Satellite navigation interference signal type identification method and system

Abstract

The invention provides a satellite navigation interference signal type identification method and system, and belongs to the field of satellite navigation. The method comprises the following steps: collecting a complete navigation signal from a satellite receiver, calculating carrier-to-noise ratio and pseudo-range residual error information, and judging whether a navigation interference signal exists in the complete navigation signal; if a navigation interference signal is detected, collecting intermediate frequency data in the complete navigation signal, performing spectral analysis, adding a sliding window in the intermediate frequency data, extracting frequency domain characteristics of the complete navigation signal based on the sliding window and the weight of the sliding window in combination with a gating circulation unit, and performing time domain analysis on the intermediate frequency data in the complete navigation signal to obtain a navigation interference signal; extracting time domain features of the complete navigation signal; and calculating the type of the navigation interference signal by combining the frequency domain characteristics of the complete navigation signal and the time domain characteristics of the complete navigation signal. According to the invention, the adaptability of satellite navigation interference signal type identification and the accuracy of the result are improved.

Description

Technical Field

[0001] The present invention belongs to the field of satellite navigation, and particularly relates to a method and system for identifying types of satellite navigation interference signals. Background Art

[0002] Satellite navigation refers to a technology that uses navigation satellites to perform navigation and positioning for ground, ocean, air, and space users. Common satellite navigation systems include the GPS navigation system, the Beidou navigation system, etc. With its advantages of high precision, all-weather, and global coverage, it has become an indispensable navigation means in many fields. When evaluating navigation capabilities, anti-interference is an important indicator. Interference signals in satellite navigation include jamming interference, spoofing interference, and perturbation interference, which will seriously affect the reception quality of satellite navigation signals, resulting in a decrease in positioning accuracy or even complete failure. Identifying the types of satellite navigation interference signals refers to analyzing and processing the received satellite navigation signals to accurately determine whether there is interference in the signals and further identify the types of interference signals, which is of great significance for improving the anti-interference ability of satellite navigation systems and ensuring the stability of navigation signals.

[0003] In the prior art, in terms of identifying the types of satellite navigation interference signals, traditional methods mainly include methods based on signal processing, statistical models, and machine learning. Among them, methods based on signal processing usually rely on filtering, spectrum analysis, etc. After extracting the features of the signals, they judge the presence and types of interference signals through preset rules or thresholds. However, such methods are difficult to cope with complex and changeable interference environments, and the extraction and identification of signal features rely on manual experience and preset rules, lacking flexibility and self-adaptability; methods based on statistical models identify interference signals by establishing statistical models of signals and using statistical parameters. Although the accuracy and robustness of identification are improved to a certain extent, the process of establishing and optimizing the models is complex, and it is difficult to adapt to rapidly changing interference environments; when facing numerous interference types, methods based on machine learning usually adopt the strategy of equipping each subtype with a dedicated identification algorithm. Although the identification accuracy of specific interference is improved, the algorithm complexity and maintenance cost are increased, a large amount of resources are consumed, and false alarms may be caused by conflicts between algorithms, affecting the overall identification efficiency. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, the present invention aims to provide a method and system for identifying types of satellite navigation interference signals to solve the problems of insufficient adaptability of traditional methods for identifying types of satellite navigation interference signals and false alarms caused by algorithm conflicts.

[0005] To achieve the above object, the embodiments of the present invention adopt the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a method for identifying types of satellite navigation interference signals, the method comprising:

[0007] Step S1, collect complete navigation signals from a satellite receiver;

[0008] Step S2, calculate the carrier-to-noise ratio and pseudorange residual information of the complete navigation signals, and determine whether there are navigation interference signals in the complete navigation signals;

[0009] Step S3, if navigation interference signals are detected, extract intermediate frequency data from the complete navigation signals and perform preprocessing; if navigation interference signals are not detected, return to Step S1;

[0010] Step S4, perform spectral analysis on the intermediate frequency data to extract the frequency domain features of the complete navigation signals;

[0011] Step S5, perform time domain analysis on the intermediate frequency data in the complete navigation signals to extract the time domain features of the complete navigation signals;

[0012] Step S6, construct a navigation interference signal recognition network, and combine the frequency domain features and time domain features of the complete navigation signals to calculate the types of navigation interference signals.

[0013] As a preferred embodiment of the present invention, Step S1 further comprises: using a satellite receiver to receive navigation signals from no less than 4 satellites, demodulating the received navigation signals, and converting them from radio frequency signals to baseband signals as complete navigation signals.

[0014] As a preferred embodiment of the present invention, Step S2 further comprises:

[0015] Step S21, decode the complete navigation signals, extract information such as satellite orbit parameters, signal propagation time, and receiver position, and at the same time record the signal power through the radio frequency front-end circuit of the receiver and record the noise power through the noise measurement circuit of the receiver;

[0016] Step S22, calculate the carrier-to-noise ratio using the signal power and noise power, and the calculation formula (1) is as follows:

[0017]

[0018] In formula (1), CNR is the carrier-to-noise ratio, lg() is the logarithmic operation with base 10, P1 is the signal power, and P2 is the noise power;

[0019] Step S23: Multiply the signal propagation time measured by the receiver by the speed of light to calculate the actual measured pseudorange; use the satellite orbit parameters and the receiver position, and calculate the theoretical pseudorange using the least squares method; subtract the theoretical pseudorange from the actual measured pseudorange to obtain the pseudorange residual ρ.

[0020] Step S24: Calculate the probability of the existence of interference signals, and determine whether there are interference signals according to the probability of the existence of interference signals.

[0021] As a preferred embodiment of the present invention, the formula (2) for calculating the probability of the existence of interference signals in step S24 is as follows:

[0022] pro=sigmoid(MLP(ρ,CNR)) (2)

[0023]

[0024] In formula (2), pro is the probability of the existence of interference signals, sigmoid() is the sigmoid function, and MLP() is the multi-layer perceptron operation;

[0025] When determining whether there are interference signals, if pro = 1, it is considered that there are navigation interference signals; if pro = 0, it is considered that there are no navigation interference signals.

[0026] As a preferred embodiment of the present invention, step S4 further includes:

[0027] Step S41: Use the fast Fourier transform algorithm to convert the intermediate frequency data in the complete navigation signal into a frequency domain signal;

[0028] Step S42: Add a sliding window to the frequency domain signal, and extract the frequency domain features of the complete navigation signal based on the frequency domain signal in the sliding window. The extraction formulas (3)-(7) are as follows:

[0029] F frequency =

[0030] MLP(Concat(F weighted window ,F global ),F global ,F weighted window ) (3)

[0031] In formula (3), F weighted window is the weighted sliding window feature; F global is the global feature; and,

[0032] F weighted window =∑ i w i ·F i,window (4)

[0033]

[0034] w i = Attention(F i,window ) (6)

[0035]

[0036] In formulas (4)-(7), w i is the attention weight, F i,window is the feature of the i-th sliding window, i is the sliding window index, n_window is the number of sliding windows; MLP() is the multi-layer perceptron operation, Concat() is the concatenation operation, W is the window size, X i is the frequency domain signal of the i-th sliding window, CNN() is the convolutional neural network operation; Attention() is the attention mechanism operation; GRU() is the gated recurrent unit operation, and X is the frequency domain signal of the intermediate frequency data in the complete navigation signal.

[0037] As a preferred embodiment of the present invention, step S5 further includes:

[0038] Step S51, using a gated recurrent unit network, extract preliminary time domain features based on the intermediate frequency data in the complete navigation signal, and the calculation formula (8) is as follows:

[0039] h T = GRU(T) (8)

[0040] In formula (8), h T is the preliminary time domain feature, and T is the intermediate frequency data in the complete navigation signal;

[0041] Step S52, using the attention mechanism, calculate the weights of each time step and normalize them. The calculation formula is:

[0042] w' t = Attention(h t ) (9)

[0043]

[0044] In formulas (9) and (10), t is the time step index, w' t is the weight of each time step, and a t is the normalized weight of each time step;

[0045] Step S53, sort the normalized attention weights a t in ascending order, and select the first n time steps. n is the number of time steps that satisfy the cumulative attention weight exceeding the threshold p. The calculation formula (11) is as follows:

[0046]

[0047] In Equation (11), TimeSteps is the set of time steps, and j is the counting index;

[0048] Step S54: Modify the gated recurrent unit network. The specific operations are as follows: For the selected n time steps, add a neural network connection between two adjacent time steps so that information can be directly transmitted among these n time steps; the modified gated recurrent unit network is denoted as GRU′();

[0049] Step S55: Calculate the time-domain features of the complete navigation signal. The calculation formula (12) is as follows:

[0050]

[0051] h′ T = GRU′(T)

[0052] In Equation (12), F time is the time-domain feature of the complete navigation signal, are the hidden states corresponding to the 1st, 2nd, …, nth time steps in the time-step set TimeSteps respectively, and h′ T is the modified preliminary time-domain feature.

[0053] As a preferred embodiment of the present invention, Step S6 further includes:

[0054] Step S61: Calculate the navigation interference signal type features based on the frequency-domain features and the time-domain features. The calculation formula is as follows:

[0055] f1 = MLP(Concat(F time , F frequency )) (13)

[0056]

[0057] g1 = tanh(β1·f2 + b1) (15)

[0058] g2 = tanh(β2·f1 + b2) (16)

[0059] f′1 = f1·g1 (17)

[0060] f′2 = f2·g2 (18)

[0061] f = MLP(Concat(f′1, f′2)) (19)

[0062] In formulas (13)-(19), f1 is the longitudinal fusion feature, f2 is the transverse fusion feature, CNN (1) is the longitudinal one-dimensional convolution operation, g1 is the longitudinal gate, tanh() is the hyperbolic tangent function, β1 is the longitudinal parameter, b1 is the longitudinal bias term, g2 is the transverse gate, β2 is the transverse parameter, b2 is the transverse bias term, and f is the navigation interference signal type feature;

[0063] Step S62: Calculate the probabilities of various types of navigation interference signals based on the navigation interference signal type feature; identify the navigation interference signal type according to the calculated probabilities of various types of navigation interference signals.

[0064] As a preferred embodiment of the present invention, the formula (20) for calculating the probabilities of various types of navigation interference signals is as follows:

[0065] p = softmax(f) (20)

[0066] In formula (20), p is the probability of various types of navigation interference signals, and softmax() is the softmax function.

[0067] As a preferred embodiment of the present invention, the method further includes:

[0068] Step S7: Optimize the navigation interference signal recognition network by using the stochastic gradient descent optimization algorithm in combination with the L2 regularization technique;

[0069] And the specific operations of step S7 are as follows:

[0070] Step S71: Randomly initialize the weight and bias parameters of the navigation interference signal recognition network, set the learning rate to 0.001, and the regularization coefficient to 0.01;

[0071] Step S72: Use the cross-entropy loss function to measure the difference between the network prediction result and the true label;

[0072] Step S73: Use the stochastic gradient descent algorithm to update the network parameters until the navigation interference signal recognition network converges.

[0073] In a second aspect, an embodiment of the present invention further provides a satellite navigation interference signal type recognition system, which includes: a signal acquisition module, a presence probability calculation module, a preliminary judgment module, an intermediate frequency data acquisition module, a frequency domain feature extraction module, a time domain feature extraction module, and an interference type recognition module; wherein,

[0074] The signal acquisition module is used to acquire complete navigation signals from a satellite receiver;

[0075] The existence probability calculation module is used to calculate the carrier-to-noise ratio and pseudorange residual information of the complete navigation signal, determine whether there is a navigation interference signal in the complete navigation signal, and send the judgment result to the preliminary judgment module;

[0076] The preliminary judgment module is used to start the intermediate frequency data acquisition module when receiving the result of the existence of a navigation interference signal; and start the signal acquisition module when receiving the result of the non-existence of a navigation interference signal;

[0077] The intermediate frequency data acquisition module is used to extract the intermediate frequency data in the complete navigation signal and perform preprocessing when detecting a navigation interference signal;

[0078] The frequency domain feature extraction module is used to perform spectrum analysis on the intermediate frequency data and extract the frequency domain features of the complete navigation signal;

[0079] The time domain feature extraction module is used to perform time domain analysis on the intermediate frequency data in the complete navigation signal and extract the time domain features of the complete navigation signal;

[0080] The interference type recognition module is used to calculate the navigation interference signal type by combining the frequency domain features and time domain features of the complete navigation signal.

[0081] The technical solution provided by the embodiment of the present invention has the following beneficial effects:

[0082] (1) By introducing the idea of deep learning, first, based on the carrier-to-noise ratio and pseudorange residual information, a preliminary judgment is made on whether there is interference in the satellite navigation signal; then, feature extraction is performed separately on the frequency domain and time domain of the navigation signal. Finally, a navigation interference signal recognition network is constructed to fuse the frequency domain and time domain features, improving the adaptability of the satellite navigation interference signal type recognition method and avoiding the problem that each interference type requires a specific algorithm for targeted recognition; and since the present invention adopts a fusion recognition technical idea and there is no step of separately judging the interference types, the situation of simultaneous alarms of various signal interference recognition algorithms is also avoided;

[0083] (2) Innovated the frequency domain feature extraction method, introduced the calculation idea of a sliding window for local feature extraction, and used the attention mechanism to assign weights to each sliding window feature, considering the importance of local features; at the same time, applied GRU to extract features from the global frequency domain signal and splice them with the weighted sliding window features. Finally, the frequency domain features of the complete navigation signal are obtained through a multi-layer perceptron, improving the effect of frequency domain feature extraction and laying a data foundation for subsequent navigation interference signal classification;

[0084] (3) Innovated the time-domain feature extraction method. Considering the problem of long-term information loss in the traditional gated recurrent unit network, which is not conducive to analyzing the dynamic characteristics of signal interference, based on the importance of time steps, an information transmission channel was added in the gated recurrent unit network, enabling information to be transmitted jumpily, reducing information attenuation, and improving the overall time-domain feature extraction ability;

[0085] (4) Innovated the feature fusion method in the frequency domain and time domain, and proposed a gating mechanism based on a multi-layer perceptron. Considering the characteristics of information in both vertical and horizontal dimensions, important features were highlighted and irrelevant features were suppressed; compared with the feature fusion based on the attention mechanism, the gating mechanism proposed in the embodiments of the present invention is based on simple mathematical operations, with a more concise and efficient calculation process; and the intensity of each gating signal directly reflects the importance of the corresponding feature, which is helpful for understanding and debugging the model.

[0086] Of course, it is not necessary for any product or method implementing the present invention to simultaneously achieve all the above-mentioned advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0088] Figure 1 It is a flowchart of the method for identifying satellite navigation interference signal types described in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0089] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. The components of the embodiments of the present invention described and illustrated in the drawings here can be arranged and designed in various different configurations. It should be noted that, without conflict, the embodiments and features in the embodiments of the present invention can also be combined with each other.

[0090] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In the description of the present invention, the terms "first", "second", "third", "fourth", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0091] In view of the problems existing in the recognition of satellite navigation interference signal types, embodiments of the present invention provide a method and system for recognizing satellite navigation interference signal types, aiming to solve the problems of insufficient adaptability of traditional satellite navigation interference signal type recognition methods and false alarm problems caused by algorithm conflicts.

[0092] See Figure 1 , the method for recognizing satellite navigation interference signal types includes the following steps:

[0093] Step S1, collect complete navigation signals from a satellite receiver.

[0094] In this step, use a satellite receiver to receive navigation signals from no less than 4 satellites, demodulate the received navigation signals, and convert them from radio frequency signals to baseband signals as complete navigation signals.

[0095] Step S2, calculate the carrier-to-noise ratio and pseudorange residual information of the complete navigation signal, and determine whether there is a navigation interference signal in the complete navigation signal.

[0096] Specifically, this step further includes:

[0097] Step S21, decode the complete navigation signal, extract information such as satellite orbit parameters, signal propagation time, and receiver position, and at the same time record the signal power through the radio frequency front-end circuit of the receiver and record the noise power through the noise measurement circuit of the receiver;

[0098] Step S22, calculate the carrier-to-noise ratio using the signal power and noise power, and the calculation formula (1) is as follows:

[0099]

[0100] In formula (1), CNR is the carrier-to-noise ratio, lg() is the logarithmic operation with base 10, P1 is the signal power, and P2 is the noise power.

[0101] Step S23, multiply the signal propagation time measured by the receiver by the speed of light to calculate the actual measured pseudorange; use the satellite orbit parameters and receiver position, and calculate the theoretical pseudorange using the least squares method; subtract the theoretical pseudorange from the actual measured pseudorange to obtain the pseudorange residual ρ;

[0102] Step S24, calculate the probability of the existence of an interference signal, and the calculation formula (2) is as follows:

[0103] pro = sigmoid(MLP(ρ, CNR)) (2)

[0104]

[0105] In formula (2), pro is the probability of the existence of interference signals, sigmoid() is the sigmoid function, and MLP() is the multi-layer perceptron operation;

[0106] Based on the probability of the existence of interference signals, it is determined whether there are interference signals. The specific determination process is as follows: If pro = 1, it is considered that there are navigation interference signals; if pro = 0, it is considered that there are no navigation interference signals.

[0107] Step S3, if navigation interference signals are detected, the intermediate-frequency data in the complete navigation signal is extracted and preprocessed; if navigation interference signals are not detected, return to step S1.

[0108] In this step, the preprocessing of the intermediate-frequency data includes: removing the DC component and downsampling the intermediate-frequency data; removing the DC component is to eliminate the DC offset in the signal and avoid affecting subsequent analysis; downsampling reduces the computational complexity and removes redundant data at the same time.

[0109] Step S4, perform spectrum analysis on the intermediate-frequency data to extract the frequency-domain characteristics of the complete navigation signal.

[0110] Specifically, this step further includes:

[0111] Step S41, using the fast Fourier transform algorithm, convert the intermediate-frequency data in the complete navigation signal into a frequency-domain signal;

[0112] Step S42, add a sliding window to the frequency-domain signal, and extract the frequency-domain characteristics of the complete navigation signal based on the frequency-domain signal in the sliding window. The extraction formulas (3)-(7) are as follows: Σ

[0113] F frequency =

[0114] MLP(Concat(F weighted window ,F global ),F global ,F weighted window ) (3)

[0115] In formula (3), F weightedwindow is the weighted sliding window feature; F global is the global feature; and,

[0116] F weighted window =∑ i w i ·F i,window (4)

[0117]

[0118] w i =Attention(Fi,window ) (6)

[0119]

[0120] In formulas (4)-(7), w i is the attention weight, F i,window is the feature of the i-th sliding window, i is the sliding window index, n_window is the number of sliding windows; MLP() is the multi-layer perceptron operation, Concta() is the concatenation operation, W is the window size, and X i is the frequency-domain signal of the i-th sliding window, CNN() is the convolutional neural network operation; Attention() is the attention mechanism operation; GRU() is the gated recurrent unit operation, and X is the frequency-domain signal of the intermediate-frequency data in the complete navigation signal.

[0121] In this step, when performing frequency-domain feature extraction, the calculation idea of the sliding window is introduced. Each sliding window can perform local feature extraction, and when performing local feature extraction, the convolutional neural network and the mean feature are integrated; then these local features are concatenated and further processed by MLP to obtain the sliding window features; then, considering that not all the information of the sliding windows should be treated equally, but the key information should be extracted and the irrelevant information should be ignored, the attention mechanism is used to assign weights to each sliding window feature, and the weighted sum is used to obtain the weighted sliding window features; at the same time, the GRU is applied to the global frequency-domain signal to extract features, and they are concatenated with the weighted sliding window features, and finally the frequency-domain features of the complete navigation signal are obtained through the multi-layer perceptron.

[0122] Traditional frequency-domain feature extraction methods rely on spectral analysis or single feature extraction and cannot fully capture the complex interference features in the navigation signal; and due to limited feature extraction ability, when identifying the types of interference signals subsequently, only different algorithm models can be used to compare one by one; when a certain interference pattern meets the discrimination conditions of several algorithms at the same time, the traditional method will fail; and in this step, by introducing the sliding window and weights, a new network feature extraction structure is constructed, and the extracted frequency-domain features provide rich input information for the subsequent navigation interference signal recognition network. Therefore, when identifying the type subsequently, it only needs to be regarded as a multi-classification problem and solved.

[0123] Step S5: Perform time-domain analysis on the intermediate-frequency data in the complete navigation signal to extract the time-domain features of the complete navigation signal.

[0124] Specifically, this step further includes:

[0125] Step S51: Use the gated recurrent unit network to extract the preliminary time-domain features based on the intermediate-frequency data in the complete navigation signal. The calculation formula (8) is as follows:

[0126] h T= GRU(T) (8)

[0127] In formula (8), h T is the preliminary time-domain feature, and T is the intermediate-frequency data in the complete navigation signal.

[0128] Step S52: Using the attention mechanism, calculate the weights for each time step and normalize them. The calculation formula is as follows:

[0129] w' t = Attention(h t ) (9)

[0130]

[0131] In formulas (9) and (10), t is the time-step index, w' t is the weight for each time step, and a t is the normalized weight for each time step.

[0132] Step S53: Sort the normalized attention weights a t in ascending order, and select the first n time steps. n is the number of time steps that satisfy the cumulative attention weight exceeding the threshold p. The calculation formula (11) is as follows:

[0133]

[0134] In formula (11), TimeSteps is the set of time steps, and j is the counting index.

[0135] Step S54: Modify the gated recurrent unit network. The specific operation is as follows: For the selected n time steps, add a neural network connection between adjacent time steps so that information can be directly transmitted among these n time steps. The modified gated recurrent unit network is denoted as GRU'().

[0136] Step S55: Calculate the time-domain feature of the complete navigation signal. The calculation formula (12) is as follows:

[0137]

[0138] h' T = CRU'(T)

[0139] In formula (12), F time is the time-domain feature of the complete navigation signal, are the hidden states corresponding to the 1st, 2nd, …, nth time steps in the time-step set TimeSteps respectively, and h' T is the modified preliminary time-domain feature.

[0140] In this step, the time-domain information is the key to understanding the dynamic behavior of the navigation signal. Traditional methods only rely on time-domain analysis (time-domain waveform analysis, probability density analysis, etc.) or gated recurrent unit networks. The traditional time-domain analysis methods cannot fully capture the complex time-domain features in the navigation signal, while the gated recurrent unit network has the problem of long-term information loss. In contrast, in this step, a gated recurrent unit network is first used to perform preliminary time-domain feature extraction on the navigation signal to capture the time dependence of the signal. Then, the attention mechanism is used to calculate the weights of each time step to highlight the information of important time steps. Finally, through a multi-layer perceptron and feature fusion operations, the features of these key time steps are integrated, and considering the global features updated by the gated recurrent unit network, the time-domain features of the complete navigation signal are obtained.

[0141] The traditional way of combining the attention mechanism with the gated recurrent unit directly takes the last hidden state as the network output. However, in this step, the first n key time steps are selected according to the attention weights, and neural network connections are added between these time steps, so that information can be transmitted through the newly added neural network connections, solving the problem of long-term information loss in the gated recurrent unit.

[0142] Step S6: Construct a navigation interference signal recognition network, and calculate the type of navigation interference signal by combining the frequency-domain features of the complete navigation signal and the time-domain features of the complete navigation signal.

[0143] Specifically, this step further includes:

[0144] Step S61: Based on the frequency-domain features and time-domain features, calculate the feature of the navigation interference signal type. The calculation formula is as follows:

[0145] f1 = MLP(Concat(F time ,F frequency )) (13)

[0146]

[0147] g1 = tanh(β1·f2 + b1) (15)

[0148] g2 = tanh(β2·f1 + b2) (16)

[0149] f′1 = f1·g1 (17)

[0150] f′2 = f2·g2 (18)

[0151] f = MLP(Concat(f′1,f′2)) (19)

[0152] In formulas (13)-(19), f1 is the longitudinal fusion feature, f2 is the transverse fusion feature, CNN(1) is a vertical one-dimensional convolution operation, g1 is the vertical gate, tanh() is the hyperbolic tangent function, β1 is the vertical parameter, b1 is the vertical bias term, g2 is the horizontal gate, β2 is the horizontal parameter, b2 is the horizontal bias term, and f is the feature of the navigation interference signal type.

[0153] In this step, the vertical fusion feature and the horizontal fusion feature are calculated respectively. The vertical fusion feature aims to directly fuse the frequency domain and time domain features; the horizontal fusion feature aims to analyze the information of the time domain and frequency domain step by step through one-dimensional convolution to capture the local correlation between features. At the same time, a gating mechanism based on a multi-layer perceptron is proposed. Considering the characteristics of the vertical and horizontal dimensional information, the fusion features are weighted to highlight important features and suppress irrelevant features. Compared with the traditional attention mechanism, the gating mechanism is based on simple mathematical operations, and the calculation process is more concise and efficient. In addition, the intensity of each gating signal directly reflects the importance of the corresponding feature, which helps to understand and debug the model. Although the attention mechanism can also perform feature weighting, its internal operation process is more complex and not easy to be intuitively explained.

[0154] Step S62: Calculate the probabilities of each type of navigation interference signal according to the feature of the navigation interference signal type. The calculation formula (20) is as follows:

[0155] p = softmax(f) (20)

[0156] In formula (20), p is the probability of each type of navigation interference signal, and softmax() is the softmax function;

[0157] Identify the type of navigation interference signal according to the calculated probabilities of each type of navigation interference signal.

[0158] For example, if the calculated probabilities p of each type of navigation interference signal are: [0.2, 0.45, 0.35], it means that the probability that the signal is a jamming interference is 0.2, the probability that it is a spoofing interference is 0.45, and the probability that it is a perturbation interference is 0.35. Then it will finally be determined that the type of this navigation interference signal is spoofing interference.

[0159] Here, the situation of incorrect preliminary judgment in step S1 can also be solved by adding an option for non-interference signals, that is, when it is considered that there is interference in the preliminary judgment, but in fact, there is no interference. For example, when a certain signal has a calculated pro = 0.501 in step S1, according to the judgment rule, this signal is preliminarily judged to have interference; however, after in-depth judgment of time-domain and frequency-domain feature extraction and fusion, it is found that this signal is a normal signal both in the time domain and the frequency domain; to cope with this situation, a new type needs to be added to the probabilities of various types of navigation interference signals, changing it from a 3D vector to a 4D vector. For example, [0.2, 0.25, 0.2, 0.35] means that the probability of this signal being "non-interfering" is the largest, and finally, the type of this navigation interference signal will be judged as non-interfering, and the preliminary judgment result will be corrected.

[0160] This embodiment may further include:

[0161] Step S7, using the stochastic gradient descent optimization algorithm and combining with the L2 regularization technique to optimize the navigation interference signal recognition network.

[0162] Specifically, this step further includes:

[0163] Step S71, randomly initialize the weight and bias parameters of the navigation interference signal recognition network, set the learning rate to 0.001, and the regularization coefficient to 0.01;

[0164] Step S72, use the cross-entropy loss function to measure the difference between the network prediction result and the true label;

[0165] Step S73, use the stochastic gradient descent algorithm to update the network parameters until the navigation interference signal recognition network converges.

[0166] Based on the same idea, an embodiment of the present invention also provides a satellite navigation interference signal type recognition system. The system includes:

[0167] The system includes: a signal acquisition module, an existence probability calculation module, a preliminary judgment module, an intermediate frequency data acquisition module, a frequency-domain feature extraction module, a time-domain feature extraction module, and an interference type recognition module; preferably, the system may further include an algorithm optimization module; wherein,

[0168] The signal acquisition module is used to acquire complete navigation signals from a satellite receiver;

[0169] The existence probability calculation module is used to calculate the carrier-to-noise ratio and pseudorange residual information of the complete navigation signal, judge whether there is a navigation interference signal in the complete navigation signal, and send the judgment result to the preliminary judgment module;

[0170] The preliminary judgment module is used to start the intermediate frequency data acquisition module when receiving the result of the existence of navigation interference signals; and start the signal acquisition module when receiving the result of the non-existence of navigation interference signals.

[0171] The intermediate frequency data acquisition module is used to extract the intermediate frequency data in the complete navigation signal and perform preprocessing when detecting navigation interference signals.

[0172] The frequency domain feature extraction module is used to perform spectrum analysis on the intermediate frequency data and extract the frequency domain features of the complete navigation signal.

[0173] The time domain feature extraction module is used to perform time domain analysis on the intermediate frequency data in the complete navigation signal and extract the time domain features of the complete navigation signal.

[0174] The interference type identification module is used to calculate the navigation interference signal type by combining the frequency domain features and time domain features of the complete navigation signal.

[0175] The algorithm optimization module is used to optimize the navigation interference signal recognition network by adopting the stochastic gradient descent optimization algorithm and combining with the L2 regularization technique.

[0176] In this embodiment, each module is implemented by a processor, and a memory is appropriately added when storage is required. Among them, the processor may be, but is not limited to, a microprocessor MPU, a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), other programmable logic devices, discrete gate, transistor logic devices, discrete hardware components, etc. The memory may include a random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far from the aforementioned processor.

[0177] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.).

[0178] In addition, it should be noted that the satellite navigation interference signal type recognition system described in this embodiment corresponds to the satellite navigation interference signal type recognition method. The description and limitation of the method also apply to the system, and will not be repeated here.

[0179] From the above technical solutions, it can be seen that the satellite navigation interference signal type recognition method and system provided by the embodiments of the present invention collect complete navigation signals from a satellite receiver, calculate the carrier-to-noise ratio and pseudorange residual information, and determine whether there is a navigation interference signal in the complete navigation signal; if a navigation interference signal is detected, the intermediate frequency data in the complete navigation signal is collected and spectrum analysis is performed. A sliding window is added to the intermediate frequency data, and based on the sliding window and the sliding window weight, combined with a gated recurrent unit, the frequency domain features of the complete navigation signal are extracted, and the time domain analysis of the intermediate frequency data in the complete navigation signal is performed to extract the time domain features of the complete navigation signal; combining the frequency domain features of the complete navigation signal and the time domain features of the complete navigation signal, the type of the navigation interference signal is calculated. The present invention has the following beneficial effects:

[0180] (1) Introducing the idea of deep learning, first make a preliminary judgment on whether there is interference in the satellite navigation signal according to the carrier-to-noise ratio and pseudorange residual information; then extract features from the frequency domain and time domain of the navigation signal respectively, and finally construct a navigation interference signal recognition network to fuse the frequency domain and time domain features, improving the adaptability of the satellite navigation interference signal type recognition method and avoiding the problem that each interference type requires a specific algorithm for targeted recognition; and because the present invention adopts the technical idea of fusion recognition and there is no step of separately judging for each interference type, the situation of simultaneous alarms of various signal interference recognition algorithms is also avoided.

[0181] (2) Innovated the frequency-domain feature extraction method, introduced the calculation idea of a sliding window, performed local feature extraction, and used the attention mechanism to assign weights to each sliding window feature, considering the importance of local features. At the same time, applied GRU to extract features from the global frequency-domain signal and concatenated them with the weighted sliding window features. Finally, obtained the frequency-domain features of the complete navigation signal through a multi-layer perceptron, improving the effect of frequency-domain feature extraction and laying a data foundation for subsequent navigation interference signal classification.

[0182] (3) Innovated the time-domain feature extraction method. Considering the problem of long-term information loss in the traditional gated recurrent unit network, which is not conducive to analyzing the dynamic characteristics of signal interference, the present invention adds an information transmission channel in the gated recurrent unit network based on the importance of time steps, enabling information to be transmitted jumpwise, reducing information attenuation, and improving the overall time-domain feature extraction ability.

[0183] (4) Innovated the feature fusion method in the frequency domain and time domain, proposed a gating mechanism based on a multi-layer perceptron. Considering the characteristics of information in both longitudinal and transverse dimensions, it highlights important features and suppresses irrelevant features. Compared with the feature fusion based on the attention mechanism, the gating mechanism proposed in the present invention is based on simple mathematical operations, with a more concise and efficient calculation process. Moreover, the intensity of each gating signal directly reflects the importance of the corresponding feature, which is helpful for understanding and debugging the model.

[0184] The above description is only a preferred embodiment of the present invention and an explanation of the applied technical principles, and is not intended to limit the scope of the present invention claimed, but only represents the preferred embodiments of the present invention. Those skilled in the art should understand that the scope of the invention involved in the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

Claims

1. A method for identifying the type of satellite navigation interference signal, characterized in that: The method comprises: Step S1, collecting complete navigation signals from a satellite receiver; Step S2, calculating the carrier-to-noise ratio and pseudorange residual information of the complete navigation signal, and determining whether there is a navigation interference signal in the complete navigation signal; Step S3, if a navigation interference signal is detected, extracting the intermediate frequency data in the complete navigation signal and preprocessing it; if no navigation interference signal is detected, returning to step S1; Step S4, performing spectrum analysis on the intermediate frequency data to extract frequency domain features of the complete navigation signal; Step S5, performing time domain analysis on the intermediate frequency data in the complete navigation signal to extract the time domain features of the complete navigation signal; Step S6, constructing a navigation interference signal identification network, combining the frequency domain characteristics of the complete navigation signal and the time domain characteristics of the complete navigation signal, and calculating the type of the navigation interference signal.

2. The satellite navigation interference signal type identification method according to claim 1, characterized in that: Step S1 further includes: using a satellite receiver to receive navigation signals from no less than 4 satellites, and demodulating the received navigation signals, converting them from radio frequency signals to baseband signals as complete navigation signals.

3. The satellite navigation interference signal type identification method according to claim 1, characterized in that: Step S2 further comprises: Step S21, decoding the complete navigation signal, extracting information such as satellite orbit parameters, signal propagation time and receiver position, and recording the signal power through the RF front-end circuit of the receiver and the noise power through the noise measurement circuit of the receiver; Step S22, using the signal power and the noise power to calculate the carrier-to-noise ratio, the calculation formula (1) is as follows: In formula (1), CNR is the carrier-to-noise ratio, lg() is the logarithm operation with base 10, P1 is the signal power, and P2 is the noise power; Step S23, multiplying the signal propagation time measured by the receiver by the speed of light to calculate the actual measured pseudorange; using the satellite orbit parameters and the receiver position, the least square method is used to calculate the theoretical pseudorange; subtracting the theoretical pseudorange from the actual measured pseudorange to obtain the pseudorange residual ρ; Step S24, calculating the probability of the existence of an interference signal, and judging whether an interference signal exists according to the probability of the existence of the interference signal.

4. The satellite navigation interference signal type identification method according to claim 3 is characterized in that: Step S24 calculates the probability formula (2) of the presence of an interference signal as follows: pro=sigmoid(MLP(ρ,CNR)) (2) In formula (2), pro is the probability of the existence of interference signal, sigmoid() is the sigmoid function, and MLP ( ) is the multi-layer perceptron operation; When judging whether there is an interference signal, if pro=1, it is considered that there is a navigation interference signal; if pro=0, it is considered that there is no navigation interference signal.

5. The satellite navigation interference signal type identification method according to claim 1, characterized in that: Step S4 further comprises: Step S41, using a fast Fourier transform algorithm to convert the intermediate frequency data in the complete navigation signal into a frequency domain signal; Step S42, adding a sliding window to the frequency domain signal, extracting the frequency domain features of the complete navigation signal based on the frequency domain signal in the sliding window, and extracting formulas (3)-(7) as follows: F frequency ＝ MLP(Concat(F weightedwindow ,F global ),F global ,F weightedwindow ) (3) In formula (3), F weightedwindow is the weighted sliding window feature; F global is a global feature; and F weightedwindow =∑ i w i ·F i,window (4) w i =Attention(F i,window ) (6) In formulas (4)-(7), w i is the attention weight, F i,window is the feature of the i-th sliding window, i is the sliding window index, n_window is the number of sliding windows; MLP() is the multi-layer perceptron operation, Concat() is the concatenation operation, W is the window size, X i is the frequency domain signal of the i-th sliding window, CNN() is the convolutional neural network operation; Attention() is the attention mechanism operation; GRU() is the gated recurrent unit operation, and X is the frequency domain signal of the intermediate frequency data in the complete navigation signal.

6. The method for identifying the type of satellite navigation interference signal according to claim 5, characterized in that: Step S5 further comprises: Step S51, using a gated recurrent unit network, extracting preliminary time domain features based on the intermediate frequency data in the complete navigation signal, the calculation formula (8) is as follows: h T =GRU(T) (8) In formula (8), h t is the preliminary time domain feature, T is the intermediate frequency data in the complete navigation signal; Step S52, using the attention mechanism, calculates the weight of each time step and normalizes it. The calculation formula is: w′ t =Attention(h t ) (9) In formulas (9) and (10), t is the time step index, w′ t is the weight of each time step, a t is the normalized weight of each time step; Step S53: normalize the attention weight a t Sort in positive order and select the first n time steps, where n is the number of time steps that satisfy the cumulative attention weight exceeding the threshold p. The calculation formula (11) is as follows: In formula (11), TimeSteps is the set of time steps, j is the count index; Step S54, modifying the gated recurrent unit network, the specific operation is as follows: for the selected n time steps, add a neural network connection between two adjacent time steps, so that information can be directly transmitted on these n time steps; the modified gated recurrent unit network is represented as GRU′(); Step S55, calculate the time domain characteristics of the complete navigation signal, and the calculation formula (12) is as follows: h′ T =GRU′(T) In formula (12), F time is the time domain characteristics of the complete navigation signal, are the hidden states corresponding to the 1st, 2nd, ..., nth time steps in the time step set TimeSteps, respectively, h′ T It is the modified preliminary time domain feature.

7. The method for identifying the type of satellite navigation interference signal according to claim 5, characterized in that: Step S6 further comprises: Step S61, based on the frequency domain characteristics and the time domain characteristics, the navigation interference signal type characteristics are calculated, and the calculation formula is as follows: f1=MLP(Concat(F time ,F frequency )) (13) g1=tanh(β1·f2+b1) (15) g2=tanh(β2·f1+b2) (16) f′1=f1·g1 (17) f′2=f2·g2 (18) f=MLP(Concat(f′1,f′2)) (19) In formulas (13)-(19), f1 is the vertical fusion feature, f2 is the horizontal fusion feature, and CNN (1) is the longitudinal one-dimensional convolution operation, g1 is the longitudinal gate, tanh( ) is the hyperbolic tangent function, β1 is the longitudinal parameter, b1 is the longitudinal bias term, g2 is the transverse gate, β2 is the transverse parameter, b2 is the transverse bias term, and f is the navigation interference signal type feature; Step S62, calculating the probability of each type of navigation interference signal according to the type characteristics of the navigation interference signal; and identifying the type of the navigation interference signal according to the calculated probability of each type of navigation interference signal.

8. The method for identifying the type of satellite navigation interference signal according to claim 7, characterized in that: The probability formula (20) for calculating each type of navigation interference signal is as follows: p = softmax(f) (20) In formula (20), p is the probability of each type of navigation interference signal, and softmax( ) is the softmax function.

9. The satellite navigation interference signal type identification method according to claim 1, characterized in that: The method further comprises: Step S7, using a stochastic gradient descent optimization algorithm combined with L2 regularization technology to optimize the navigation interference signal recognition network; The specific operations of step S7 are as follows: Step S71, randomly initializing the weights and bias parameters of the navigation interference signal recognition network, setting the learning rate to 0.001, and the regularization coefficient to 0.01; Step S72, using a cross entropy loss function to measure the difference between the network prediction result and the true label; Step S73, using the stochastic gradient descent algorithm to update the network parameters until the navigation interference signal recognition network converges.

10. A satellite navigation interference signal type identification system, characterized in that: The system includes: a signal acquisition module, an existence probability calculation module, a preliminary judgment module, an intermediate frequency data acquisition module, a frequency domain feature extraction module, a time domain feature extraction module and an interference type identification module; wherein, The signal acquisition module is used to acquire complete navigation signals from a satellite receiver; The existence probability calculation module is used to calculate the carrier-to-noise ratio and pseudo-range residual information of the complete navigation signal, determine whether there is a navigation interference signal in the complete navigation signal, and send the determination result to the preliminary determination module; The preliminary judgment module is used to start the intermediate frequency data acquisition module when receiving a result that a navigation interference signal exists; and start the signal acquisition module when receiving a result that a navigation interference signal does not exist; The intermediate frequency data acquisition module is used to extract the intermediate frequency data in the complete navigation signal and perform preprocessing when a navigation interference signal is detected; The frequency domain feature extraction module is used to perform spectrum analysis on the intermediate frequency data to extract the frequency domain features of the complete navigation signal; The time domain feature extraction module is used to perform time domain analysis on the intermediate frequency data in the complete navigation signal to extract the time domain features of the complete navigation signal; The interference type identification module is used to calculate the type of navigation interference signal by combining the frequency domain characteristics of the complete navigation signal and the time domain characteristics of the complete navigation signal.

Images