An efficient interference identification method for satellite navigation systems

Abstract

This invention designs an efficient interference identification method for satellite navigation systems, relating to the fields of communication technology and satellite navigation and positioning. First, signals are received through a navigation satellite receiver, and the received signals are preprocessed. Then, features are extracted from the preprocessed received signals to establish a feature parameter dataset. An improved Sparrow Algorithm-based BP neural network model classifier, SSA-BP, is established to classify the feature parameter dataset, resulting in the identification result. This method can accurately identify the type of interference present in the received signal, facilitating rapid implementation of corresponding measures and improving anti-interference capabilities. It is applicable to satellite navigation receiving systems and devices.

Description

Technical Field

[0001] This invention relates to the fields of communication technology and satellite navigation and positioning, and specifically to an efficient interference identification method for satellite navigation systems. Background Technology

[0002] Satellite navigation systems provide users with all-weather, all-time, high-precision positioning, navigation, and timing services. Because satellite platforms operate in open environments, they are susceptible to various types of interference, affecting the accuracy of navigation and positioning. Therefore, how to quickly identify interference signals is a crucial part of research in navigation interference signal identification technology. Generally, interference identification techniques design suitable classifiers based on feature extraction to identify interference signals. A classifier can be seen as a transformation process that maps the received signal from a feature parameter space to a decision space. Under a certain discrimination criterion, it classifies each unknown sample in the feature vector space into different discrimination types, thereby achieving the classification and identification of interference signals. However, current identification techniques mostly rely on supervised learning methods from machine learning, such as decision trees, SVM, and KNN. While these methods can effectively identify various interference signals under certain conditions, they still suffer from slow identification speed and insufficient accuracy. Therefore, designing efficient interference identification methods is crucial and key to improving the anti-interference capability of navigation systems. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention presents a highly efficient interference identification method for satellite navigation systems.

[0004] An efficient interference identification method for satellite navigation systems includes the following steps:

[0005] Step 1: The navigation satellite receiver receives signals, which include navigation signals, interference signals, and noise; among which the interference signals are single-tone interference, multi-tone interference, pulse interference, frequency sweep interference, BPSK interference, and noise interference; noise interference refers to noise amplitude modulation interference, broadband noise interference, and narrowband noise interference.

[0006] Step 2: Preprocess the received signal. The preprocessing process includes normalization and noise reduction. The normalization is to normalize the power of the signal to [-1,1] and use wavelet transform for filtering and noise reduction.

[0007] Step 3: Perform feature extraction on the preprocessed received signal S to effectively identify interference signals; extract the following eight feature parameters from S:

[0008] (1) Temporal kurtosis factor k u ;

[0009]

[0010] Temporal kurtosis factor k u It indicates the smoothness of the signal waveform and is used to describe the distribution of the signal amplitude; thus, it can be used to distinguish pulse interference from other types of interference.

[0011] (2) Time-domain waveform factor W;

[0012]

[0013]

[0014]

[0015] The time-domain waveform factor W is the effective value S of the received signal. rms With the rectified average value S arv The ratio; where N represents the number of sampling points; S i Where i = 1, 2, ..., N represents the i-th received signal; this feature is used to distinguish multi-tone interference from other interferences;

[0016] (3) Time-domain peak factor C;

[0017]

[0018] χ is the difference between the peaks and troughs of the received signal waveform, also known as the peak-to-peak value; narrowband noise interference is distinguished by the time-domain peak factor C.

[0019] (4) Time-domain pulse factor I;

[0020]

[0021] The time-domain pulse factor I is the peak value S of the received signal. peak The average value of the rectified average, i.e., the absolute value, S arv The ratio; used to distinguish pulse interference from other interference;

[0022] (5) Time margin factor L;

[0023] The time margin factor L is the ratio of the peak value to the root square amplitude of the received signal. Similar to the time peak factor C, the root square amplitude and the root mean square value, i.e. the effective value of the received signal, are corresponding. The root mean square is the arithmetic square root of the average of the sum of squares of the signal, and the root square amplitude is the square of the average of the arithmetic square roots. This feature is used to distinguish frequency sweep interference from other interference.

[0024] (6) Carrier factor coefficient G;

[0025]

[0026] Performing a Fast Fourier Transform on the received signal S yields a discrete frequency domain signal X[λ]. Sorting these signals according to their amplitude, we obtain X[λ1], X[λ2], ..., X[λ]. NFFT The ratio of the maximum spectral value X[λ1] to the adjacent larger value X[λ2] is defined as the carrier factor coefficient G; this is used to distinguish single-tone interference from other interferences.

[0027] (7) Normalized average flatness coefficient F se ;

[0028]

[0029] The preprocessed received signal S is then subjected to a windowed moving average to obtain the flat portion of the normalized signal amplitude. Then use the normalized signal amplitude, i.e., R p (i) Subtracting its flat portion yields the normalized impulse component; taking the arithmetic square root of this impulse component is defined as the normalized average flatness coefficient, denoted by F. se This characteristic is used to distinguish noise interference from other types of interference.

[0030] (8) The partial mean of the time-domain moments α;

[0031]

[0032] Where μ is the mean of the received signal S and σ is the standard deviation of the received signal S; this feature is used to distinguish noise amplitude modulation interference from other interferences.

[0033] Step 4: Based on the feature parameters extracted in Step 3, establish a feature parameter dataset for the received signal; the dataset format is as follows:

[0034] The dataset contains nine dimensions in total. The first eight dimensions represent eight extracted feature parameters, where the first dimension corresponds to the temporal kurtosis factor k. u The second dimension corresponds to the time-domain waveform factor W, the third dimension to the time-domain peak factor C, the fourth dimension to the time-domain impulse factor I, the fifth dimension to the time-domain margin factor L, the sixth dimension to the carrier factor coefficient G, and the seventh dimension to the normalized average flatness coefficient F. se The eighth dimension corresponds to the partial mean of the time-domain moments α; the ninth dimension corresponds to the type of interference signal.

[0035] Step 5: Considering that traditional BP neural networks have problems such as slow learning speed and easy getting trapped in local optima, we establish a sparrow algorithm to optimize the BP neural network model classifier SSA-BP by improving the sparrow population initialization method and changing the identity formula of the discoverer and joiner, and classify the feature parameter dataset established in Step 4.

[0036] First, divide the feature parameter dataset generated in step 4 into a training set and a test set in a 7:3 ratio. Input the divided training set data and test set data and normalize them. Then, use the training set data to train the SSA-BP network model. Stop training when the maximum number of iterations or the minimum error is reached. Then, use the test set data to verify and classify the trained SSA-BP network model, thereby realizing the classification and identification of various interference signals.

[0037] The specific recognition process using a classifier based on an improved sparrow search and optimized BP neural network model is as follows:

[0038] Step 5.1: Establish the BP neural network structure, that is, determine the number of nodes in the input layer, hidden layer, and output layer of the BP neural network;

[0039] Step 5.2: Initialize the BP neural network;

[0040] Step 5.3: Initialize the population and objective function; establish the improved sparrow algorithm optimizer, i.e., set H in the sparrow algorithm. max , n, Θ, Ω, Ξ parameter values; where H max : Maximum number of iterations; n: Population size, i.e., number of sparrows; Θ: Number of discoverers; Ω: Number of sparrows sensing danger; Ξ: Safety value;

[0041] Step 5.4: Establish the improved chaotic mapping formula y n+1 =2(sin(2arccosy) n )) 2 -1, where y represents the sparrow population location and n is the number of sparrows; initialize the sparrow population location and calculate the fitness of each individual sparrow;

[0042] F = min(MSE) TrainingSet +MSE TestingSet )

[0043] In the formula, TrainingSet and TestingSet are the training set and the test set, respectively; then the fitness function F is the minimum mean square error of the entire training set and the test set.

[0044] Step 5.5: Establish an improved discoverer location update formula to update the discoverer's location;

[0045]

[0046] In the formula, exp() represents an exponential function with base e, randn(0,1) represents generating a random number that follows a normal distribution with an expected value of 0 and a standard deviation of 1, t represents the current iteration number, and Y... i,jLet represent the position information of the i-th sparrow in the j-th dimension, where a is a random number ∈ [0,1], and R1 (R1∈[0,1]) and Ξ (Ξ∈[0.5,1]) represent the warning value and the safety value, respectively.

[0047] Step 5.6: Establish an improved follower position update formula to update the joiner position;

[0048]

[0049]

[0050] Z is calculated as follows:

[0051]

[0052] Where Γ(1+ξ)=(1+ξ-1)!, ξ∈[0,2], r1, r2 follow a uniform distribution and are random numbers in the range [0,1]. It is the best location currently occupied by the discoverer, Y worst Let Q be the position of the sparrow with the lowest fitness, and let Q be a random number following a standard normal distribution. Let represent the Kronecker product, Levy(d) represent the Levy distribution, and d represent the dimension to be optimized.

[0053] Step 5.7: Update the position of the early warning agent using the early warning agent formula from the original Sparrow Algorithm;

[0054] Step 5.8: Determine if the termination condition is met; that is, whether the maximum number of iterations or the minimum error required by the target has been reached. If it is met, output the optimal individual parameters and assign them to the optimal weights and threshold parameters of the BP neural network; otherwise, return to step 5.4.

[0055] Step 5.9: Input the normalized training set data into the BP neural network for training;

[0056] Step 5.10: If the maximum number of iterations is reached or the minimum error required by the target is achieved, the training ends and the network model SSA-BP is output; otherwise, return to step 5.9.

[0057] Step 5.11: Test the output network model SSA-BP with normalized test set data and output labels to determine which type of interference it belongs to, thereby achieving BP neural network classification and obtaining the recognition result.

[0058] Beneficial technical effects of the present invention:

[0059] Leveraging the advantages of BP neural network classifiers—fast classification speed and high recognition accuracy—and overcoming their susceptibility to initial weight thresholds, this invention avoids getting trapped in local optima. It incorporates the fitness parameter of the sparrow algorithm, introduces chaotic mapping and the Lévy flight mechanism, and draws inspiration from the bird flocking algorithm to construct an efficient interference identification method for satellite navigation systems. This method can be applied to various application areas requiring interference identification, such as military, navigation communication, and speech recognition. When the interference signal-to-noise ratio is above 4dB, the method achieves an accuracy rate of over 98% for each type of interference signal, accurately identifying the type of interference present in the received signal. This method facilitates rapid implementation of corresponding measures, improving anti-interference capabilities. This invention is applicable to satellite navigation receiving systems and devices. Attached Figure Description

[0060] Figure 1 A flowchart of an interference identification method for a satellite navigation system according to an embodiment of the present invention;

[0061] Figure 2 The recognition accuracy of the SSA-BP neural network model classifier under a fixed signal-to-noise ratio in this embodiment of the invention;

[0062] Figure 3 A schematic diagram comparing the overall interference identification curves of an embodiment of the present invention. Detailed Implementation

[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments;

[0064] An efficient interference identification method for satellite navigation systems, as shown in the appendix. Figure 1 As shown, the specific steps include:

[0065] Step 1: The navigation satellite receiver receives signals, which include navigation signals, interference signals, and noise; among which the interference signals are single-tone interference, multi-tone interference, pulse interference, frequency sweep interference, BPSK interference, and noise interference; noise interference refers to noise amplitude modulation interference, broadband noise interference, and narrowband noise interference.

[0066] Step 2: Preprocess the received signal to make it comparable after feature extraction; the preprocessing process includes normalization and noise reduction; normalization normalizes the power of the signal to [-1,1], and wavelet transform is used for filtering and noise reduction;

[0067] Step 3: Perform feature extraction on the preprocessed received signal S to effectively identify interference signals; extract the following eight feature parameters from S:

[0068] (1) Temporal kurtosis factor k u ;

[0069]

[0070] Temporal kurtosis factor k u It indicates the smoothness of the signal waveform and is used to describe the distribution of signal amplitude; when the kurtosis is less than or equal to 3, the distribution curve will be "flatter", and when it is greater than 3, the distribution curve will be "steeper"; this is used to distinguish pulse interference from other interference.

[0071] (2) Time-domain waveform factor W;

[0072]

[0073]

[0074]

[0075] The time-domain waveform factor W is the effective value S of the received signal. rms With the rectified average value S arv The ratio; where N represents the number of sampling points; S i Where i = 1, 2, ..., N represents the i-th received signal; this feature is used to distinguish multi-tone interference from other interferences;

[0076] (3) Time-domain peak factor C;

[0077]

[0078] χ is the difference between the peaks and troughs of the received signal waveform, also known as the peak-to-peak value; narrowband noise interference is distinguished by the time-domain peak factor C.

[0079] (4) Time-domain pulse factor I;

[0080]

[0081] The time-domain pulse factor I is the peak value S of the received signal. peak The average value of the rectified average, i.e., the absolute value, S arv The ratio; used to distinguish pulse interference from other interference;

[0082] (5) Time margin factor L;

[0083] The time margin factor L is the ratio of the peak value to the root square amplitude of the received signal. Similar to the time peak factor C, the root square amplitude and the root mean square value, i.e. the effective value of the received signal, are corresponding. The root mean square is the arithmetic square root of the average of the sum of squares of the signal, and the root square amplitude is the square of the average of the arithmetic square roots. This feature is used to distinguish frequency sweep interference from other interference.

[0084] (6) Carrier factor coefficient G;

[0085]

[0086] Performing a Fast Fourier Transform on the received signal S yields a discrete frequency domain signal X[λ]. Sorting these signals according to their amplitude, we obtain X[λ1], X[λ2], ..., X[λ]. NFFT The ratio of the maximum spectral value X[λ1] to the adjacent larger value X[λ2] is defined as the carrier factor coefficient G; this is used to distinguish single-tone interference from other interferences.

[0087] (7) Normalized average flatness coefficient F se ;

[0088]

[0089] The preprocessed received signal S is then subjected to a windowed moving average to obtain the flat portion of the normalized signal amplitude. Then use the normalized signal amplitude, i.e., R p (i) Subtracting its flat portion yields the normalized impulse component; taking the arithmetic square root of this impulse component is defined as the normalized average flatness coefficient, denoted by F. se This characteristic is used to distinguish noise interference from other types of interference.

[0090] (8) The partial mean of the time-domain moments α;

[0091]

[0092] Where μ is the mean of the received signal S and σ is the standard deviation of the received signal S; this feature is used to distinguish noise amplitude modulation interference from other interferences.

[0093] Step 4: Based on the feature parameters extracted in Step 3, establish a feature parameter dataset for the received signal; the dataset format is as follows:

[0094] First dimension Second dimension Third dimension Fourth dimension Fifth dimension Sixth dimension Seventh Dimension Eighth dimension Ninth Dimension <![CDATA[k u ]]> W C I L G <![CDATA[F se ]]> α Interference Category

[0095] The dataset contains a total of 9 dimensions. The first 8 dimensions represent the eight extracted feature parameters, where the first dimension corresponds to the temporal kurtosis factor k. u The second dimension corresponds to the time-domain waveform factor W, the third dimension to the time-domain peak factor C, the fourth dimension to the time-domain impulse factor I, the fifth dimension to the time-domain margin factor L, the sixth dimension to the carrier factor coefficient G, and the seventh dimension to the normalized average flatness coefficient F. se The eighth dimension corresponds to the partial mean of the time-domain moments α; the ninth dimension corresponds to the type of interference signal.

[0096] Step 5: Considering that traditional BP neural networks have problems such as slow learning speed and easy getting trapped in local optima, we establish a sparrow algorithm to optimize the BP neural network model classifier SSA-BP by improving the sparrow population initialization method and changing the identity formula of the discoverer and joiner, and classify the feature parameter dataset established in Step 4.

[0097] First, divide the feature parameter dataset generated in step 4 into a training set and a test set in a 7:3 ratio. Input the divided training set data and test set data and normalize them. Then, use the training set data to train the SSA-BP network model. Stop training when the maximum number of iterations or the minimum error is reached. Then, use the test set data to verify and classify the trained SSA-BP network model, thereby realizing the classification and identification of various interference signals.

[0098] The specific recognition process using a classifier based on an improved sparrow search and optimized BP neural network model is as follows:

[0099] Step 5.1: Establish the BP neural network structure, that is, determine the number of nodes in the input layer, hidden layer, and output layer of the BP neural network;

[0100] Step 5.2: Initialize the BP neural network;

[0101] Step 5.3: Initialize the population and objective function; establish the improved sparrow algorithm optimizer, i.e., set H in the sparrow algorithm. max , n, Θ, Ω, Ξ parameter values; where H max : Maximum number of iterations; n: Population size, i.e., number of sparrows; Θ: Number of discoverers; Ω: Number of sparrows sensing danger; Ξ: Safety value;

[0102] Step 5.4: Establish the improved chaotic mapping formula y n+1 =2(sin(2arccosy) n )) 2 -1, where y represents the sparrow population location and n is the number of sparrows; initialize the sparrow population location and calculate the fitness of each individual sparrow;

[0103] F = min(MSE) TrainingSet +MSE TestingSet )

[0104] In the formula, TrainingSet and TestingSet are the training set and the test set, respectively; then the fitness function F is the minimum mean square error of the entire training set and the test set.

[0105] Step 5.5: Establish an improved discoverer location update formula to update the discoverer's location;

[0106]

[0107] In the formula, exp() represents an exponential function with base e, randn(0,1) represents generating a random number that follows a normal distribution with an expected value of 0 and a standard deviation of 1, t represents the current iteration number, and Y... i,j Let represent the position information of the i-th sparrow in the j-th dimension, where a is a random number ∈ [0,1], and R1 (R1∈[0,1]) and Ξ (Ξ∈[0.5,1]) represent the warning value and the safety value, respectively.

[0108] Step 5.6: Establish an improved follower position update formula to update the joiner position;

[0109]

[0110]

[0111] Z is calculated as follows:

[0112]

[0113] Where Γ(1+ξ)=(1+ξ-1)!, ξ∈[0,2], r1, r2 follow a uniform distribution and are random numbers in the range [0,1]. It is the best location currently occupied by the discoverer, Y worst Let Q be the position of the sparrow with the lowest fitness, and let Q be a random number following a standard normal distribution. Let represent the Kronecker product, Levy(d) represent the Levy distribution, and d represent the dimension to be optimized.

[0114] Step 5.7: Update the position of the early warning agent using the early warning agent formula from the original Sparrow Algorithm;

[0115] Step 5.8: Determine if the termination condition is met; that is, whether the maximum number of iterations or the minimum error required by the target has been reached. If it is met, output the optimal individual parameters and assign them to the optimal weights and threshold parameters of the BP neural network; otherwise, return to step 5.4.

[0116] Step 5.9: Input the normalized training set data into the BP neural network for training;

[0117] Step 5.10: If the maximum number of iterations is reached or the minimum error required by the target is achieved, the training ends and the network model SSA-BP is output; otherwise, return to step 5.9.

[0118] Step 5.11: Test the output network model SSA-BP with normalized test set data and output labels to determine which type of interference it belongs to, thereby achieving BP neural network classification and obtaining the recognition result.

[0119] The test results verifying the above method;

[0120] The simulation parameters are set as follows:

[0121]

[0122] The signal-to-noise ratio (JSNR) is set to a range of -5 to 15 dB, with a JSNR interval of 1 dB. Single-tone interference (STO) and multi-tone interference (MTI) frequencies are randomly generated within the signal bandwidth, with a center frequency of 1561.098 MHz. STO interference includes one frequency point, and MTI includes five frequency points. Pulse interference is generated by rectangular pulses. The sweep frequency interference signal has an initial frequency of 1558 MHz and a cutoff frequency of 1562 MHz. The BPSK interference signal has a center frequency of 1561.098 MHz. Wideband noise interference, narrowband noise interference, and amplitude-modulated noise interference are all generated by Gaussian white noise, with wideband bandwidth accounting for more than 1 / 2 and narrowband bandwidth accounting for less than 1 / 5.

[0123] The specific parameter settings for the SSA-BP classifier are as follows:

[0124] The sparrow population size is set to 15, the maximum number of iterations is 15, and the upper and lower thresholds of the independent variables are 5 and -5, respectively. ST = 0.6 represents the safety value; PD = 0.7 represents the proportion of discoverers in the entire population, with the rest being followers; SD = 0.2 represents the proportion of sparrows that can sense danger in the entire population, i.e., the proportion of alarm sparrows in the entire population. The number of input layer nodes in the BP-network is set to 8, representing eight feature parameters; the number of hidden layer nodes is set to 12; and the number of output layer nodes is set to 8, representing eight types of interference signals. The network iterations are 500, the error resolution is 1e-3, and the learning rate is 0.001. The fitness function set in this paper is as follows:

[0125] F = min(MSE) TrainingSet +MSE TestingSet )

[0126] In the formula, TrainingSet and TestingSet are the training set and the test set, respectively. The fitness function is the minimum mean squared error of the entire training and test sets.

[0127] The simulation experimental environment is as follows:

[0128] The simulation software used was Matlab 2020a, the operating system was Microsoft Win10, and the hardware configuration was AMD Ryzen 7 4800H with Radeon Graphics 2.90GHz and 16GB of memory.

[0129] The simulation experiments were conducted on two different datasets:

[0130] The first type is a fixed signal-to-noise ratio (SNR) dataset. This means that the eight sample datasets used for training the model are generated under the same SNR conditions each time. Each SNR dataset contains 1200 signal interference features across 8 classes, with an average of 150 feature samples per class. Using a 7:3 ratio, 840 datasets are selected as training samples to train the network, and the remaining 360 datasets are used as test samples to evaluate the network model's classification and recognition efficiency.

[0131] The second type is a mixed interference signal-to-noise ratio (SNR) dataset, which consists of a mixture of all signal interference feature samples under different SNR conditions, totaling 25,200 data sets. Using a 7:3 ratio, 17,640 data sets were selected as training samples for network training, and the remaining 7,560 data sets were used as test samples to test the classification and recognition efficiency of the network model.

[0132] Under a fixed signal-to-noise ratio, the recognition accuracy of the SSA-BP neural network model classifier is as follows: Figure 2 As shown;

[0133] Under mixed signal-to-noise ratio conditions, the interference correct recognition rate of the SSA-BP neural network model classifier is shown in the table below; the statistical average recognition rate is approximately 75.7%.

[0134]

[0135] The training performance is significantly worse than that of neural network models under a fixed signal-to-noise ratio, indicating that the classification performance of the network model will decrease as the complexity of the samples increases.

[0136] according to Figure 3 The comparison of the overall interference recognition curves shown indicates that the SSA-BP network model classifier has improved interference recognition performance compared to the traditional BP network model classifier.

Claims

1. A highly efficient interference identification method for a satellite navigation system, characterized in that, Specifically, the following steps are included: Step 1: The navigation satellite receiver receives signals, which include navigation signals, interference signals, and noise; wherein the interference signals are single-tone interference, multi-tone interference, pulse interference, frequency sweep interference, BPSK interference, and noise interference. Noise interference refers to amplitude modulation noise interference, broadband noise interference, and narrowband noise interference; Step 2: Preprocess the received signal, including normalization and noise reduction. The normalization involves normalizing the power of the signal to [-1, 1] and then using wavelet transform for filtering and noise reduction. Step 3: Process the pre-processed received signal Feature extraction is performed to enable effective identification of interference signals; Step 4: Based on the feature parameters extracted in Step 3, establish a feature parameter dataset for the received signal; The dataset contains nine dimensions in total. The first eight dimensions represent eight extracted feature parameters, with the first dimension corresponding to the temporal kurtosis factor. The second dimension corresponds to the time-domain waveform factor. The third dimension corresponds to the time-domain peak factor. The fourth dimension corresponds to the time-domain pulse factor. The fifth dimension corresponds to the time-domain margin factor. The sixth dimension corresponds to the carrier factor coefficient. The seventh dimension corresponds to the normalized average flatness coefficient. The eighth dimension corresponds to the partial mean of the time-domain moments. The ninth dimension corresponds to the type of interference signal. Step 5: Considering that traditional BP neural networks have problems such as slow learning speed and easy getting trapped in local optima, we establish a sparrow algorithm to optimize the BP neural network model classifier SSA-BP by improving the sparrow population initialization method and changing the identity formula of the discoverer and joiner, and classify the feature parameter dataset established in Step 4. First, divide the feature parameter dataset generated in step 4 into a training set and a test set in a 7:3 ratio. Input the training set data and test set data after dividing them in the ratio and normalize them. Then, use the training set data to train the SSA-BP network model. Stop training when the maximum number of iterations or the minimum error is reached. Then, use the test set data to verify and classify the trained SSA-BP network model, thereby realizing the classification and identification of various interference signals. Step 5.1: Establish the BP neural network structure, that is, determine the number of nodes in the input layer, hidden layer, and output layer of the BP neural network; Step 5.2: Initialize the BP neural network; Step 5.3: Initialize the population and objective function; establish the improved sparrow algorithm optimizer, i.e., set the parameters in the sparrow algorithm... , , , , Parameter values; where Maximum number of iterations; Population size, i.e., the number of sparrows; Number of discoverers; The number of sparrows that sense danger; : Safety value; Step 5.4: Establish the improved chaotic mapping formula ,in Represents the location of the sparrow population. This represents the number of sparrows; initialize the sparrow population location and calculate the fitness of each individual sparrow. ； In the formula, TrainingSet and TestingSet are the training set and the test set, respectively; then the fitness function F is the minimum mean square error of the entire training set and the test set. Step 5.5: Establish an improved discoverer location update formula to update the discoverer's location; ； In the formula, exp() represents an exponential function with base e. This represents generating a random number that follows a normal distribution with an expected value of 0 and a standard deviation of 1. Represents the current iteration number. Indicates the first The sparrow in the first Location information in the dimension yes random numbers, and These represent the warning value and the safe value, respectively. Step 5.6: Establish an improved follower position update formula to update the joiner position; ； ； in, The calculation method is as follows: ； in, , , , It follows a uniform distribution and is a random number in the range [0,1]. This is the best location currently occupied by the discoverers. The position of the sparrow with the lowest fitness. For random numbers that follow a standard normal distribution, Representing Kronecker, Representing the Lévy distribution, Indicates the dimension that needs to be optimized; Step 5.7: Update the position of the early warning agent using the early warning agent formula from the original Sparrow Algorithm; Step 5.8: Determine if the termination condition is met; that is, whether the maximum number of iterations or the minimum error required by the target has been reached. If it is met, output the optimal individual parameters and assign them to the optimal weights and threshold parameters of the BP neural network; otherwise, return to step 5.

4. Step 5.9: Input the normalized training set data into the BP neural network for training; Step 5.10: If the maximum number of iterations is reached or the minimum error required by the target is achieved, the training ends and the network model SSA-BP is output; otherwise, return to step 5.

9. Step 5.11: Test the output network model SSA-BP with normalized test set data and output labels to determine which type of interference it belongs to, thereby achieving BP neural network classification and obtaining the recognition result.

2. The efficient interference identification method for a satellite navigation system according to claim 1, characterized in that, Step 3 Extract the following eight feature parameters respectively: (1) Temporal kurtosis factor ; ； Temporal kurtosis factor It indicates the smoothness of the signal waveform and is used to describe the distribution of the signal amplitude; thus, it can be used to distinguish pulse interference from other types of interference. (2) Time-domain waveform factor ; ； ； ； Time-domain wave factor It is the effective value of the received signal. With rectified average value The ratio; where, Indicates the number of sampling points; Indicates the received number One signal, among which This characteristic is used to distinguish multi-tone interference from other types of interference. (3) Time-domain peak factor ; ； It is the difference between the peaks and troughs of the received signal waveform, also known as the peak-to-peak value; measured by the time-domain peak factor. To distinguish narrowband noise interference; (4) Time-domain pulse factor ; ； Time-domain pulse factor It is the peak value of the received signal. The average of the rectified average value, i.e., the average of the absolute values. The ratio; used to distinguish pulse interference from other interference; (5) Time-domain margin factor ; Time margin factor It is the ratio of the peak value to the root square amplitude of the received signal; and the time-domain peak factor. Similarly, the root square amplitude and the root mean square value, which are the effective values ​​of the received signal, are corresponding. The root mean square is the arithmetic square root of the average of the sum of squares of the signal, and the root square amplitude is the square of the average of the arithmetic square roots. This feature is used to distinguish frequency sweep interference from other interference. (6) Carrier factor coefficient ; ； For the received signal Perform a Fast Fourier Transform to obtain the frequency domain discrete signal Sort by amplitude to obtain Its maximum spectral value Larger adjacent values The ratio is defined as the carrier factor coefficient. This is used to distinguish monotone interference from other types of interference. (7) Normalized average flatness coefficient ; ； For the pre-processed received signal By performing a windowed moving average, the flat portion of the normalized signal amplitude is obtained. Then use the normalized signal amplitude, i.e. Subtracting the flat portion yields the normalized impulse component; taking the arithmetic square root of this impulse component is defined as the normalized average flatness coefficient. This characteristic is used to distinguish noise interference from other types of interference. (8) Partial mean of time-domain moments ; ； in, To receive signals The mean, To receive signals The standard deviation; this characteristic is used to distinguish amplitude modulation noise interference from other interferences.

Images