Signal identification method based on satellite radio signal characteristic data
By augmenting satellite radio signals with data and extracting various features, combined with machine learning models, the problems of sample scarcity and insufficient robustness in satellite radio signal identification were solved, achieving high-precision signal identification and improving the recognition rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- PINGHU SPACE PERCEPTION LAB TECH CO LTD
- Filing Date
- 2025-12-12
- Publication Date
- 2026-04-17
AI Technical Summary
Existing radio signal identification technologies suffer from problems such as scarce satellite radio signal samples, insufficient feature dimensions, insufficient robustness, and low identification accuracy, making it difficult to meet the real-time processing requirements in complex electromagnetic environments.
By performing data augmentation on the original satellite radio signals, various signal features are extracted, a feature dataset is constructed, and feature extraction and classification are performed based on machine learning models. These include data augmentation methods such as random window processing, downsampling, time-series reversal, and random discarding, as well as various signal features such as the maximum value of the spectral density of the zero-center normalized instantaneous amplitude. Support vector machines, decision trees, Naive Bayes, and k-nearest neighbor algorithms are used for identification.
It effectively alleviates the model overfitting problem caused by the scarcity of satellite radio signal samples, improves the model's generalization ability, enriches the effective feature categories of radio signals, achieves high-precision signal recognition, and breaks through the recognition rate bottleneck of traditional methods.
Smart Images
Figure CN121881065A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of radio signal processing technology, and in particular to a signal identification method based on satellite radio signal feature data. Background Technology
[0002] With the rapid development of satellite communication technology and the continuous increase in the number of satellites in orbit, the radio spectrum environment is becoming increasingly complex. Accurate identification of satellite radio signals has become a key technical aspect of spectrum monitoring, electromagnetic space security management, and interference source localization. Satellite radio signal characteristics are the "fingerprints" of radio signals, serving as a crucial basis for distinguishing different signal sources, modulation methods, and transmitting equipment. Satellite radio signal characteristics refer to quantitative indicators extracted from the satellite radio signal itself or its observable parameters, capable of characterizing its intrinsic attributes. Due to the diverse types, wide frequency band distribution, open transmission channels, and large dynamic range of signal-to-noise ratio of satellite radio signals, their characteristics are easily affected by noise interference and channel fading. Traditional signal identification methods mainly rely on manual experience for feature extraction and pattern judgment, which is not only inefficient but also difficult to meet the real-time processing needs of massive data and complex electromagnetic environments.
[0003] Currently, radio signal feature extraction technology has yielded relatively abundant research results. Popular key features of radio signals include time-domain and frequency-domain signal features such as signal envelope kurtosis, cyclic spectrum, number of phase pulses, wavelet transform, higher-order moments, signal statistics, characteristic spectrum, signal constellation, maximum power spectral density, and instantaneous phase standard deviation. The effectiveness of these features for radio signal analysis and identification tasks varies.
[0004] However, several shortcomings remain in the feature extraction process. Existing feature extraction methods are fragmented, lacking a systematic feature selection and evaluation mechanism tailored to the characteristics of satellite radio signals. The effectiveness of various features for recognition tasks varies significantly but lacks quantitative standards, making it difficult to form optimal feature combinations. Satellite signal samples are typically scarce; directly using limited samples to train models easily leads to overfitting, while traditional methods lack effective data augmentation strategies, limiting model generalization ability. Existing features lack robustness under low signal-to-noise ratio conditions, making it difficult to adapt to common channel fading and noise interference during satellite signal transmission. Furthermore, the recognition accuracy of single features or simple feature combinations faces performance bottlenecks; the recognition rate of traditional machine learning methods often struggles to break through these limitations, and in practical applications, the recognition rate typically fails to meet engineering requirements.
[0005] In summary, existing radio signal identification technologies have limitations in feature system construction, data augmentation, and robust design. Therefore, there is an urgent need for a signal identification scheme capable of constructing satellite radio signal feature datasets, fully mining the essential features of signals, and combining machine learning methods to overcome the performance bottlenecks of existing technologies and improve the accuracy and reliability of satellite radio signal identification. Summary of the Invention
[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies, specifically the scarcity of satellite radio signal samples and insufficient feature dimensions. Specifically, it provides a signal identification method based on satellite radio signal feature data, as detailed below: 1) In a first aspect, the present invention provides a signal identification method based on satellite radio signal feature data, the specific technical solution of which is as follows: S1, perform data augmentation processing on the collected raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; S2, extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; S3. Based on the feature dataset, train a machine learning model. The trained machine learning model is used to extract and classify features of the satellite radio signals to be identified, and the signal identification result is obtained.
[0007] The beneficial effects of the signal identification method based on satellite radio signal feature data provided by this invention are as follows: By augmenting the original signals to expand the sample size, the overfitting problem caused by the scarcity of satellite radio signal samples is effectively alleviated, thus improving the model's generalization ability. Multi-class and multi-domain features are extracted to construct a feature dataset, significantly enriching the effective feature categories of radio signals and providing comprehensive fingerprint information for signal identification. A machine learning model is trained based on this feature dataset to achieve high-precision identification of satellite radio signal sources, overcoming the bottleneck of insufficient recognition rate in traditional methods.
[0008] Based on the above solution, the present invention can be further improved as follows.
[0009] Furthermore, the data augmentation process includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
[0010] Furthermore, the various signal features include: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
[0011] Furthermore, before training the machine learning model based on the feature dataset, the process further includes: Each signal feature in the feature dataset is preprocessed to eliminate the influence of dimensions in the feature dataset.
[0012] 2) In a second aspect, the present invention also provides a signal identification system based on satellite radio signal feature data, the specific technical solution of which includes: a signal augmentation module, a feature construction module, and a classification and identification module; The signal augmentation module is used to perform data augmentation processing on the acquired raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; The feature construction module is used to extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; The classification and recognition module is used to train a machine learning model based on the feature dataset, and to extract and classify the features of the satellite radio signal to be identified through the trained machine learning model to obtain the signal recognition result.
[0013] Based on the above solution, the present invention can be further improved as follows.
[0014] Furthermore, the data augmentation process includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
[0015] Furthermore, the various signal features include: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
[0016] Furthermore, before training the machine learning model based on the feature dataset, the process further includes: Each signal feature in the feature dataset is preprocessed to eliminate the influence of dimensions in the feature dataset.
[0017] 3) In a third aspect, the present invention also provides a computer device, the computer device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the computer device to implement any of the above methods.
[0018] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to enable a computer to perform any of the above methods.
[0019] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description
[0020] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the steps of a signal identification method based on satellite radio signal feature data according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the processing flow of a signal identification method based on satellite radio signal feature data according to an embodiment of the present invention; Figure 3 This is a structural block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
[0022] like Figure 1 As shown in the figure, a signal identification method based on satellite radio signal feature data according to an embodiment of the present invention includes the following steps: S1, perform data augmentation processing on the collected raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; S2, extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; S3 trains a machine learning model based on the feature dataset. The trained machine learning model is used to extract features and classify the satellite radio signals to be identified, and the signal identification results are obtained.
[0023] The beneficial effects of the signal identification method based on satellite radio signal feature data provided by this invention are as follows: By augmenting the original signals to expand the sample size, the overfitting problem caused by the scarcity of satellite radio signal samples is effectively alleviated, thus improving the model's generalization ability. Multi-class and multi-domain features are extracted to construct a feature dataset, significantly enriching the effective feature categories of radio signals and providing comprehensive fingerprint information for signal identification. A machine learning model is trained based on this feature dataset to achieve high-precision identification of satellite radio signal sources, overcoming the bottleneck of insufficient recognition rate in traditional methods.
[0024] It should be noted that, for ease of understanding, the technical terms used in this solution will be explained one by one, and will not be repeated hereafter: Raw satellite radio signals: refer to unprocessed radio frequency electromagnetic wave signals transmitted or relayed by satellites, containing basic communication parameters such as modulation information and carrier frequency.
[0025] Data augmentation processing refers to a method of increasing the data sample size by transforming the original satellite radio signals. In this scheme, data augmentation processing includes at least one of the following: random window processing, downsampling processing, time-series reversal processing, and random discarding processing.
[0026] Augmented signal sample set: refers to the set of signal samples that are many times larger than the original satellite radio signals after data augmentation processing, providing a sufficient data foundation for subsequent feature extraction.
[0027] Signal sample: refers to a single discrete digital signal sequence.
[0028] Signal characteristics: refers to the quantitative indicators extracted from signal samples that can characterize their intrinsic attributes. In this scheme, ten signal characteristics are specifically included.
[0029] Feature dataset: refers to a data set consisting of various signal features of all signal samples.
[0030] Machine learning model: refers to a classifier built using data-driven automatic learning algorithms. In this scheme, the learning algorithm can employ support vector machines, decision trees, Naive Bayes, and the k-nearest neighbor method. Among these: Support Vector Machine (SVM): A classic supervised learning algorithm widely used in classification, regression, and outlier detection tasks. Its core idea is to find an optimal hyperplane in the feature space to perform classification, maximizing the margin.
[0031] Naive Bayes algorithm: a probability and statistics-based classification method that uses Bayes' theorem and the feature conditional independence assumption to predict the category of a sample.
[0032] Decision tree algorithm: It classifies or predicts data step by step by setting a series of rules, and is suitable for classification and regression tasks.
[0033] k-Nearest Neighbors Algorithm: This algorithm predicts the class of an unknown sample by the majority vote of its k nearest neighbors. It is typically suitable for classification or regression tasks with small-scale, low-dimensional data.
[0034] Satellite radio signals to be identified: These refer to unknown satellite radio signals that need to be classified and identified in practical applications. The processing method is the same as that of the samples after signal augmentation during the training phase.
[0035] Signal recognition result: refers to the signal category label or classification decision output by the machine learning model. Processing the signal using the recognition method in this solution significantly improves the recognition rate.
[0036] Random window processing refers to the operation of randomly selecting a starting position in a radio signal to extract a fixed-length segment of signal. In this scheme, for a signal x(k), a window length L is set, and a random index s is used, where s takes values ranging from 1 to N-L+1, to ensure that the extracted window signal falls completely within the original satellite radio signal sequence. The extracted window signal (i.e., the value after random window processing of x(k)) is... .
[0037] It should be noted that the objects of random window processing, downsampling processing, time reversal processing, and random discarding processing can be the original satellite radio signals or radio signals obtained after other processing. This will not be elaborated further below, and the object used in data augmentation processing will be represented by signal x(k).
[0038] On the other hand, it should be noted that in this solution, the radio signal participating in processing / calculation is represented by x(k), and the value of its subscript k depends on the actual processing / calculation. For example, for a signal sample, k = 1, 2, …, N, where N is the signal length. For the signal length in data augmentation processing, it can be L (i.e., the window length L). In other words, when the object of operation in data augmentation processing is the original satellite radio signal, the signal length is N. When the object of operation in data augmentation processing is the radio signal processed by a random window, the signal length is L (i.e., after being intercepted by the window length L, the total length of the radio signal is L). This will not be elaborated further below, and only N is used to represent the signal length for the setting of the value of k.
[0039] Downsampling processing: It refers to the operation of extracting signal samples from the radio signal at a fixed interval to reduce the sampling rate. In this solution, a sampling factor d (i.e., the number of samples at a fixed interval) is set, and the downsampled signal is , where .
[0040] Time reversal processing: It refers to the operation of reversing the signal sequence along the time axis. In this solution, the reversed signal is , and most of the statistical characteristics of the reversed signal remain unchanged.
[0041] Random discarding processing: It refers to the operation of randomly setting some samples of the signal to zero with a certain probability. In this solution, a discarding probability p is set, where 0 < p < 1, and a mask m is generated: ; the signal after random discarding is .
[0042] The maximum value of the spectral density of the zero-centered normalized instantaneous amplitude : It refers to the maximum spectral density value obtained after performing a fast Fourier transform (i.e., FFT) on the normalized centered instantaneous amplitude sequence. The calculation method is as follows: For the radio signal x(k), k = 1, 2, …, N, calculate its instantaneous amplitude sequence a(k) as: a(k) = |x(k)|; calculate the mean value E(a) of the instantaneous amplitude sequence a(k) as: E(a) = ; calculate the zero-centered normalized instantaneous amplitude sequence as: ; perform a fast Fourier transform on the zero-centered normalized instantaneous amplitude sequence to obtain the frequency-domain representation FFT , calculate the squared modulus of the transformation result to obtain the power spectral density, find the maximum value among all frequency points and divide it by the signal length N to obtain the maximum value of the spectral density of the zero-centered normalized instantaneous amplitude as: , where f() is a function.
[0043] Standard deviation of the absolute value of the zero-center normalized instantaneous amplitude This refers to the statistical standard deviation of the absolute values of the normalized instantaneous amplitude sequence. It is calculated as follows: using zero-centered normalized instantaneous amplitude sequence... Calculate the zero-center normalized instantaneous amplitude sequence The average of the squares (i.e., the first mean): ; Calculate the zero-center normalized instantaneous amplitude sequence The average of the absolute values (i.e., the second mean): Calculate the square of the second mean. The first difference is obtained by subtracting the square of the second mean from the first mean. Taking the square root of the first difference yields the standard deviation of the zero-center normalized absolute value of the instantaneous amplitude. .
[0044] Standard deviation of instantaneous amplitude of non-weak signal segment with zero center normalization This refers to the standard deviation of the normalized instantaneous amplitude after removing non-weak signal segments. Non-weak signals are determined by a threshold. The calculation method is as follows: Set a threshold t for determining non-weak signals. a Traverse the instantaneous amplitude sequence a(k) to find all values greater than the threshold t. a The sampling point locations are used to normalize the instantaneous amplitude sequence with zero centers corresponding to these locations. The sampling points constitute a non-weak signal segment. The number of sampling points in the non-weak signal segment is denoted as the length C of the non-weak signal segment. The square mean (i.e., the third mean) of all zero-center normalized instantaneous amplitude values within the non-weak signal segment is calculated. ; Calculate the average of all zero-center normalized instantaneous amplitude values within the non-weak signal segment (i.e., the fourth mean): ; Calculate the square of the fourth mean The second difference is obtained by subtracting the square of the fourth mean from the third mean. The square root of the second difference is then used to obtain the standard deviation of the instantaneous amplitude of the non-weak signal segment with zero center normalization. .
[0045] Standard deviation of instantaneous phase nonlinear component in zero-center non-weak signal segment This refers to the standard deviation of the instantaneous phase nonlinear component after removing non-weak signal segments. The calculation method is as follows: Calculate the instantaneous phase sequence φ(k) for the signal sample x(k), and remove the linear component from the instantaneous phase sequence φ(k) to obtain the nonlinear component. Using the location determination results of the non-weak signal segment, the instantaneous phase nonlinear component is extracted. For all sampling points within the non-weak signal segment, calculate the squared average (i.e., the fifth mean) of these sampling points: Calculate the average of these sample points (i.e., the sixth mean): Calculate the square of the sixth mean. The third difference is obtained by subtracting the square of the sixth mean from the fifth mean. The square root of the third difference is then used to obtain the standard deviation of the instantaneous phase nonlinear component of the zero-center non-weak signal segment. ;in, It is the nonlinear component of the instantaneous phase of a zero-center non-weak signal.
[0046] Standard deviation of the absolute value of the zero-center normalized instantaneous frequency in the non-weak signal segment This refers to the standard deviation of the normalized absolute value of the instantaneous frequency after removing non-weak signal segments. The calculation method is as follows: calculate the instantaneous frequency sequence f(k) for the signal sample x(k), and calculate the zero-center normalized instantaneous frequency sequence f0. cn (k), removing the zero-center component from the instantaneous frequency sequence f(k) yields the instantaneous frequency f of the non-weak signal. NL (k) Using the result of the non-weak signal segment location determination, extract the instantaneous frequency f of the non-weak signal. NL (k) For all sampling points within the non-weak signal segment, calculate the squared average of the absolute values of these sampling points (i.e., the seventh mean): Calculate the average of the absolute values of these sample points (i.e., the eighth mean): Calculate the square of the eighth mean. The fourth difference is obtained by subtracting the square of the eighth mean from the seventh mean. The square root of the fourth difference is then used to obtain the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment. .
[0047] Spectral symmetry Power spectral density (PSD) is a measure of the symmetry of a signal's power spectral density about its center frequency. It is calculated as follows: Power spectral density is estimated from a signal sample x(k) to obtain a power spectral density sequence P(f), where f is a frequency variable, and the center frequency f of the signal is determined. c and the maximum frequency range f max Set the frequency offset Δf = 1 / f s f s The sampling frequency is [f]; in the interval to the left of the center frequency [f] c Summing the power spectral density values within [-Δf,0] yields the left spectrum sum: In the interval to the right of the center frequency [f c +Δf,f max The summation of the power spectral density values within the inner quadrant yields the right spectrum: The total spectrum is obtained by summing the power spectral density values over the entire frequency range: Calculate spectral symmetry .
[0048] Compactness of normalized zero-center instantaneous amplitude This refers to the ratio of the fourth moment to the square of the second moment of the normalized instantaneous amplitude sequence. The calculation method is as follows: using the zero-center normalized instantaneous amplitude sequence a... cn (k), calculate the zero-center normalized instantaneous amplitude sequence a cn The fourth moment mean of (k) (i.e., the ninth mean): ; Calculate the zero-center normalized instantaneous amplitude sequence a cn The second moment mean (i.e., the tenth mean) of (k): Calculate the square of the tenth mean, and divide the ninth mean by the square of the tenth mean to obtain the compactness of the normalized zero-center instantaneous amplitude. .
[0049] Compactness of zero-center normalized instantaneous frequency This refers to the ratio of the square of the fourth moment to the square of the second moment in the normalized instantaneous frequency sequence. The calculation method is as follows: Calculate the instantaneous frequency sequence f(k) for the signal sample x(k), and calculate the zero-center normalized instantaneous frequency sequence f... cn (k), calculate the zero-center normalized instantaneous frequency sequence f cn The fourth moment mean (i.e., the eleventh mean) of (k): ; Calculate the zero-center normalized instantaneous frequency sequence f cn The second moment mean of (k) (i.e., the twelfth mean): Calculate the square of the twelfth mean, and divide the eleventh mean by the square of the twelfth mean to obtain the compactness of the zero-centered normalized instantaneous frequency. .
[0050] higher-order features of the envelope This refers to the characteristic parameters calculated based on the higher-order statistical moments of the real part of the signal, which are obtained by combining the fourth-order moment m4 and the second-order moment m2. The calculation method is as follows: Extract the real part sequence y(k) = real(x(k)) from the signal sample x(k), k = 1, 2, ..., N, and calculate the fourth-order moment of the real part sequence y(k). ,Right now ; Calculate the second moment of the real part sequence y(k) Right now ; Calculate the second moment The square value of the fourth moment Subtract twice the second moment The square value gives the numerator Divide the molecule by four times the second moment The squared value yields higher-order features of the envelope. .
[0051] Envelope feature parameters This refers to the ratio of the variance of the signal envelope to the square of the mean, reflecting the envelope's fluctuation characteristics. The calculation method is as follows: For the signal sample x(k), calculate the instantaneous amplitude sequence a(k) = |x(k)|, and calculate the mean of the instantaneous amplitude sequence a(k) μ = E[a(k)], i.e. ; Calculate the variance of the instantaneous amplitude sequence a(k) =E[(a(k)-μ) 2 ],Right now ; Calculate the square of the mean Variance Divided by the square of the mean Obtain envelope feature parameters .
[0052] Preprocessing refers to the operations performed on each signal feature in the feature dataset to improve model performance. In this scheme, standard deviation standardization is used for data preprocessing to eliminate the influence of units. The standard deviation standardization method is as follows: ;in, Let y be the mean of the signal characteristic. Let y be the standard deviation of the signal characteristic y. These are the signal characteristics obtained after processing.
[0053] Dimension: refers to the unit of measurement of a physical quantity and its dimensions. Eliminating the influence of dimensions can achieve the unification of characteristic scales.
[0054] In another embodiment of this solution, S1 is specifically implemented as follows: S101, random window processing is performed on the acquired raw satellite radio signal during data augmentation. The raw satellite radio signal is represented as x(k), k=1,2,...,N, where N represents the signal length of the raw satellite radio signal. The window length L is set to a fixed positive integer, and L is less than N. The random index s is set to a random integer, and the value of s satisfies 0≤s≤NL. L consecutive sampling points from the (s+1)th sampling point to the (s+L)th sampling point are extracted from the raw satellite radio signal sequence to form the window signal, which is represented as... By changing the value of the random index s, multiple different window signals can be obtained, thus expanding a single raw satellite radio signal into multiple signal samples with different time segments.
[0055] S102, downsampling is performed on the window signal. The sampling factor d is set to a positive integer greater than 1. The downsampled signal is obtained by extracting one sample every d samples from the window signal. The downsampled signal is represented as follows: (k)= (k), where k = 1, 1 + d,..., 1 + nd, and n represents the maximum index value of the downsampled signal, which is determined by the floor function Determined by . The downsampling operation reduces the data volume and increases the sample diversity without changing the essential characteristics of the signal.
[0056] S103. Perform time reversal processing on the window signal. Reverse the window signal sequence completely along the time axis to obtain the reversed signal, which is denoted as (k) = (k), where k = L, L - 1,..., 1. The time reversal processing keeps the statistical characteristics of the signal unchanged while generating new signal samples.
[0057] S104. Perform random discard processing on the window signal. Set the discard probability p as a real number and satisfy 0 < p < 1. Generate the mask m(k), and the mask m(k) follows Bernoulli(1 - p). The mask m(k) is a binary sequence and each element independently takes the value 1 with probability 1 - p and takes the value 0 with probability p, which is used to control which sampling points are retained and which are set to zero. The signal after random discard is denoted as (k) = m(k) (k), where k = 1, 2,..., L, and Denotes the point - by - point multiplication operation. The random discard processing randomly sets some sampling points in the window signal to zero, thus simulating the signal loss situation and enhancing the model robustness.
[0058] In another embodiment of this solution, the specific implementation manner of S2 is as follows: Extract multiple signal features for each signal sample in the augmented signal sample set. The maximum value of the spectral density of the zero - centered normalized instantaneous amplitude, the standard deviation of the absolute value of the zero - centered normalized instantaneous amplitude, the standard deviation of the zero - centered non - weak signal segment instantaneous amplitude, the standard deviation of the non - linear component of the zero - centered non - weak signal segment instantaneous phase, the standard deviation of the absolute value of the zero - centered non - weak signal segment instantaneous frequency, spectral symmetry, the compactness of the zero - centered normalized instantaneous amplitude, the compactness of the zero - centered normalized instantaneous frequency, envelope high - order features, and envelope feature parameters can be extracted in sequence, obtaining ten corresponding eigenvalue. Arrange the ten obtained eigenvalues in a fixed order to form a ten - dimensional feature vector, and the form of the feature vector is , , , , , , , , , The feature extraction calculation process is repeated for each signal sample in the augmented signal sample set to obtain a ten-dimensional feature vector corresponding to each signal sample. All feature vectors are then organized into a matrix according to the order of the signal samples to obtain the feature dataset of the satellite radio signal. Each row of the feature dataset corresponds to the ten-dimensional feature vector of one signal sample, and each column of the feature dataset corresponds to the value of a signal feature across all signal samples. It should be noted that the calculation process of the feature values is as described in each feature extraction process above, and will not be repeated here.
[0059] In another embodiment of this solution, S3 is specifically implemented as follows: The feature dataset is divided into training and testing sets. The training set can be used to train models using Support Vector Machines, Decision Trees, Naive Bayes, and k-Nearest Neighbors algorithms, respectively. The recognition accuracy is evaluated using the testing set. Models with a recognition rate of over 94% are selected, and their parameters are fixed to form a trained machine learning model. Satellite radio signals to be identified are collected, data augmentation is performed, and ten signal features are extracted to form feature vectors to be identified. The mean and standard deviation parameters saved during the training phase are used to standardize the feature vectors to be identified. The standardized feature vectors are then input into the trained machine learning model, which outputs class labels as the signal recognition results for signal source identification.
[0060] Furthermore, data augmentation processing includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
[0061] Furthermore, multiple signal characteristics, including: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
[0062] Furthermore, before training the machine learning model based on the feature dataset, the following steps are also included: Each signal feature in the feature dataset undergoes preprocessing to eliminate the influence of dimensions. Specifically, standard deviation standardization is used to preprocess the data and eliminate the influence of dimensions. The standard deviation standardization method is as follows: ;in, Let y be the mean of the signal characteristic. Let y be the standard deviation of the signal characteristic y. The signal features are obtained after processing. Performing S3 on the standardized feature dataset obtained after preprocessing can eliminate the influence of dimensions and improve the efficiency of satellite radio signal identification.
[0063] Example 1, Figure 2 This is a schematic diagram of the processing flow of an embodiment of the present invention; as shown below. Figure 2 As shown, the raw radio signals are augmented using methods including random windowing, downsampling, time-series reversal, and random discarding to expand the originally sparse data sample. Ten signal features are extracted from the augmented signal to construct a feature dataset; these features are crucial for signal classification and recognition. Since the extracted feature values vary, standard deviation normalization is used for data preprocessing to eliminate the influence of dimensions. Subsequently, support vector machines, decision trees, Naive Bayes, and k-nearest neighbors are used for recognition, and their recognition performance is analyzed. Using the method described in this scheme, the recognition rate for satellite radio signal features can reach over 94%.
[0064] In the above embodiments, although the steps are numbered S1, S2, etc., they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation, and these situations are also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.
[0065] The present invention also provides a signal recognition system based on satellite radio signal feature data, the specific technical solution of which includes: a signal augmentation module, a feature construction module, and a classification and recognition module; The signal augmentation module is used to perform data augmentation processing on the acquired raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; The feature construction module is used to extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; The classification and recognition module is used to train a machine learning model based on the feature dataset. The trained machine learning model extracts and classifies the features of the satellite radio signals to be identified, and obtains the signal recognition results.
[0066] Based on the above solution, the present invention can be further improved as follows.
[0067] Furthermore, data augmentation processing includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
[0068] Furthermore, multiple signal characteristics, including: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
[0069] Furthermore, before training the machine learning model based on the feature dataset, the following steps are also included: Each signal feature in the feature dataset is preprocessed to eliminate the influence of dimensions in the feature dataset.
[0070] It should be noted that the beneficial effects of the signal identification system based on satellite radio signal feature data provided in the above embodiments are the same as those of the signal identification method based on satellite radio signal feature data described above, and will not be repeated here. Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, and will not be repeated here.
[0071] like Figure 3 As shown, an embodiment of the present invention provides a computer device 300, which includes a processor 320 coupled to a memory 310. The memory 310 stores at least one computer program 330, which is loaded and executed by the processor 320 to enable the computer device 300 to implement any of the above-described methods. Specifically: The computer device 300 can vary considerably due to differences in configuration or performance. It may include one or more processors 320 (Central Processing Units, CPUs) and one or more memories 310. The one or more memories 310 store at least one computer program 330, which is loaded and executed by the one or more processors 320 to enable the computer device 300 to implement the signal identification method based on satellite radio signal characteristic data provided in the above embodiments. Of course, the computer device 300 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The computer device 300 may also include other components for implementing device functions, which will not be elaborated here.
[0072] An embodiment of the present invention provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to enable a computer to implement any of the above-described methods.
[0073] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.
[0074] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform any of the above-described signal identification methods based on satellite radio signal characteristic data.
[0075] It should be noted that the terms "first," "second," etc., used in the specification of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown in the figures or description.
[0076] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.
[0077] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.
[0078] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A signal identification method based on satellite radio signal characteristic data, characterized in that, include: S1, perform data augmentation processing on the collected raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; S2, extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; S3. Based on the feature dataset, train a machine learning model. The trained machine learning model is used to extract and classify features of the satellite radio signals to be identified, and the signal identification result is obtained.
2. The signal identification method based on satellite radio signal feature data according to claim 1, characterized in that, The data augmentation process includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
3. The signal identification method based on satellite radio signal feature data according to claim 1, characterized in that, The various signal features include: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
4. The signal identification method based on satellite radio signal feature data according to claim 1, characterized in that, Before training the machine learning model based on the feature dataset, the method further includes: Each signal feature in the feature dataset is preprocessed to eliminate the influence of dimensions in the feature dataset.
5. A signal identification system based on satellite radio signal characteristic data, characterized in that, include: Signal augmentation module, feature construction module, and classification and recognition module; The signal augmentation module is used to perform data augmentation processing on the acquired raw satellite radio signals to obtain an augmented signal sample set of satellite radio signals; The feature construction module is used to extract multiple signal features for each signal sample in the augmented signal sample set, and construct a feature dataset of satellite radio signals based on the multiple signal features of all signal samples; The classification and recognition module is used to train a machine learning model based on the feature dataset, and to extract and classify the features of the satellite radio signal to be identified through the trained machine learning model to obtain the signal recognition result.
6. A signal identification system based on satellite radio signal characteristic data according to claim 5, characterized in that, The data augmentation process includes: At least one of random window processing, downsampling processing, time-reversal processing, and random discarding processing.
7. A signal identification system based on satellite radio signal characteristic data according to claim 5, characterized in that, The various signal features include: The maximum value of the spectral density of the zero-center normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-center normalized instantaneous amplitude, the standard deviation of the zero-center normalized instantaneous amplitude of the non-weak signal segment, the standard deviation of the instantaneous phase nonlinear component of the non-weak signal segment, the standard deviation of the absolute value of the zero-center normalized instantaneous frequency of the non-weak signal segment, spectral symmetry, compactness of the normalized zero-center instantaneous amplitude, compactness of the zero-center normalized instantaneous frequency, higher-order envelope features, and at least two of the following envelope feature parameters.
8. A signal identification system based on satellite radio signal characteristic data according to claim 5, characterized in that, Before training the machine learning model based on the feature dataset, the method further includes: Each signal feature in the feature dataset is preprocessed to eliminate the influence of dimensions in the feature dataset.
9. A computer device, characterized in that, The computer device includes a processor coupled to a memory storing at least one computer program, which is loaded and executed by the processor to enable the computer device to implement a signal identification method based on satellite radio signal feature data as described in any one of claims 1 to 4.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to enable the computer to implement a signal identification method based on satellite radio signal feature data as described in any one of claims 1 to 4.