Beidou satellite signal capturing method and device
By calculating the covariance matrix and hyperparameters of the Beidou satellite signal, the propagator is constructed, and the frequency and phase matrix calculation without knowing the number of signals is realized, which solves the problem of poor capture performance of Beidou satellite signal and improves real-time and accuracy.
Patent Information
- Application Number
- CN202411264004.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-10
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-09-10
AI Technical Summary
The existing Beidou satellite signal capture method has poor performance, insufficient real-time performance, and requires the number of signals to be known in advance. The inaccurate estimates lead to the capture failure.
By obtaining the covariance matrix and hyperparameters of the target satellite signal, building a propagator and calculating the frequency and phase matrix, it realizes capture without knowing the number of signals in advance, reducing the calculation amount and improving real-time performance.
It improves the accuracy and performance of Beidou satellite signal capture, avoids the risk of signal number estimation, reduces calculation time, and is suitable for real-time application scenarios.
Smart Images

Figure CN119045007A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of signal processing technology, and in particular to a Beidou satellite signal capture method and device. Background Art
[0002] The satellite navigation system can provide high-precision, all-day, all-weather positioning, navigation and timing services. It is an information resource that can be shared by military and civilian users in the sea, land and air fields. my country's self-developed "Beidou Satellite Navigation and Positioning System" (hereinafter referred to as "Beidou System") has been officially put into operation since April 2004 and is fully open to civilian users. So far, the entire system has been running stably and has played an important role in many fields such as surveying and mapping, telecommunications, water conservancy, transportation, exploration and national security, bringing great convenience to people's lives. In order to use Beidou satellite signals for positioning and navigation, it is necessary to capture Beidou satellite signals first. The quality of Beidou satellite signal capture method will directly affect the performance of positioning and navigation. Therefore, using a good Beidou satellite signal capture method is very important for the application of Beidou satellites.
[0003] Traditional Beidou satellite signal acquisition methods include time domain serial acquisition method and parallel acquisition algorithm based on fast Fourier transform. The time domain serial acquisition method is a two-dimensional search of the code phase and Doppler frequency of the signal received by the receiver. The amount of calculation is very large and the calculation takes a long time, which cannot meet the real-time requirements. The parallel acquisition algorithm based on fast Fourier transform improves the poor real-time performance of the time domain serial acquisition algorithm, but it still has some disadvantages, such as requiring a lot of computing and storage space, limiting the sampling rate and length of the signal, and having the disadvantage of too long search time in the frequency band far away from the center frequency.
[0004] Moreover, the above-mentioned existing Beidou satellite signal capture method usually requires knowing the number of Beidou satellite signals received in advance, otherwise it is impossible to capture satellite signals, which is almost impossible to achieve in real application scenarios. In order to solve such problems, the existing method is to estimate the number of Beidou satellite signals received, but this has huge risks, because once the number of signals is estimated incorrectly, the Beidou satellite signal cannot be captured correctly, resulting in navigation and positioning work cannot be carried out normally. Therefore, how to improve the capture performance of Beidou satellite signals has become a problem waiting to be solved. Summary of the invention
[0005] The present invention provides a Beidou satellite signal capture method and device, which solves the technical problem of poor performance in the existing Beidou satellite signal capture.
[0006] On one hand, the present invention provides a Beidou satellite signal acquisition method, the method comprising:
[0007] Acquire target satellite signals;
[0008] Calculating the covariance matrix of the target satellite signal;
[0009] Obtaining target hyperparameters corresponding to the target satellite signal;
[0010] constructing a propagator according to the covariance matrix and the target hyperparameter;
[0011] The frequency matrix and the phase matrix of the target satellite signal are calculated based on the propagator.
[0012] Optionally, acquiring the target satellite signal includes:
[0013] Get the initial Beidou satellite signal;
[0014] Extracting a first target signal feature from the initial Beidou satellite signal;
[0015] The first target signal feature is input into a pre-trained noise reduction model, and the target satellite signal is output.
[0016] Optionally, the acquiring a target hyperparameter corresponding to the target satellite signal includes:
[0017] Extracting a second target signal feature from the target satellite signal;
[0018] The second target signal feature is input into a pre-built regression model, and the target hyperparameter is obtained as output.
[0019] Optionally, the propagator is:
[0020] U im =I M×M -R(R H R+τI M×M ) -1 R H
[0021] U im For the propagator, I M×M is the identity matrix, R is the covariance matrix, τ is the target hyperparameter, and H represents the Hermitian transpose.
[0022] Optionally, calculating the frequency matrix and the phase matrix of the target satellite signal according to the propagator includes:
[0023] constructing a spatial spectrum function according to the propagator;
[0024] Traversing the spatial spectrum function, determining a spectrum peak of the spatial spectrum function, and determining a frequency matrix of the target satellite signal according to the spectrum peak;
[0025] According to the frequency matrix of the target satellite signal, a pseudo-inverse matrix of the frequency matrix is calculated;
[0026] Performing least squares on the pseudo-inverse matrix and the target satellite signal to obtain a target column vector;
[0027] Solving the target column vector, a phase matrix of the target satellite signal is obtained.
[0028] Optionally, the step of training the denoising model includes:
[0029] Acquire original data and noise-free signal data; the original data is signal data that integrates Beidou satellite signals and background noise;
[0030] Performing data preprocessing on the original data to obtain an original data sequence, and extracting original signal features from the original data sequence;
[0031] Build a recurrent neural network model;
[0032] The original signal features are used as the input of the recurrent neural network model, and the noise-free signal data is used as the output of the recurrent neural network model. The recurrent neural network model is trained to obtain the denoising model.
[0033] Optionally, the calculating the covariance matrix of the target satellite signal includes:
[0034] Performing signal-frequency sampling on the target satellite signal to obtain a blind separation model corresponding to the target satellite signal;
[0035] An autocorrelation matrix calculation operation is performed on the blind separation model to obtain a covariance matrix of the target satellite signal.
[0036] Optionally, the step of training the regression model includes:
[0037] Acquire an experimental data set, wherein the experimental data set includes sample signal characteristics of noise-free signal data and hyperparameter experimental values;
[0038] Constructing an initial regression model, taking the sample signal features as input of the initial regression model, taking the hyperparameter experimental values as output of the initial regression model, and training the initial regression model using a support vector regression method to obtain optimal model parameters of the initial regression model;
[0039] The regression model is constructed based on the optimal model parameters and the initial regression model.
[0040] Optionally, the blind separation model is:
[0041] x=B·S+n
[0042] x=[x(1),x(2),…x(M)] T
[0043] x∈V M×1
[0044] B=[b1,b2,…b N ]
[0045] B∈V M×N
[0046]
[0047] b k ∈V M×1
[0048] k∈{1,2,…,N}
[0049]
[0050] S∈V N×1
[0051] n=[n(1),n(2),…n(M)] T
[0052] n∈V M×1
[0053] F k =f IF +f k
[0054] Where V represents complex space, T represents transpose, and H represents Hermite transpose; x represents the BeiDou satellite signals collected by different channels at the same sampling point, M represents the number of sampling channels; tg represents the number of samples; A k represents the amplitude of the kth Beidou satellite signal received; θ k,0 represents the initial phase of the kth Beidou satellite signal received; P k represents the product of the ranging code and the navigation message code of the received k-th Beidou satellite signal; F k is the frequency of the kth BeiDou satellite signal; f IF is the intermediate frequency carrier frequency, f k and θ k , 0 respectively represent the frequency and initial phase of the kth Beidou satellite signal collected by the receiver, n(t is the background noise at time t;
[0055] The calculation formula of the covariance matrix of the target satellite signal is:
[0056] R=E(x·x H )=B·E(S·S H )·B H +σ 2 I=B·R xx ·B H +σ 2 I
[0057] Among them, R xx is the covariance matrix of the satellite signal sent by the BeiDou satellite, I is the identity matrix, E is the mathematical expectation, and R is the covariance matrix of the target satellite signal received by the receiver.
[0058] Another aspect of the present invention provides a Beidou satellite signal acquisition device, the device comprising:
[0059] A first acquisition module, used to acquire a target satellite signal;
[0060] A first calculation module, used for calculating the covariance matrix of the target satellite signal;
[0061] A second acquisition module is used to acquire the target hyperparameter corresponding to the target satellite signal;
[0062] A construction module, configured to construct a propagator according to the covariance matrix and the target hyperparameter;
[0063] The second calculation module is used to calculate the frequency matrix and phase matrix of the target satellite signal according to the propagator.
[0064] It can be seen from the above technical solutions that the present invention has the following advantages:
[0065] The present invention provides a Beidou satellite signal acquisition method, comprising acquiring a target satellite signal; calculating a covariance matrix of the target satellite signal; acquiring a target hyperparameter corresponding to the target satellite signal; constructing a propagator according to the covariance matrix and the target hyperparameter; and calculating a frequency matrix and a phase matrix of the target satellite signal according to the propagator.
[0066] In the present invention, by acquiring the target satellite signal after noise reduction, the initial screening of the Beidou satellite signal is realized, and the capture accuracy of the Beidou satellite signal is improved; and by calculating the covariance matrix of the target satellite signal, the target hyperparameter corresponding to the target satellite signal is obtained, and a propagator is constructed according to the covariance matrix and the target hyperparameter, thereby realizing the construction of the propagator without using the number of Beidou satellite signals; and by using the propagator to calculate the frequency matrix and the phase matrix of the target satellite signal, it is achieved that the frequency and phase of the Beidou satellite signal can be calculated without knowing the number of Beidou satellite signals in advance, and the capture of the Beidou satellite signal is realized, the problem of poor capture performance of the existing Beidou satellite signal is solved, and the capture performance of the Beidou satellite signal is improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0068] Figure 1 A schematic diagram of a flow chart of a Beidou satellite signal acquisition method provided by an embodiment of the present invention;
[0069] Figure 2 A schematic diagram of Beidou satellite signal reception provided by an embodiment of the present invention;
[0070] Figure 3 Another schematic diagram of a flow chart of a Beidou satellite signal acquisition method provided by an embodiment of the present invention;
[0071] Figure 4 A schematic diagram of the structure of a noise reduction model provided by an embodiment of the present invention;
[0072] Figure 5 A schematic diagram of the structure of a regression model provided by an embodiment of the present invention;
[0073] Figure 6 Another schematic flow chart of a Beidou satellite signal acquisition method provided as an application example of the present invention;
[0074] Figure 7 A schematic structural diagram of a Beidou satellite signal acquisition device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0075] In the existing Beidou satellite signal capture process, the noise subspace is usually obtained by eigenvalue decomposition, which will generate huge amount of calculation, long calculation time and poor real-time performance. In addition, this method requires knowing the number of Beidou satellite signals received by the receiver in advance, which is almost impossible to do in real application scenarios. In order to solve this problem, the most commonly used solution in existing technologies is to estimate the number of Beidou satellite signals received by the receiver, but this has huge risks. Once the estimated number is inaccurate, the correct noise subspace cannot be obtained, and the spatial spectrum function cannot be correctly constructed, and the Beidou satellite signal capture cannot be completed.
[0076] In view of this, an embodiment of the present invention provides a Beidou satellite signal capture method and device, which solves the technical problem of poor performance of existing Beidou satellite signal capture.
[0077] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0078] See also Figure 1 , an embodiment of the present invention provides a Beidou satellite signal acquisition method, comprising:
[0079] 101. Acquire target satellite signals.
[0080] It should be noted that the target satellite signal refers to the Beidou satellite signal obtained after noise reduction of the Beidou satellite signal received by the receiver.
[0081] 102. Calculate the covariance matrix of the target satellite signal.
[0082] It should be noted that the calculation process of this embodiment is to convert the target satellite signal into a blind separation model, and solve the autocorrelation matrix of the blind separation model, so as to obtain the covariance matrix of the target satellite signal.
[0083] 103. Obtain target hyperparameters corresponding to the target satellite signal.
[0084] It should be noted that there is a correlation between the target satellite signal and the target hyperparameter. This embodiment determines the corresponding target hyperparameter according to the target satellite signal and the correlation.
[0085] 104. Construct a propagator based on the covariance matrix and target hyperparameters.
[0086] It should be noted that if the number K of Beidou satellite signals received by the receiver is known, it can be used to select specific columns in the covariance matrix R to form a full-column rank subspace, construct a propagator, and use its orthogonality with the signal direction vector to construct a spatial spectrum function. However, in actual application scenarios, the number of Beidou satellite signals received by the receiver cannot be known in advance. When the number of Beidou satellite signals received by the receiver is unknown, if the number of signals is estimated, when the estimated number of signals is less than the actual number of signals, the rank of the subspace constructed by the specific columns selected in the covariance matrix R is less than K, the constructed propagator and direction vector do not meet the orthogonality, and the spatial spectrum function cannot be correctly constructed. When the estimated number of sources is greater than the actual number of sources K, the constructed subspace is not full rank, and the inversion operation in the propagator cannot be performed because it does not meet the conditions.
[0087] Since the rank of the matrix R must be greater than or equal to the number of signals, this embodiment uses the entire matrix of the covariance matrix R to construct the propagator, which can ensure that the propagator is orthogonal to the direction vector. In addition, this embodiment introduces hyperparameters to make the matrix R meet the inversion condition and solve R H R may not be full rank, which may result in the inversion operation in the propagator being unable to be performed.
[0088] The constructed propagator is as follows:
[0089] U im =I M×M -R(R H R+τI M×M ) -1 R H
[0090] Among them, τ is a positive hyperparameter, and its optimal value is determined by machine learning methods.
[0091] This embodiment uses the parameter τ to ensure that the above-mentioned propagator meets the conditions of the inversion operation. At this time, it is not necessary to know the number of Beidou satellite signals received by the receiver, and the spatial spectrum function can be correctly constructed, which is in line with the actual usage scenario, avoids eigenvalue decomposition, reduces the amount of calculation, and thus reduces the calculation time, improves the real-time performance of Beidou satellite signal capture, and improves the performance of Beidou satellite signal capture. It avoids the existing Beidou satellite signal capture method. In the process of solving the satellite signal frequency, it is usually necessary to perform eigenvalue decomposition, which results in a large amount of calculation in the solution process, consumes a long time, and cannot meet the real-time requirements well.
[0092] 105. According to the propagator, the frequency matrix and phase matrix of the target satellite signal are calculated.
[0093] It should be noted that the calculation process of this embodiment is to construct the corresponding spatial spectrum function through the propagator, and obtain the frequency matrix and phase matrix of the target satellite signal according to the spatial spectrum function. Among them, the frequency matrix represents the frequency of the target satellite signal, and the phase matrix represents the initial phase of the target satellite signal. If the initial phase and frequency have been obtained, it means that the capture of the Beidou satellite signal has been completed.
[0094] In this embodiment, by acquiring the target satellite signal after noise reduction, the initial screening of the Beidou satellite signal is achieved, and the capture accuracy of the Beidou satellite signal is improved; and by calculating the covariance matrix of the target satellite signal, the target hyperparameter corresponding to the target satellite signal is obtained, and a propagator is constructed according to the covariance matrix and the target hyperparameter, thereby realizing the construction of the propagator without using the number of Beidou satellite signals; and by using the propagator to calculate the frequency matrix and phase matrix of the target satellite signal, it is achieved that the frequency and phase of the Beidou satellite signal can be calculated without knowing the number of Beidou satellite signals in advance, thereby realizing the capture of the Beidou satellite signal, solving the problem of poor capture performance of the existing Beidou satellite signal, and improving the capture performance of the Beidou satellite signal.
[0095] It should be noted that this embodiment does not specifically limit the order between step 102 and step 103 , and step 103 may be performed before step 102 , or after step 102 , or simultaneously with step 102 .
[0096] See also Figure 3 , a Beidou satellite signal acquisition method provided by an embodiment of the present invention includes:
[0097] 201. Obtain initial Beidou satellite signal.
[0098] It should be noted that if Figure 2 As shown, in this embodiment, a receiver is used to receive Beidou satellite signals. The specific form of the Beidou satellite signals received by the receiver can be expressed by the following formula:
[0099]
[0100] Where N is the number of Beidou satellite signals received, C k (t), D k (t) respectively represent the ranging code and navigation message code of the kth Beidou satellite signal collected by the receiver at time t, A k represents the amplitude of the kth Beidou satellite signal collected by the receiver, f IF is the intermediate frequency carrier frequency, f k and θ k,0They respectively represent the frequency and initial phase of the kth Beidou satellite signal collected by the receiver, and n(t) is the background noise at time t, which obeys Gaussian distribution.
[0101] 202. Extract a first target signal feature from an initial BeiDou satellite signal.
[0102] 203. Input the first target signal feature into a pre-trained noise reduction model, and output a target satellite signal.
[0103] It should be noted that the first target signal feature refers to the features that can be used for identification in the initial Beidou satellite signal, such as time domain features and frequency domain features. Among them, the time domain features include mean, standard, etc. The frequency domain features are such as power spectrum density.
[0104] The noise reduction model is pre-built and trained. In actual use, the first target signal feature is input into the noise reduction model, which can effectively reduce the noise of the initial Beidou satellite signal.
[0105] In a specific example, the training steps of the denoising model include:
[0106] S31, obtaining original data and noise-free signal data. The original data is signal data that combines Beidou satellite signals and background noise.
[0107] It should be noted that this embodiment collects raw data and noise-free signal data including Beidou satellite signals and background noise. The noise-free signal data refers to clean Beidou satellite signals without background noise, which can be obtained from a simulation environment or by marking Beidou satellite signals that meet the requirements.
[0108] It can be understood that the present embodiment prepares in advance a training data set and labels for training a denoising model, wherein the labels are used to annotate clean noise-free signals.
[0109] S32, performing data preprocessing on the original data to obtain an original data sequence, and extracting original signal features from the original data sequence.
[0110] It should be noted that data preprocessing includes outlier removal, normalization, data segmentation and feature extraction. Specifically, firstly, outliers are removed from the original data, and normalization is performed. Then, the normalized original data is sliced and cut into multiple signal sequences, each signal sequence contains continuous time steps. Then, useful features (i.e., original signal features) are extracted from the sliced signal sequences, including time domain features (such as mean, standard deviation) and frequency domain features (such as power spectral density).
[0111] S33. Construct a recurrent neural network model.
[0112] It should be noted that the denoising model of this embodiment adopts a recurrent neural network (RNN) model, which specifically includes an input layer, a hidden layer and an output layer. The structure of the recurrent neural network (RNN) model is as follows: Figure 4 As shown, at time t, m t represents the characteristic sequence corresponding to the received signal input at this moment, q t Represents the output value of the hidden layer of the RNN network at the previous moment, m t Together with qt, it determines the value of the hidden layer at time t. is the value of the output layer at time t, y t is the true value of the training sample sequence at time t, and L is the loss function. U is the weight matrix from the input layer to the hidden layer, V is the weight matrix from the hidden layer to the output layer, and the weight matrix W represents the weight of the last value of the hidden layer as the input this time. The initial values of the weight matrices U, V, and W are set based on experience.
[0113] S34. Using the original signal features as the input of the recurrent neural network model and the noise-free signal data as the output of the recurrent neural network model, the recurrent neural network model is trained to obtain a noise reduction model.
[0114] It should be noted that the training principle of the recurrent neural network model is:
[0115] At time t, the original signal features corresponding to the initial Beidou satellite signal are taken as input m t , the corresponding noise-free signal y t As the output target, q t By m t and q t-1 After hidden layer calculation, we get:
[0116] q t =σ(Um t +Wq t-1 +b)
[0117] Among them, σ is the activation function of RNN, which is in the form of:
[0118]
[0119] Where b is the bias value vector.
[0120] After that, the output of the RNN at time t is calculated as:
[0121]
[0122] Where c is the bias value vector.
[0123] Calculate the loss function, the loss function is as follows:
[0124]
[0125] Afterwards, the values of the weight matrix and the bias vector are updated according to the value of the loss function.
[0126] Among them, the updates of V and c are as follows:
[0127]
[0128] To calculate W, U, and b, we need to consider the gradient propagation in the time series at the same time. The hidden layer at time t is defined as:
[0129]
[0130] Then the gradient update of W, U, b is:
[0131]
[0132] Iterative updates are performed according to the above training principles until the optimal model parameters are obtained.
[0133] After that, the recurrent neural network model that has been preliminarily trained is evaluated and its parameters are tuned. The cross-validation method is used to evaluate the performance of the trained RNN model, and the root mean square error is used as the evaluation indicator. According to the model evaluation results, the hyperparameters of the RNN model, such as the learning rate, the number of hidden layer units, the number of iterations, etc., are adjusted to optimize the model performance until the model performance reaches the optimal level, thereby obtaining the best recurrent neural network model, that is, the denoising model. In actual use, the denoising model is used to perform background denoising on the received initial Beidou satellite signal, which can achieve a good denoising effect.
[0134] Therefore, this embodiment uses the trained denoising model to denoise the initial Beidou satellite signal, thereby achieving a better preprocessing effect, thereby obtaining better Beidou satellite signal capture performance, and further improving the Beidou satellite signal capture performance.
[0135] 204. Calculate the covariance matrix of the target satellite signal.
[0136] It should be noted that, in this embodiment, the target satellite signal obtained after noise reduction is converted into a blind separation model, and then the covariance matrix of the target satellite signal is calculated according to the blind separation model.
[0137] Step 204 specifically includes the following sub-steps:
[0138] S41. Perform signal-frequency sampling on the target satellite signal to obtain a blind separation model corresponding to the target satellite signal.
[0139] After the receiver collects the Beidou satellite signal, the signal model received by the receiver is converted into a blind separation model through signal-frequency sampling. The blind separation model is as follows:
[0140] x=B·S+n
[0141] in,
[0142] x=[x(1),x(2),…x(M)] T , x∈V M×1 .
[0143] B=[b1,b2,…b N ],B∈V M×N ,in b k ∈V M ×1 , k∈{1,2,…,N}.
[0144] S∈V N×1 n=[n(1),n(2),…n(M)] T , n∈V M×1 .
[0145] Where V represents complex space, T represents transpose, and H represents Hermite transpose; x represents the Beidou satellite signals collected by different channels at the same sampling point, and M represents the number of sampling channels; F in matrix B k =f IF +f k , t g represents the number of samples; the parameter A in the column vector S k ,θ k,0 Respectively represent the amplitude and initial phase of the kth Beidou satellite signal received; P k Represents the product of the ranging code and the navigation message code of the received k-th Beidou satellite signal.
[0146] Since both the ranging code and the navigation message code are digital coded signals and the sampling is set to be dense enough, it can be considered that P is independent of the sampling channel.
[0147] S42, performing an autocorrelation matrix calculation operation on the blind separation model to obtain a covariance matrix of the target satellite signal.
[0148] It should be noted that according to the calculation formula of the blind separation model, the autocorrelation matrix is calculated on both sides at the same time, and combined with the frequency-invariant characteristic of the sinusoidal signal correlation function, the covariance matrix R of the received signal is obtained, and its calculation formula is specifically as follows:
[0149] R=E(x·x H )=B·E(S·S H )·BH +σ 2 I=B·R xx ·B H +σ 2 I
[0150] Among them, R xx is the covariance matrix of the satellite signal sent by the BeiDou satellite, I is the identity matrix, E is the mathematical expectation, and R is the covariance matrix of the target satellite signal received by the receiver.
[0151] 205. Extract a second target signal feature from the target satellite signal, input the second target signal feature into a pre-built regression model, and output a target hyperparameter.
[0152] It should be noted that the training steps of the regression model include:
[0153] S51. Acquire an experimental data set, where the experimental data set includes sample signal characteristics of noise-free signal data and hyperparameter experimental values.
[0154] It should be noted that in the stage of training the regression model, multiple simulation experiments are conducted in advance, and the characteristics of the signal, hyperparameters, obtained frequencies and other values obtained in each experiment are recorded to obtain the original data set. After that, the original data set is preprocessed to process the duplicate values, missing values and outliers in the original data set, split the data set into a training set and a test set, and the data set is feature-scaled using a normalization method to obtain an experimental data set for training the initial regression model.
[0155] S52. Construct an initial regression model, use the sample signal characteristics as the input of the initial regression model, use the hyperparameter experimental value as the output of the initial regression model, and use the support vector regression method to train the initial regression model to obtain the optimal model parameters of the initial regression model.
[0156] It should be noted that obtaining the value of a hyperparameter is a linear regression problem, which involves finding the correlation between variables through training, determining the relationship between variables, and finally inputting known variables to predict the value of the required hyperparameter. The model function used in this embodiment is y=wx+b.
[0157] Among them, y is the hyperparameter to be solved, x is the variable that affects the value of the hyperparameter, such as the signal-to-noise ratio, w is the weight, and b is the bias.
[0158] It is understandable that the purpose of training the initial regression model is to solve the unknowns w and b and obtain the specific functional form of y. When the characteristics of the signal are known, the value of the hyperparameter can be obtained. For linear regression problems, the goal is to fit a regression line to the data to minimize the error caused by the deviation.
[0159] Specifically, this embodiment uses the support vector regression method (SVR) to train the initial regression model, and sets a threshold error tolerance ε around the regression line, such as Figure 5 As shown. For samples outside the tolerance range, the loss value is calculated, that is, the difference between the predicted value and the true value is calculated. After obtaining the loss value, the model updates each parameter through back propagation to reduce the loss between the true value and the predicted value, so that the predicted value generated by the model moves closer to the true value, thereby achieving the purpose of learning.
[0160] Specifically, this embodiment constructs a slack variable ξ i , to allow some samples to be outside the interval band, such as Figure 5 As shown, Represents the distance between the projection of the upper edge sample point on the upper edge line and the sample point, ξ i Represents the distance between the projection of the lower edge sample point on the lower edge line and the sample point.
[0161] SVR optimizes the model by maximizing the tolerance width and minimizing the total loss. The SVR objective function can be expressed as:
[0162]
[0163] in,
[0164]
[0165] and i The value of is:
[0166]
[0167] In order to minimize the objective function, this embodiment constructs a Lagrangian function according to the constraints, and respectively calculates w, b, ξ i , Find the partial derivative and set it to 0.
[0168] Finally, this embodiment uses the sequential minimum optimization algorithm (SMO) to solve the coefficients w, b of the regression model. At this time, the optimal model parameters obtained are the coefficients w, b.
[0169] S53. Construct a regression model based on the optimal model parameters and the initial regression model.
[0170] It should be noted that the regression model required for use in this embodiment can be obtained by inputting the optimal model parameters into the initial regression model.
[0171] In actual use, the extracted second target signal features are input into the regression model to obtain the optimal hyperparameter, that is, the target hyperparameter.
[0172] It can be understood that the second target signal feature is consistent with the signal feature used to train the initial regression model, such as the signal-to-noise ratio.
[0173] 206. Construct a propagator based on the covariance matrix and target hyperparameters.
[0174] It should be noted that step 206 may refer to step 104, which will not be described in detail in this embodiment.
[0175] 207. According to the propagator, the frequency matrix and phase matrix of the target satellite signal are calculated.
[0176] It should be noted that step 207 specifically includes the following sub-steps:
[0177] S71. Construct a spatial spectrum function based on the propagator.
[0178] It should be noted that the use of propagator U im , construct the following spatial spectrum function:
[0179]
[0180] Among them, G(f) is the spatial spectrum function.
[0181] S72, traversing the spatial spectrum function, determining the spectrum peak of the spatial spectrum function, and determining the frequency matrix of the target satellite signal according to the spectrum peak.
[0182] It should be noted that this embodiment determines a spectrum peak within the preset frequency range by traversing the spatial spectrum within the preset frequency range, and the frequency corresponding to the spectrum peak is the frequency of the target satellite signal, wherein the frequency is expressed in the form of a matrix, that is, a frequency matrix is obtained.
[0183] Among them, the preset frequency range refers to the range of Beidou satellite signal civil frequencies defined in the official documents of the Beidou satellite navigation system.
[0184] S73, calculating a pseudo-inverse matrix of the frequency matrix according to the frequency matrix of the target satellite signal;
[0185] S74, performing least squares on the pseudo-inverse matrix and the target satellite signal to obtain a target column vector;
[0186] S75. Solve the target column vector to obtain the phase matrix of the target satellite signal.
[0187] It should be noted that, according to the obtained frequency of the received Beidou satellite signal, this embodiment can directly perform blind estimation on the initial phase of the signal. Specifically, after calculating the frequency matrix, the pseudo-inverse matrix is calculated, and the pseudo-inverse matrix is used to perform linear least squares on the satellite signal after noise reduction to obtain:
[0188] S=B'·x
[0189] Among them, B' is the pseudo-inverse matrix of the frequency matrix B. Then the initial phase of the Beidou satellite signal can be solved as follows:
[0190]
[0191] Where θ=[θ 1,0 ,θ 2,0 ,…θ N,0 T ,θ∈V N×1 .
[0192] From the above, it can be seen that the frequency and initial phase of the received initial Beidou satellite signal have been obtained, that is, the capture of the Beidou satellite signal has been completed.
[0193] In a specific application example, during the specific implementation process, it can be Figure 6 The indicated process steps implement the present invention. The process specifically includes starting, collecting Beidou satellite signals; determining the specific form of the signal; using a recurrent neural network algorithm to reduce noise on the signal; determining a blind separation model of Beidou satellite signals combined with a blind separation model; obtaining a covariance matrix of the received signal; introducing hyperparameters, constructing a propagator that does not require the use of the number of signals; constructing a spatial spectrum function, traversing the frequency, and obtaining the satellite signal frequency; combining the obtained signal frequency, solving the initial phase of the satellite signal, and ending.
[0194] Among them, the stages of using the recurrent neural network algorithm to reduce the noise of the signal include: data collection and preprocessing; data segmentation and labeling, feature extraction and conversion in the signal sequence; building the RNN model and training the model; evaluating the model using the root mean square error as an indicator; parameter tuning to optimize model performance.
[0195] Among them, introducing hyperparameters and constructing a propagation sub-stage that does not require the use of signal numbers includes repeating experiments, recording data, and obtaining the original data set; preprocessing the data; determining the form of the model function; using the support vector regression method to train and optimize the model; solving the specific form of the model function and obtaining the value of the hyperparameter.
[0196] As can be seen from the above, the present invention solves the problem that the existing Beidou satellite capture algorithm needs to know the number of received Beidou satellite signals in advance, while avoiding estimating the number of signals, and solving the huge risk brought by inaccurate signal number estimation from the root. At the same time, the present invention solves the problem that the existing Beidou satellite signal capture method needs to perform eigenvalue decomposition in the process of solving the satellite signal frequency, thereby generating a large amount of calculation and a long calculation time, hoping to obtain better real-time performance. In addition, the present invention combines high-performance neural network algorithms and machine learning algorithms to improve the preprocessing process of the existing Beidou signal capture method, and obtains better Beidou satellite signal capture performance.
[0197] The present invention constructs a propagator that does not need to use the number of Beidou satellite signals, replacing the traditional method of using eigenvalue decomposition to obtain the noise subspace. The propagator of the present invention is used to construct a spatial spectrum function and perform a spatial spectrum search. The frequency of the Beidou satellite signal required can be obtained when the number of Beidou satellite signals is unknown, and then the signal phase is solved according to the obtained frequency of the satellite signal, thereby completing the capture of the Beidou satellite signal. Therefore, the use of the present invention does not require prior knowledge of the number of Beidou satellite signals to be captured, which is more in line with realistic usage scenarios and is not constrained by prior knowledge of the number of signals. At the same time, since the present invention avoids eigenvalue decomposition, it can significantly reduce the amount of calculation, thereby reducing the calculation time, and can obtain better real-time performance, which is more conducive to application in realistic scenarios. In addition, the present invention uses a high-performance neural network algorithm in the signal preprocessing module and a machine learning algorithm in the hyperparameter value module, which can achieve a better Beidou satellite signal capture effect and obtain better capture performance.
[0198] See also Figure 6 , an embodiment of the present invention provides a Beidou satellite signal acquisition device, comprising:
[0199] A first acquisition module 601 is used to acquire a target satellite signal;
[0200] A first calculation module 602 is used to calculate the covariance matrix of the target satellite signal;
[0201] The second acquisition module 603 is used to acquire the target hyperparameter corresponding to the target satellite signal;
[0202] A construction module 604 is used to construct a propagator according to the covariance matrix and the target hyperparameters;
[0203] The second calculation module 605 is used to calculate the frequency matrix and phase matrix of the target satellite signal according to the propagator.
[0204] In a specific embodiment, the first acquisition module 601 includes:
[0205] A first acquisition unit, used to acquire an initial Beidou satellite signal;
[0206] A first extraction unit is used to extract a first target signal feature from an initial Beidou satellite signal;
[0207] The target satellite signal output unit is used to input the first target signal feature into a pre-trained noise reduction model to output the target satellite signal.
[0208] The first acquisition module 601 also includes: a denoising model training unit.
[0209] In a specific embodiment, the denoising model training unit includes:
[0210] The first acquisition subunit is used to acquire original data and noise-free signal data; the original data is signal data that integrates Beidou satellite signals and background noise;
[0211] The first extraction subunit is used to perform data preprocessing on the original data to obtain an original data sequence, and extract original signal features from the original data sequence;
[0212] A first construction subunit is used to construct a recurrent neural network model;
[0213] The training subunit is used to use the original signal features as the input of the recurrent neural network model and the noise-free signal data as the output of the recurrent neural network model to train the recurrent neural network model and obtain a denoising model.
[0214] In a specific embodiment, the first calculation module 602 includes:
[0215] A model conversion unit is used to perform signal-frequency sampling on the target satellite signal to obtain a blind separation model corresponding to the target satellite signal;
[0216] The first calculation unit is used to perform an autocorrelation matrix calculation operation on the blind separation model to obtain a covariance matrix of the target satellite signal.
[0217] In a specific embodiment, the second acquisition module 603 includes:
[0218] A second extraction unit, used to extract a second target signal feature from the target satellite signal;
[0219] The target hyperparameter output unit is used to input the second target signal feature into a pre-built regression model and output the target hyperparameter.
[0220] In a specific embodiment, the second acquisition module 603 further includes: a regression model training unit,
[0221] The regression model training unit specifically includes:
[0222] A second acquisition subunit is used to acquire an experimental data set, where the experimental data set includes sample signal characteristics of noise-free signal data and hyperparameter experimental values;
[0223] The second training subunit is used to construct an initial regression model, using the sample signal characteristics as the input of the initial regression model, and the hyperparameter experimental value as the output of the initial regression model, and using the support vector regression method to train the initial regression model to obtain the optimal model parameters of the initial regression model;
[0224] The second construction subunit is used to construct a regression model according to the optimal model parameters and the initial regression model.
[0225] In a specific embodiment, the second calculation module 605 includes:
[0226] A space spectrum function construction unit, used for constructing a space spectrum function according to a propagator;
[0227] A frequency matrix determination unit, used for traversing the spatial spectrum function, determining the spectrum peak of the spatial spectrum function, and determining the frequency matrix of the target satellite signal according to the spectrum peak;
[0228] A pseudo-inverse matrix calculation unit, used for calculating a pseudo-inverse matrix of the frequency matrix according to the frequency matrix of the target satellite signal;
[0229] A calculation unit, used for performing least squares on the pseudo-inverse matrix and the target satellite signal to obtain a target column vector;
[0230] The solving unit is used to solve the target column vector to obtain the phase matrix of the target satellite signal.
[0231] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0232] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0233] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0234] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each functional unit may be physically separate, or two or more functional units may be integrated into one processing unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0235] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or partly contributed to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0236] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0237] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A Beidou satellite signal acquisition method, characterized in that: The method comprises: Acquire target satellite signals; Calculating the covariance matrix of the target satellite signal; Obtaining target hyperparameters corresponding to the target satellite signal; constructing a propagator according to the covariance matrix and the target hyperparameter; The frequency matrix and the phase matrix of the target satellite signal are calculated based on the propagator.
2. The method according to claim 1, characterized in that The acquiring of the target satellite signal comprises: Get the initial Beidou satellite signal; Extracting a first target signal feature from the initial Beidou satellite signal; The first target signal feature is input into a pre-trained noise reduction model, and the target satellite signal is output.
3. The method according to claim 2, characterized in that The acquiring of the target hyperparameter corresponding to the target satellite signal comprises: Extracting a second target signal feature from the target satellite signal; The second target signal feature is input into a pre-built regression model, and the target hyperparameter is obtained as output.
4. The method according to claim 1, characterized in that: The propagator is: U im =I M×M -R(R H R+τI M×M ) -1 R H U im For the propagator, I M×M is the identity matrix, R is the covariance matrix, τ is the target hyperparameter, and H represents the Hermitian transpose.
5. The method according to claim 1, characterized in that The step of calculating the frequency matrix and the phase matrix of the target satellite signal according to the propagator comprises: constructing a spatial spectrum function according to the propagator; Traversing the spatial spectrum function, determining a spectrum peak of the spatial spectrum function, and determining a frequency matrix of the target satellite signal according to the spectrum peak; According to the frequency matrix of the target satellite signal, a pseudo-inverse matrix of the frequency matrix is calculated; Performing least squares on the pseudo-inverse matrix and the target satellite signal to obtain a target column vector; Solving the target column vector, a phase matrix of the target satellite signal is obtained.
6. The method according to claim 2, characterized in that The training steps of the denoising model include: Acquire original data and noise-free signal data; the original data is signal data that integrates Beidou satellite signals and background noise; Performing data preprocessing on the original data to obtain an original data sequence, and extracting original signal features from the original data sequence; Build a recurrent neural network model; The original signal features are used as the input of the recurrent neural network model, and the noise-free signal data is used as the output of the recurrent neural network model. The recurrent neural network model is trained to obtain the denoising model.
7. The method according to claim 1, characterized in that The calculating the covariance matrix of the target satellite signal comprises: Performing signal-frequency sampling on the target satellite signal to obtain a blind separation model corresponding to the target satellite signal; An autocorrelation matrix calculation operation is performed on the blind separation model to obtain a covariance matrix of the target satellite signal.
8. The method according to claim 3, characterized in that The training steps of the regression model include: Acquire an experimental data set, wherein the experimental data set includes sample signal characteristics of noise-free signal data and hyperparameter experimental values; Constructing an initial regression model, taking the sample signal features as input of the initial regression model, taking the hyperparameter experimental values as output of the initial regression model, and training the initial regression model using a support vector regression method to obtain optimal model parameters of the initial regression model; The regression model is constructed based on the optimal model parameters and the initial regression model.
9. The method according to claim 7, characterized in that: The blind separation model is: x=B·S+n x=[x(1),x(2),…x(M)] T x∈V M×1 B=[b1,b2,…b N ] B∈V M×N b k ∈V M×1 k∈{1,2,…,N} S∈V N×1 n=[n(1),n(2),…n(M)] T n∈V M×1 F k =f IF +f k Where V represents complex space, T represents transpose, and H represents Hermite transpose; x represents the BeiDou satellite signals collected by different channels at the same sampling point, M represents the number of sampling channels; tg represents the number of samples; A k represents the amplitude of the kth Beidou satellite signal received; θ k,0 represents the initial phase of the kth Beidou satellite signal received; P k represents the product of the ranging code and the navigation message code of the received k-th Beidou satellite signal; F k is the frequency of the kth BeiDou satellite signal; f IF is the intermediate frequency carrier frequency, f k and θ k,0 They represent the frequency and initial phase of the kth Beidou satellite signal collected by the receiver, n(t is the background noise at time t; The calculation formula of the covariance matrix of the target satellite signal is: R=E(x·x H )=B·E(S·S H )·B H +σ 2 I=B·R xx ·B H +σ 2 I Among them, R xx is the covariance matrix of the satellite signal sent by the BeiDou satellite, I is the identity matrix, E is the mathematical expectation, and R is the covariance matrix of the target satellite signal received by the receiver.
10. A Beidou satellite signal acquisition device, characterized in that: The device comprises A first acquisition module, used to acquire a target satellite signal; A first calculation module, used for calculating the covariance matrix of the target satellite signal; A second acquisition module is used to acquire the target hyperparameter corresponding to the target satellite signal; A construction module, configured to construct a propagator according to the covariance matrix and the target hyperparameter; The second calculation module is used to calculate the frequency matrix and phase matrix of the target satellite signal according to the propagator.
Citation Information
Patent Citations
Beidou satellite signal high-precision capture model based on blind separation of variance ratio
CN108983261A
Beidou satellite signal high-precision blind capture method based on wavelet transform
CN109975842A
Satellite communication anti-interference method based on improved propagation operator
CN116054909A
Medium-long code satellite navigation signal capturing method and device
CN118604853A
