral density data that fall into the box is calculated, which is obtained by calculating the ratio of the number of data points to the volume of the box. The box dimension is calculated using the coverage. The box dimension is the slope obtained by linear regression of the coverage of boxes of different sizes. The linear relationship between the box count function and the box size is used to fit the data, and the slope is the box dimension. The slope obtained according to the calculation is the power spectrum fractal dimension.

[0086] Step S33: performing independent signal power spectrum peak analysis on the low-loss conversion data to obtain a signal power spectrum peak;

[0087] In the embodiment of the present invention, the low-loss conversion data is preprocessed, including removing noise, filtering, and normalization operations, dividing the data into multiple frames, each frame containing a certain number of sampling points, applying Fourier transform (fast Fourier transform, FFT) to the data of each frame to convert the signal to the frequency domain, performing peak detection on the power spectrum density data of each frame, identifying the peak point in the power spectrum, and achieving this by finding the local maximum. Feature extraction is performed on the detected peak, including peak position, peak amplitude, and peak width, and the peak features detected in each frame are integrated to obtain the signal power spectrum peak of the low-loss conversion data.

[0088] Step S34: calculating the peak-bandwidth correlation characteristic index of the signal power spectrum peak to obtain the peak-bandwidth characteristic index;

[0089] In the embodiment of the present invention, the size of the correlation window used to calculate the peak-bandwidth characteristic index is determined. For each power spectrum peak, the correlation window is applied to the spectrum range near it. Within the correlation window, the frequency position of the peak and the amplitude of the peak are determined. The bandwidth characteristics are extracted by calculating the spectrum width around the peak, including the half-peak width, -3dB bandwidth or other indicators to measure the spectrum width. The peak characteristics and bandwidth characteristics are combined to calculate the peak-bandwidth characteristic index, and all peak-bandwidth characteristic indices are integrated to obtain the peak-bandwidth characteristic index of the signal.

[0090] Step S35: labeling the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label.

[0091] In an embodiment of the present invention, a label definition rule is formulated based on the value range of the power spectrum fractal dimension and the peak-bandwidth characteristic index, and the fractal dimensions and characteristic indices in different ranges are respectively marked as different RF signal categories. According to the label definition rule, the fractal dimension and characteristic index of each low-loss conversion data are compared with the label definition to determine the RF signal category to which it belongs, and each low-loss conversion data is associated with the corresponding classification label to obtain a label for the RF signal.

[0092] Preferably, step S4 comprises the following steps:

[0093] Step S41: when the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency analysis is performed on the strong RF signal tag to obtain signal frequency characteristic data;

[0094] Step S42: capturing the change pattern of the signal frequency feature data to generate a strong tag frequency change pattern;

[0095] Step S43: performing phase analysis on the strong radio frequency signal tag to obtain signal phase data; performing phase correlation analysis on the signal phase data to obtain signal phase correlation data;

[0096] Step S44: performing strong signal identification on the strong radio frequency signal tag based on the strong tag frequency change pattern and signal phase correlation data to obtain strong signal specific data.

[0097] The benefit of the present invention is that by setting the thresholds of the power spectrum fractal dimension and the peak-bandwidth characteristic index, signals with high complexity and spectrum characteristics can be marked as strong radio frequency signal tags. This helps to screen out signals with significant characteristics and focus on key radio frequency signals. Frequency analysis of strong radio frequency signal tags can extract the frequency characteristic data of the signal. This helps to understand the frequency distribution, frequency components and frequency variation law of the signal, and provides a basis for subsequent signal processing and analysis. Capturing the change pattern of the signal frequency characteristic data can reveal the frequency variation law of the signal. This helps to identify the pattern, periodic changes and trends of the signal, and provides an effective basis for subsequent signal analysis and prediction. By performing phase analysis and phase correlation analysis on strong radio frequency signal tags, the phase data and phase correlation data of the signal can be extracted. This helps to understand the phase characteristics, phase variation law and phase relationship between different frequency components of the signal. Based on the strong tag frequency variation pattern and signal phase correlation data, strong signal identification can be performed on the strong radio frequency signal tag. This helps to identify strong signals with important information and high energy, and further analyze and process these signals to obtain more specific data and features. By defining strong signal tags, performing frequency analysis, changing pattern capture, phase resolution and phase correlation analysis, and strong signal identification, a series of beneficial steps are provided for identifying RF data signals. These steps help extract key features from complex RF signals, identify important signals, and obtain specific data and feature information.

[0098] As an example of the present invention, in this example, step S4 includes:

[0099] Step S41: when the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency analysis is performed on the strong RF signal tag to obtain signal frequency characteristic data;

[0100] In an embodiment of the present invention, the power spectrum of each RF signal is analyzed, and its fractal dimension is calculated using the box counting method. The preset fractal dimension threshold is set to 1.5, and the peak-bandwidth characteristic index of the power spectrum of each RF signal is calculated. The peak in the signal spectrum is found, and the ratio of the peak to the corresponding bandwidth is calculated as the characteristic index. The preset characteristic index threshold is set to 0.6. For RF signals with a fractal dimension greater than 1.5 and a characteristic index greater than 0.6, they are defined as strong RF signal labels, and frequency analysis is performed on the strong RF signal labels to extract signal frequency characteristic data. For example, Fourier transform or wavelet transform can be used to perform spectral analysis on the signal, and the main frequency components or frequency band information in the spectrum can be extracted.

[0101] Step S42: capturing the change pattern of the signal frequency feature data to generate a strong tag frequency change pattern;

[0102] In an embodiment of the present invention, the frequency feature data is preprocessed, such as removing noise and smoothing data. A filter and an average sliding window method are used for preprocessing, and a time series analysis method is applied to capture the change pattern of the preprocessed frequency feature data. For example, cluster analysis, peak detection, and volatility analysis methods can be used. Based on the captured change pattern, a strong label frequency change pattern is generated. Statistical features and waveform analysis methods can be used to characterize the pattern of frequency change.

[0103] Step S43: performing phase analysis on the strong radio frequency signal tag to obtain signal phase data; performing phase correlation analysis on the signal phase data to obtain signal phase correlation data;

[0104] In an embodiment of the present invention, a strong radio frequency signal tag is phase parsed, and a parsing method is used to extract phase information from the signal. For example, the phase data of the signal can be obtained using a Hilbert transform or a phase-locked loop method, and a phase correlation analysis can be performed on the phase data to explore the correlation of the signal phase. Phase correlation analysis can be performed using an autocorrelation function, a cross-correlation function, or a phase difference statistical method, and the phase correlation data of the signal can be obtained based on the results of the phase correlation analysis. For example, the average value, standard deviation, and correlation coefficient index of the phase difference can be obtained to describe the phase correlation of the signal.

[0105] Step S44: performing strong signal identification on the strong radio frequency signal tag based on the strong tag frequency change pattern and signal phase correlation data to obtain strong signal specific data.

[0106] In an embodiment of the present invention, strong signal identification is performed according to the frequency change pattern of the strong radio frequency signal tag. By analyzing the frequency change trend of the strong tag, such as the increase or decrease of the detection frequency and the periodic change, it is determined whether a strong signal exists. For example, the frequency change rate is calculated using a sliding window or a differential method. If the frequency change rate exceeds a preset threshold, it is identified as a strong signal. The strong radio frequency signal tag is further identified by using the signal phase correlation data. By analyzing the phase relationship between the strong tags, such as the stability of the phase difference and the phase offset, it is determined whether a strong signal exists. For example, the phase correlation between different tags is calculated using a correlation coefficient or a phase difference criterion. If the phase correlation exceeds a preset threshold, it is confirmed as a strong signal. According to the identification result of the strong signal, the specific data of the strong signal is extracted. For example, specific information such as the amplitude, frequency, and phase of the strong signal can be extracted from the corresponding radio frequency data to obtain detailed characteristics and data of the strong signal.

[0107] Preferably, step S5 comprises the following steps:

[0108] Step S51: when the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and statistical feature extraction is performed on the weak RF signal tag to obtain weak tag statistical feature data;

[0109] Step S52: performing feature matching on the weak tag statistical feature data according to the preset historical radio frequency data signal to obtain historical feature matching radio frequency data;

[0110] Step S53: using historical feature matching radio frequency data to simulate and enhance the weak radio frequency signal tag to generate a simulated enhanced tag;

[0111] Step S54: performing fuzzy value estimation on the weak RF signal tag based on the simulated enhanced tag to obtain fuzzy data of the weak tag;

[0112] Step S55: Use the fuzzy data of the weak tag to assist in identifying the weak RF signal tag to obtain specific data of the weak signal.

[0113] The benefit of the present invention is that the radio frequency signal is classified by the power spectrum fractal dimension and the peak-bandwidth characteristic index, and is defined as a weak radio frequency signal tag. This can effectively distinguish weak signals from other types of signals, and provide a clear target for subsequent processing. Statistical feature extraction is performed on the weak radio frequency signal tag. This helps to extract meaningful information from the weak signal, and the statistical feature can provide a summary and description of the signal characteristics, which is convenient for subsequent processing and analysis. The weak tag statistical feature data is feature matched using the preset historical radio frequency data signal to obtain the historical feature matching radio frequency data. This can find data with similar features to the weak signal by comparing with the historical data, providing a reference and basis for subsequent processing. The weak radio frequency signal tag is simulated enhanced and fuzzy value estimated using the historical feature matching radio frequency data and the simulated enhanced tag. This can improve the recognizability and credibility of the weak signal and enhance the understanding and analysis capabilities of the weak signal. The weak radio frequency signal tag is auxiliary identified using the weak tag fuzzy data to obtain the specific data of the weak signal. This can further improve the accuracy and completeness of the recognition of the weak signal through the analysis and reasoning of the fuzzy data. Through different processing and analysis methods, useful information can be extracted from weak RF signals, signal understanding capabilities can be enhanced, and specific data of weak signals can be identified and obtained. This is of great significance for the processing and application of RF data signals.

[0114] In an embodiment of the present invention, when the power spectrum fractal dimension is less than a preset fractal dimension threshold (for example, set to 2.5), and the peak-bandwidth characteristic index is less than a preset characteristic index threshold (for example, set to 0.6), the radio frequency signal tag is defined as a weak radio frequency signal tag, and statistical feature extraction is performed on the weak radio frequency signal tag to obtain weak tag statistical feature data. For example, for spectrum data, its fractal dimension is calculated and compared with a preset threshold, and the peak-bandwidth characteristic index is calculated and compared with the preset threshold, and the weak tag statistical feature data is feature matched according to the preset historical radio frequency data signal. For example, the weak tag statistical feature data is compared with the corresponding features in the historical data, and a similarity measurement method (such as correlation coefficient, Euclidean distance) is used to perform matching calculation to obtain historical feature matching radio frequency data. The weak radio frequency signal tag is simulated enhanced using the historical feature matching radio frequency data to generate a simulated enhanced tag. For example, similar features in the historical feature matching radio frequency data are used to generate simulated enhanced tag data for the original weak radio frequency signal tag through interpolation, curve fitting or generative model methods. Fuzzy value estimation is performed on the weak radio frequency signal tag based on the simulated enhanced tag to obtain weak tag fuzzy data. For example, using fuzzy logic methods, the features of the simulated enhanced tags are matched with a set of preset fuzzy sets, and the fuzzy membership is calculated to obtain the fuzzy data of the weak tags. The weak tag fuzzy data is used to assist in identifying the weak RF signal tag to obtain the specific data of the weak signal. For example, based on the membership distribution of the weak tag fuzzy data, the fuzzy reasoning method is used to infer to determine the value or range of the specific data of the weak signal.

[0115] Preferably, step S53 includes the following steps:

[0116] Step S531: performing signal label analysis on the historical feature matching radio frequency data to obtain a historical similar label set;

[0117] Step S532: constructing a simulation model for the historical similar label set to generate a label simulation model;

[0118] Step S533: Use the tag simulation model to perform complete tag simulation on the weak RF signal tag to generate a simulated enhanced tag.

[0119] The benefit of the present invention is that the signal label analysis is performed on the historical feature matching radio frequency data to obtain the historical similar label set. By analyzing the label of the historical feature matching radio frequency data, the label data with similar characteristics to the weak radio frequency signal can be identified. This helps to further understand and classify the weak signal and provides basic data for building a label simulation model. The simulation model is constructed for the historical similar label set to generate a label simulation model. By constructing a simulation model for the historical similar label set, a generation model for simulated label data can be established. This model can capture the characteristics and laws of the historical similar label data and provide accuracy and reliability guarantees for subsequent label simulation. The weak radio frequency signal label is completely simulated by using the label simulation model to generate a simulated enhanced label. By applying the label simulation model, the weak radio frequency signal label can be completely simulated to generate a simulated enhanced label with similar characteristics and distribution to the historical similar label data. This helps to enhance the recognizability and understanding of weak signals and provide richer and more complete label data as the basis for subsequent processing and analysis. By analyzing and modeling the historical feature matching radio frequency data and simulating the weak signal label using the simulation model, the understanding and recognition ability of weak signals can be improved, and the quality and availability of signal data can be enhanced. This is of great significance for the processing and application of RF data signals.

[0120] In an embodiment of the present invention, a signal label analysis is performed on the radio frequency data matched with historical features to obtain a historical similar label set. For example, based on the label information of the radio frequency data matched with historical features, a cluster analysis or similarity measurement is performed on it, and similar labels are classified into the same category to obtain a historical similar label set. The K-means clustering algorithm is used, and the number of clusters is set to a preset value, or a similarity calculation method based on a distance metric, such as cosine similarity or Euclidean distance, is used to construct a simulation model for the historical similar label set to generate a label simulation model. For example, based on the statistical characteristics of the historical similar label set, a probability distribution model (such as a Gaussian mixture model) or a machine learning algorithm (such as a generative adversarial network) is used to construct a label simulation model. A Gaussian mixture model is selected as the simulation model, and model parameters (mean, covariance matrix, weight) are set or a generative adversarial network architecture is selected. The label simulation model is used to perform complete label simulation on the weak radio frequency signal label to generate a simulated enhanced label. For example, the weak radio frequency signal label is input into the label simulation model, and a simulated enhanced label with similar characteristics and distribution to the historical similar label is generated by the model. Use sampling methods from generative models, such as sampling from Gaussian mixture models or generating data through generative adversarial networks.

[0121] In this specification, a system for identifying a radio frequency data signal is provided, which is used to perform the method for identifying a radio frequency data signal as described above. The system for identifying a radio frequency data signal includes:

[0122] The spectrum fractal module is used to obtain radio frequency data; perform spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; perform fractal clustering on the scale spectrum graph to obtain spectrum fractal data;

[0123] The progressive conversion module is used to reconstruct the spectrum of the scale spectrum diagram to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data;

[0124] The label definition module is used to perform spectral fractal analysis on the low-loss conversion data to generate the power spectrum fractal dimension; perform peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain the peak-bandwidth characteristic index; perform label definition on the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain the radio frequency signal label;

[0125] A strong signal identification module is used to define the RF signal tag as a strong RF signal tag when the power spectrum fractal dimension is greater than a preset fractal dimension threshold and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, and to perform frequency characteristic pattern analysis on the strong RF signal tag to generate a strong tag frequency change pattern; to perform phase characteristic correlation analysis on the strong RF signal tag to obtain signal phase correlation data; to perform strong signal identification on the strong RF signal tag based on the strong tag frequency change pattern and the signal phase correlation data to obtain specific data of the strong signal;

[0126] The weak signal identification module is used to define the RF signal tag as a weak RF signal tag when the power spectrum fractal dimension is less than a preset fractal dimension threshold and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, and simulate and complete the weak RF signal tag based on the preset historical RF data signal to generate a simulated enhanced tag; and perform auxiliary identification on the weak RF signal tag based on the simulated enhanced tag to obtain specific data of the weak signal.

[0127] The benefit of the present invention lies in that the original radio frequency data can be effectively converted into a more structured and characteristic scale spectrum diagram and spectrum fractal data through spectrum hierarchical analysis and fractal clustering, so that the signal characteristics are clearer and distinguishable. By performing spectrum optimization and progressive conversion on the scale sub-signal and spectrum fractal data, it is helpful to reduce the loss in the signal transmission process and improve the quality and reliability of the signal. According to the power spectrum fractal dimension and the peak-bandwidth characteristic index, the low-loss conversion data is labeled and defined, which can effectively distinguish strong radio frequency signals from weak radio frequency signals, and perform corresponding identification and classification. For strong signals, frequency characteristic pattern analysis and phase characteristic correlation analysis can be further performed to obtain specific data of strong signals, which helps to deeply understand and utilize the information of strong signals; while for weak signals, by simulating enhanced labels and auxiliary identification, the recognition ability can be enhanced and the integrity and accuracy of data acquisition can be improved. Different processing is performed for strong signals and weak signals respectively. For strong signals, by frequency characteristic pattern analysis and phase characteristic correlation analysis, the characteristics of the signal can be further understood and specific data information can be obtained therefrom. For weak signals, by simulating enhanced labels and auxiliary identification, the recognition ability of the signal can be improved, so as to obtain more complete data information. Therefore, the present invention effectively processes radio frequency data, extracts useful information therein, and realizes accurate identification and classification of signals by combining signal processing, feature extraction and classification recognition technologies, thereby providing important support for the optimization and performance improvement of radio frequency communication systems.

[0128] The present invention also provides a device for identifying radio frequency data signals, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, for use in the method for identifying radio frequency data signals as described above.

[0129] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points of view, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the same elements of the application documents are included in the present invention.

[0130] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for identifying a radio frequency data signal, characterized in that: The following steps are involved: Step S1: Acquire radio frequency data; perform spectrum layering analysis on the radio frequency data to generate a scale spectrum diagram; Perform fractal clustering on the scale spectrum graph to obtain spectrum fractal data; Step S2: reconstruct the spectrum of the scaled spectrum graph to obtain sub-signals of each scale; Perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data; wherein step S2 includes the following steps: Step S21: performing inverse Fourier transform on the scaled spectrum to obtain reconstructed time-domain sub-signal components; performing multi-layer decomposition on the reconstructed time-domain sub-signal components to obtain sub-signals of various scales; Step S22: extracting fractal features from the frequency spectrum fractal data to obtain frequency spectrum fractal feature data; Step S23: Adaptively reconstruct the RF data using the sub-signals of each scale and the spectrum fractal feature data to generate an optimized reconstructed spectrum; wherein step S23 includes the following steps: Step S231: performing high-dimensional spectrum analysis on each scale sub-signal to obtain multi-modal sub-spectrum feature data; Step S232: performing generalized fractal dimension calculation on the multimodal sub-spectrum feature data to obtain the generalized fractal dimension; Step S233: dynamically transform and model the generalized fractal dimension and the frequency spectrum fractal feature data to generate a dynamic transformation model; Step S234: using a dynamic conversion model to perform nonlinear dynamic transformation simulation on the radio frequency data to obtain an optimized reconstructed spectrum; Step S24: performing low-loss conversion on the radio frequency data according to the optimized reconstructed spectrum to obtain low-loss conversion data; Step S3: performing spectral fractal analysis on the low-loss conversion data to generate a power spectrum fractal dimension; performing peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain a peak-bandwidth characteristic index; defining a label for the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label; Step S4: When the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency characteristic pattern analysis is performed on the strong RF signal tag to generate a strong tag frequency change pattern; a phase characteristic correlation analysis is performed on the strong RF signal tag to obtain signal phase correlation data; based on the strong tag frequency change pattern and the signal phase correlation data, a strong signal identification is performed on the strong RF signal tag to obtain strong signal specific data; Step S5: When the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and the weak RF signal tag is simulated and completed based on the preset historical RF data signal to generate a simulated enhanced tag; the weak RF signal tag is assisted in identification based on the simulated enhanced tag to obtain specific data of the weak signal.

2. The method for identifying a radio frequency data signal according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: acquiring the highest frequency component of the collected signal; collecting radio frequency signals based on the highest frequency component of the collected signal to obtain radio frequency data; Step S12: performing spectrum filtering on the radio frequency data to obtain a filtered radio frequency data signal; performing multi-frequency analysis on the filtered radio frequency data signal to generate radio frequency signal components of various scales; Step S13: Performing rapid spectrum analysis on each scale radio frequency signal component to obtain a scale spectrum diagram; Step S14: performing cluster analysis on the scale spectrum graph to obtain scale spectrum clusters; Step S15: Estimating the fractal dimension of the scaled frequency spectrum clusters to obtain frequency spectrum fractal data.

3. The method for identifying a radio frequency data signal according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing power spectrum density analysis on the low-loss conversion data to obtain signal power spectrum density data; Step S32: Calculate the box dimension of the signal power spectrum density data to generate the power spectrum fractal dimension; Step S33: performing independent signal power spectrum peak analysis on the low-loss conversion data to obtain a signal power spectrum peak; Step S34: calculating the peak-bandwidth correlation characteristic index of the signal power spectrum peak to obtain the peak-bandwidth characteristic index; Step S35: labeling the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain a radio frequency signal label.

4. The method for identifying a radio frequency data signal according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: when the power spectrum fractal dimension is greater than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, the RF signal tag is defined as a strong RF signal tag, and a frequency analysis is performed on the strong RF signal tag to obtain signal frequency characteristic data; Step S42: capturing the change pattern of the signal frequency feature data to generate a strong tag frequency change pattern; Step S43: performing phase analysis on the strong radio frequency signal tag to obtain signal phase data; performing phase correlation analysis on the signal phase data to obtain signal phase correlation data; Step S44: performing strong signal identification on the strong radio frequency signal tag based on the strong tag frequency change pattern and signal phase correlation data to obtain strong signal specific data.

5. The method for identifying a radio frequency data signal according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: when the power spectrum fractal dimension is less than a preset fractal dimension threshold, and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, the RF signal tag is defined as a weak RF signal tag, and statistical feature extraction is performed on the weak RF signal tag to obtain weak tag statistical feature data; Step S52: performing feature matching on the weak tag statistical feature data according to the preset historical radio frequency data signal to obtain historical feature matching radio frequency data; Step S53: using historical feature matching radio frequency data to simulate and enhance the weak radio frequency signal tag to generate a simulated enhanced tag; Step S54: performing fuzzy value estimation on the weak RF signal tag based on the simulated enhanced tag to obtain fuzzy data of the weak tag; Step S55: Use the fuzzy data of the weak tag to assist in identifying the weak RF signal tag to obtain specific data of the weak signal.

6. The method for identifying a radio frequency data signal according to claim 5, characterized in that: Step S53 includes the following steps: Step S531: performing signal label analysis on the historical feature matching radio frequency data to obtain a historical similar label set; Step S532: constructing a simulation model for the historical similar label set to generate a label simulation model; Step S533: Use the tag simulation model to perform complete tag simulation on the weak RF signal tag to generate a simulated enhanced tag.

7. A system for identifying radio frequency data signals, characterized in that: For executing the method for identifying a radio frequency data signal as claimed in claim 1, the system for identifying a radio frequency data signal comprises: The spectrum fractal module is used to obtain radio frequency data; perform spectrum layering analysis on the radio frequency data to generate a scale spectrum graph; perform fractal clustering on the scale spectrum graph to obtain spectrum fractal data; The progressive conversion module is used to reconstruct the spectrum of the scale spectrum diagram to obtain sub-signals of each scale; perform spectrum optimization progressive conversion on each scale sub-signal and spectrum fractal data to obtain low-loss conversion data; The label definition module is used to perform spectral fractal analysis on the low-loss conversion data to generate the power spectrum fractal dimension; perform peak-bandwidth characteristic index analysis on the low-loss conversion data to obtain the peak-bandwidth characteristic index; perform label definition on the low-loss conversion data based on the power spectrum fractal dimension and the peak-bandwidth characteristic index to obtain the radio frequency signal label; A strong signal identification module is used to define the RF signal tag as a strong RF signal tag when the power spectrum fractal dimension is greater than a preset fractal dimension threshold and the peak-bandwidth characteristic index is greater than a preset characteristic index threshold, and to perform frequency characteristic pattern analysis on the strong RF signal tag to generate a strong tag frequency change pattern; to perform phase characteristic correlation analysis on the strong RF signal tag to obtain signal phase correlation data; to perform strong signal identification on the strong RF signal tag based on the strong tag frequency change pattern and the signal phase correlation data to obtain specific data of the strong signal; The weak signal identification module is used to define the RF signal tag as a weak RF signal tag when the power spectrum fractal dimension is less than a preset fractal dimension threshold and the peak-bandwidth characteristic index is less than a preset characteristic index threshold, and simulate and complete the weak RF signal tag based on the preset historical RF data signal to generate a simulated enhanced tag; and perform auxiliary identification on the weak RF signal tag based on the simulated enhanced tag to obtain specific data of the weak signal.

8. A device for identifying a radio frequency data signal, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, and is used to execute the method for identifying a radio frequency data signal as claimed in claim 1.

Images