Cycle slip signal generation method and system, electronic device and storage medium

Abstract

The application discloses a cycle slip signal generation method and system, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring carrier phase data; preprocessing the carrier phase data to obtain preprocessed data; performing wavelet decomposition processing on the preprocessed data to obtain a low-frequency signal and a first high-frequency signal; performing singular value interpolation processing on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal; performing construction processing on the second high-frequency signal according to a time-varying sequence generation formula to obtain a third high-frequency signal; and performing synthetic reconstruction processing on the third high-frequency signal and the low-frequency signal to generate a cycle slip signal. The embodiment of the application can quickly and efficiently generate a cycle slip signal, expands a data set in a cycle slip detection field based on deep learning, and can be widely applied to the technical field of satellite signal processing.

Description

Technical Field

[0001] This invention relates to the field of satellite signal processing technology, and in particular to a cycle slip signal generation method, system, electronic device, and storage medium. Background Technology

[0002] With the continuous development of navigation technology, people's requirements for positioning accuracy are constantly increasing. In the process of satellite navigation, the preprocessing of satellite data is particularly important for the reliability and accuracy of positioning. Among them, accurate and efficient cycle slip detection and repair is also a key issue in the data processing of high-precision global satellite navigation systems. Using deep learning for cycle slip detection has sufficient robustness and, after targeted training, can have the ability to detect small cycle slips and insensitive cycle slip pairs.

[0003] However, building deep learning neural networks often requires a large amount of data for training, and the data quality is crucial. Researchers find it difficult to determine whether cycle slips in all data have been detected or cleaned, making it hard to guarantee data quality. Furthermore, due to the inherent uncertainty in cycle slip generation, ensuring data consistency is challenging, adding significant difficulty to dataset construction. In summary, the technical problems in these related technologies urgently need to be addressed. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a cycle slip signal generation method, system, electronic device, and storage medium to improve the efficiency of dataset construction.

[0005] On one hand, the present invention provides a method for generating cycle slip signals, the method comprising:

[0006] Acquire carrier phase data;

[0007] The carrier phase data is preprocessed to obtain preprocessed data;

[0008] The preprocessed data is subjected to wavelet decomposition to obtain a low-frequency signal and a first high-frequency signal;

[0009] The first high-frequency signal is subjected to singular value interpolation based on a preset cycle slip value to obtain the second high-frequency signal;

[0010] The second high-frequency signal is processed using a time-varying sequence generative adversarial network to obtain the third high-frequency signal;

[0011] The third high-frequency signal and the low-frequency signal are combined and reconstructed to generate a cycle slip signal.

[0012] Optionally, the preprocessing of the carrier phase data to obtain preprocessed data includes:

[0013] The carrier phase data is filtered according to a preset threshold to obtain filtered data;

[0014] The filtered data is subjected to cycle slip repair processing to obtain repaired data;

[0015] The repaired data is subjected to consistency processing to obtain preprocessed data.

[0016] Optionally, performing wavelet decomposition on the preprocessed data to obtain a low-frequency signal and a first high-frequency signal includes:

[0017] Based on the preprocessed data, wavelet functions are selected and processed to obtain orthogonal biorthogonal wavelet functions;

[0018] The preprocessed data is decomposed according to the positive bi-orthogonal wavelet function to obtain a low-frequency signal and a first high-frequency signal.

[0019] Optionally, the step of performing singular value interpolation on the first high-frequency signal according to a preset cycle slip value to obtain the second high-frequency signal includes:

[0020] Get the first epoch of the week jump to be added;

[0021] The value of the second epoch in the first high-frequency signal is adjusted according to a preset cycle slip value to obtain a second high-frequency signal, wherein the value of the second epoch is half the value of the first epoch.

[0022] Optionally, the step of constructing and processing the second high-frequency signal based on the time-varying sequence generative adversarial network to obtain the third high-frequency signal includes:

[0023] The second high-frequency signal is extracted and processed to obtain a training dataset;

[0024] The training dataset is input into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal; the time-varying sequence generative adversarial network includes a generator, a discriminator, an embedding network, and a recovery network.

[0025] Optionally, the step of extracting and processing the second high-frequency signal to obtain a training dataset includes:

[0026] The second high-frequency signal is processed to extract its components, resulting in the first dataset;

[0027] The first dataset is subjected to time-series reversal processing and sorted in reverse chronological order to obtain the second dataset;

[0028] The second dataset is normalized to obtain the third dataset;

[0029] The data in the third dataset is segmented according to a preset sequence length, and then the segmented data is randomly arranged to obtain the training dataset.

[0030] Optionally, the step of inputting the training dataset into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal includes:

[0031] Obtain the random noise vector;

[0032] The training dataset is input into the embedding network and the recovery network for training;

[0033] The random noise vector is input into the generator and the discriminator for training.

[0034] The time-varying sequence generative adversarial network is trained based on a global optimization loss function to generate a third high-frequency signal.

[0035] On the other hand, embodiments of the present invention also provide a cycle slip signal generation system, comprising:

[0036] The first module is used to acquire carrier phase data;

[0037] The second module is used to preprocess the carrier phase data to obtain preprocessed data;

[0038] The third module is used to perform wavelet decomposition on the preprocessed data to obtain a low-frequency signal and a first high-frequency signal;

[0039] The fourth module is used to perform singular value interpolation on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal;

[0040] The fifth module is used to construct and process the second high-frequency signal based on the time-varying sequence generative adversarial network to obtain the third high-frequency signal;

[0041] The sixth module is used to synthesize and reconstruct the third high-frequency signal and the low-frequency signal to generate a cycle slip signal.

[0042] On the other hand, embodiments of the present invention also disclose an electronic device, including a processor and a memory;

[0043] The memory is used to store programs;

[0044] The processor executes the program to implement the method described above.

[0045] On the other hand, embodiments of the present invention also disclose a computer-readable storage medium storing a program that is executed by a processor to implement the methods described above.

[0046] In another aspect, the present application also discloses a computer program product or computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the foregoing method.

[0047] Compared with the prior art, the present application has the following technical effects: by wavelet decomposition processing on the pretreated data, low-frequency signals and first high-frequency signals are obtained, and the positioning accuracy of cycle slip signals is improved by wavelet transform for data decomposition and reconstruction; in addition, the second high-frequency signals are constructed and processed according to a time-varying sequence generation formula to obtain third high-frequency signals; the ability of the adversarial network to learn data distribution is used to generate synthetic signals similar to real cycle slip signals, expand the data set in the cycle slip detection field, and improve the data set construction efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Figure 1 is a flowchart of a cycle slip signal generation method provided by an embodiment of the present application;

[0050] Figure 2 is a schematic diagram of wavelet decomposition of data provided by an embodiment of the present application;

[0051] Figure 3 is a structure schematic diagram of a time-varying neural network provided by an embodiment of the present application;

[0052] Figure 4 is a training flowchart of a time-varying network provided by an embodiment of the present application;

[0053] Figure 5 is a schematic diagram of data combination reconstruction provided by an embodiment of the present application;

[0054] Figure 6 is a flowchart of a cycle slip signal generation method provided by an embodiment of the present application;

[0055] Figure 7 is a structure schematic diagram of a cycle slip signal generation system provided by an embodiment of the present application;

[0056] Figure 8 Fig. 1 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0058] Firstly, several terms involved in the present application are analyzed:

[0059] Cycle slips refer to the jump or interruption of whole number count caused by the loss of lock of satellite signals in the carrier phase measurement of global navigation satellite system (GNSS) technology. Correctly detecting and recovering cycle slips is one of the most important and necessary problems to be solved in carrier phase measurement.

[0060] Carrier phase measurement is a technology and method for measuring carrier phase observation values or differential observation values by a receiver, and calculating baseline vector coordinates difference between two synchronous observation stations through baseline vector calculation.

[0061] In the related art, cycle slip detection is performed through deep learning. However, the construction of a neural network for deep learning often requires a large amount of data for training, and has a high requirement for the quality of data, making it difficult to accurately construct a cycle slip signal data set.

[0062] Therefore, in the embodiments of the present application, a cycle slip signal generation method is provided. The cycle slip signal generation method in the embodiments of the present application can be applied in a terminal, can be applied in a server, and can also be software running in a terminal or a server, etc. The terminal can be a tablet computer, a notebook computer, a desktop computer, etc., but is not limited thereto. The server can be a standalone physical server, can be a server cluster or a distributed system composed of multiple physical servers, and can also be a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms.

[0063] With reference to Figure 1 The embodiments of the present application provide a cycle slip signal generation method, which comprises:

[0064] S101, acquiring carrier phase data;

[0065] S102, preprocessing the carrier phase data to obtain preprocessed data;

[0066] S103, wavelet decomposition processing is performed on the pretreated data to obtain a low-frequency signal and a first high-frequency signal;

[0067] S104, singular value interpolation processing is performed on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal;

[0068] S105, the second high-frequency signal is constructed by a time-varying sequence generative adversarial network to obtain a third high-frequency signal;

[0069] S106, the third high-frequency signal and the low-frequency signal are synthesized and reconstructed to generate a cycle slip signal.

[0070] Wherein, for the difficulty of collecting cycle slip signals and the like, the embodiment of the application provides a cycle slip signal synthesis method, by wavelet decomposition of the original signal to obtain a low-frequency signal and a first high-frequency signal, after adding a singular value to the first high-frequency signal part, a second high-frequency signal is obtained and sent to a time-varying sequence generative adversarial network for training, and then the third high-frequency signal obtained by synthesis and the original low-frequency signal are reconstructed, so as to realize the synthesis of the cycle slip signal, facilitate researchers to expand the data set, and help researchers to develop new cycle slip detection methods based on deep learning.

[0071] Further, in the step S101, carrier phase data is obtained.

[0072] Wherein, the carrier phase data of all satellites in the same sampling interval and observation time period can be obtained by a receiver or the like.

[0073] Further, in the step S102, the carrier phase data is pretreated to obtain pretreated data, including:

[0074] The carrier phase data is screened according to a preset threshold to obtain screened data;

[0075] The screened data is repaired for cycle slip to obtain repaired data;

[0076] The repaired data is processed for consistency to obtain pretreated data.

[0077] In the embodiment of the present application, first, the carrier phase data is filtered according to a preset threshold, wherein the preset threshold can be set to 10° to 20° according to experience, and satellites with a low elevation angle are removed according to the elevation angle of the satellite in the obtained time period. The carrier phase data below the preset threshold is filtered and removed to obtain filtered data. Then, the cycle slip of the filtered data is repaired by performing cycle slip repair processing on each satellite carrier phase sequence in the filtered data to ensure that there is no cycle slip in the original sequence for subsequent difference value interpolation processing. The embodiment of the present application can select data obtained by a high-precision station as original carrier phase data, perform cycle slip detection and repair processing on the data by using a forward and backward moving window filtering algorithm (FBMWA) and other cycle slip post-processing methods for pre-processing the data, thereby ensuring that there is no cycle slip in the sequence, improving the accuracy and reliability of the data, and avoiding the negative impact of data errors on the performance of the cycle slip signal. Finally, the repaired data is processed for consistency. Let the carrier phase observation value of satellite i (i = 1, 2, 3,...) be with a length of N, and the initial value is processed to zero, as shown in the following formula:

[0078]

[0079] The initial value of the repaired data is processed to zero to ensure the consistency of the data.

[0080] Further, as an optional implementation, in the step S103, the wavelet decomposition processing on the pre-processed data to obtain a low-frequency signal and a first high-frequency signal comprises:

[0081] According to the pre-processed data, a wavelet function is selected to obtain a biorthogonal wavelet function;

[0082] According to the biorthogonal wavelet function, the pre-processed data is decomposed to obtain a low-frequency signal and a first high-frequency signal.

[0083] In the embodiment of the present application, since the wavelet method has the advantage of accurate positioning, the original signal is decomposed by wavelet and then the cycle slip is added, which has higher accuracy than directly adding the cycle slip in the original signal. There is a difference between the signal decomposition capabilities of different wavelet functions, and the selection of a suitable wavelet for the addition of cycle slip is of great significance. The Bior3.1 wavelet is a kind of approximately symmetric wavelet function, which is a biorthogonal wavelet. Two wavelets are used for signal decomposition and reconstruction, respectively, and its decay speed is faster than other wavelets. Among them, 3 represents the reconstruction length, and 1 represents the decomposition length. In the embodiment of the present application, the pre-processed data is subjected to Bior3.1 wavelet transform to decompose into low-frequency and high-frequency two sub-signals. Referring to Figure 2When the signal passes through the high-pass filter, a first high-frequency signal is obtained, and the instantaneous change of the high-frequency signal is more obvious, and the change part frequency does not cause a great influence on the whole signal. Therefore, by using this characteristic, the singular value interpolation can be carried out on the basis of the high-frequency signal, so that the reconstructed signal reflects the cycle slip on the original signal. If a small cycle slip is directly added to the original signal, that is, the preprocessed data, the fitting may not be good, that is, the synthesized cycle slip epoch is biased or the number of synthesized cycle slips cannot meet the requirements, so the feature extraction needs to be carried out on the preprocessing, so that the part that is easy to be recognized by the network is exposed. The wavelet transform is essentially a feature extraction process. Therefore, in the embodiment of the application, the preprocessed data is decomposed based on the wavelet transform to obtain a low-frequency signal and a first high-frequency signal, so that the cycle slip can be accurately positioned, and a basis for subsequent cycle slip interpolation processing of the first high-frequency signal is provided.

[0084] Further, as an optional implementation, in the step S104, the singular value interpolation processing of the first high-frequency signal according to the preset cycle slip value to obtain a second high-frequency signal comprises:

[0085] A first epoch to which a cycle slip is to be added is obtained.

[0086] The value of the second epoch in the first high-frequency signal is adjusted according to the preset cycle slip value to obtain a second high-frequency signal, and the value of the second epoch is half of the value of the first epoch.

[0087] In the embodiment of the application, first, the epoch of the first high-frequency signal to which a cycle slip is to be added is determined. If a cycle slip ε is to be added to the nth epoch in the first high-frequency signal, since each wavelet decomposition will make the signal length of the next layer halved, the value of the second epoch needs to be adjusted, that is, by replacing the value of the n / 2th epoch in the high-frequency signal of the first layer, specifically, the modulus value of the n / 2th epoch and the (n+1) / 2th epoch is added to obtain the cycle slip value ε / 2 to be added, and this process is called singular value interpolation of the first high-frequency signal. In the signal reconstruction stage, the interpolated value will be mapped to show the properties of the cycle slip. By performing the singular value interpolation on the first high-frequency signal, the cycle slip information can be added to the signal, thereby providing a basis for generating a data set for subsequent training in the neural network.

[0088] Further, as an optional implementation, in the step S105, the second high-frequency signal is constructed by using the time-varying sequence generative adversarial network to obtain a third high-frequency signal, comprising:

[0089] The second high-frequency signal is extracted to obtain a training data set.

[0090] Input the training data set into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal; the time-varying sequence generative adversarial network comprises a generator, a discriminator, an embedding network and a restoration network.

[0091] In the embodiment of the application, the second high-frequency signal is extracted to construct a training data set, which is used to input the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal, which is a cycle slip data generated by a neural network and having similar characteristics to the second high-frequency signal. The traditional generative adversarial network (GAN) is composed of a generator and a discriminator. The generator is responsible for generating synthetic data samples similar to real data, while the discriminator is responsible for judging whether the input data is real data or data generated by the generator. Through adversarial training, the generator and the discriminator compete with each other, and finally the generator can generate realistic data samples. However, the traditional GAN does not perform well in processing time series, and cannot add labels to data. Therefore, the embodiment of the application adopts a time-varying sequence generative adversarial network (TIMEGAN), which is a time series data generation model used for data analysis in the fields of finance, meteorology, etc. The TIMEGAN in the embodiment of the application learns the time dependence and data distribution of time series data through adversarial training by training different networks in the model respectively. The input time series is encoded to a hidden layer by a neural network, and then the hidden representation is decoded by a decoder into synthetic time series data having a similar distribution to the original time series data. Referring to Figure 3 In addition to the generator and the discriminator of the traditional GAN, the TIMEGAN in the embodiment of the application additionally introduces two new networks: an embedding network and a restoration network. The design purpose of the time-varying sequence generative adversarial network is to capture the characteristics of real data and the complex dynamics of these characteristics over time. The newly introduced embedding network provides a reversible mapping between features and latent representations, and the embedding network and the restoration network maintain the relationship between latent vectors and features in the latent space, while the generator and the discriminator network remain unchanged. The embodiment of the application generates synthetic cycle slip signals that are very similar to real cycle slip signals according to the labels by using the TIMEGAN combined with the method of conditional labels, which facilitates researchers to expand the data set, and helps researchers to develop new cycle slip detection methods based on deep learning.

[0092] Further, as an optional implementation, the extracting processing on the second high-frequency signal to obtain the training data set comprises:

[0093] The second high-frequency signal is extracted to obtain a first data set;

[0094] performing time reversal processing on the first data set and arranging in time reverse order to obtain a second data set;

[0095] performing normalization processing on the second data set to obtain a third data set;

[0096] performing cutting processing on data in the third data set according to a preset sequence length, and then performing random arrangement on the cut data to obtain a training data set.

[0097] In the embodiment of the present application, the changed high-frequency signal is extracted to form a first data set. The data in the first data set is time-reversed and arranged in time reverse order to obtain a second data set. The MinMaxScaler function is used to normalize the data of the second data set, and all values are scaled to the range of [0, 1] to obtain a third data set. Finally, the data in the third data set is cut at a set sequence length, and then randomly arranged to obtain a training data set. The training data set is used to input the time-varying sequence generative adversarial network for data generation processing. It should be noted that, in addition to constructing the training data set itself, the network parameters such as sequence length, dimension, hidden dimension, neural network layer number, training batch, and iteration number are also added to the feedforward to realize the optimization of the model training. In a feasible implementation manner, the number of hidden dimensions of the time-varying sequence generative adversarial network can be set to 24, the number of neural network layers can be set to 3, and the number of iterations can be set to 10,000 times.

[0098] Further, as an optional implementation manner, the inputting the training data set into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal comprises:

[0099] obtaining a random noise vector;

[0100] inputting the training data set into the embedding network and the restoration network for training;

[0101] inputting the random noise vector into the generator and the discriminator for training;

[0102] training the time-varying sequence generative adversarial network based on a global optimization loss function to generate a third high-frequency signal.

[0103] In the embodiment of the present application, the time-varying sequence generative adversarial network is trained based on the training data set, and the time-varying sequence generative adversarial network comprises an embedding network, a restoration network, a generator, and a discriminator. Figure 4, where s is the static feature vector for different satellite scenes and x is its temporal property. The supervised loss reflects how well the generated data approximates the data produced by the autoencoder. In calculating the supervised loss, the model captures the conditional distribution of data over time through the supervision of the original data. The model simultaneously generates static and temporal data and passes it to the embedding network, which provides assistance in calculating the next time data in the latent space. For the embedding network in the figure, it is essentially a convolutional encoding layer that maps the original high-dimensional features into a low-dimensional space, represented as:

[0104]

[0105] where s represents the static feature vector, h s represents the mapped low-dimensional static feature vector, h t-1 represents the static feature vector at time t-1, x t represents the high-dimensional dynamic feature vector at time t, and l represents the conditional label, implemented by the recurrent network e.

[0106] The reconstruction network attempts to reconstruct the original high-dimensional static and dynamic features from the low-dimensional feature vector, represented as:

[0107]

[0108] where, represents the reconstructed high-dimensional static feature vector, represents the reconstructed high-dimensional dynamic feature vector at time t, and the function r is implemented by a bidirectional recurrent network with a feedforward output layer.

[0109] The joint loss function of the embedding network and the reconstruction network is:

[0110]

[0111] This is also Figure 4 the reconstruction loss, reflecting the loss of the autoencoder part. It refers to the degree of mastery of the intrinsic pattern of the input data by the autoencoder, that is, the encoding level of the autoencoder is tested to score.

[0112] For the generator, since the conditional label is added, it can be represented as:

[0113]

[0114] where z s and z t are random samples from two vector spaces, are the generated static and dynamic feature vectors, respectively, and l represents the conditional label, implemented by the recurrent network function g.

[0115] For the discriminator, also because of the addition of conditional labels, it is expressed as:

[0116]

[0117] wherein, represents the static feature of the data to generate the classification results, represents the dynamic feature of the data to generate the classification results at time t, is the joint encoding output result, and l' is the conditional label to be generated.

[0118] In general, in the generation network, gradient ascent is used to find the local maximum, while in the part of the discriminator network, gradient descent is used to obtain the global minimum, and this process is trained in cycles according to the number of iterations until the model learns the cycle signal and the conditional label carried by it, and generates synthetic data with similar characteristics to the actual data.

[0119] The joint loss function of the generator and the discriminator is defined as:

[0120]

[0121] is also Figure 2 The unsupervised loss in the above equation only relies on It is not enough to encourage the generator to capture the distribution of the conditional label, so an additional supervised loss is introduced to let the generator accept and Alternating training, the supervised loss is defined as:

[0122]

[0123] In order to let the conditional TimeGAN capture the joint probability distribution of the real data set as accurately as possible from the data set, so that the probability distribution of the generated data sequence can be as close as possible to the conditional TimeGAN, the optimization objective of local training is also designed in TimeGAN.

[0124] Let θe, θr, θg and θd be the weight parameter sets of the embedding network, the recovery network, the generator network and the discriminator network respectively, and the global optimization loss function of TIMEGAN is:

[0125]

[0126] Where λ and η are manually set hyperparameters.

[0127] It should be noted that the RMSProp optimizer is used in the embodiment of the present application instead of the Adam optimizer, so that the model has better convergence and generalization ability, can effectively solve the problem of model collapse, and the RMSProp has fewer hyperparameters, so that the parameter adjustment in the model training process can be simplified. In a feasible implementation manner, the learning rate of the optimizer can be set to 0.001, the decay rate is 0.9, and the momentum is 0.1.

[0128] In the step S106, the third high-frequency signal and the low-frequency signal are combined and reconstructed

[0129] The cycle slip signal is generated.

[0130] The cycle slip signal is generated. Figure 5 Since the low-frequency signal is obtained by wavelet decomposition and reconstruction of the original signal, and the high-frequency signal is a synthetic high-frequency signal obtained by the TIMEGAN network, the synthetic cycle slip signal can be obtained after combining and reconstructing them. Since the carrier phase sequence without cycle slip is relatively smooth, and the attenuation speed of the bior wavelet is also fast, the wavelet reconstruction basically does not lose information, and the signal reconstruction can be realized, that is, the carrier phase signal, and the high-frequency signal is mapped to the cycle slip, thereby generating the cycle slip signal.

[0131] The cycle slip signal is generated. Figure 6 In a feasible embodiment of the present application, first, the original carrier phase data is obtained, and the original carrier phase data is decomposed by wavelet transform to obtain high-frequency signals and low-frequency signals. Then, the epoch value of the high-frequency signal is changed by the singular value difference method and input into the TIMEGAN network for training. The training process of the network is that the embedding network and the recovery network are trained separately, then the generator and the discriminator are trained separately, and finally the four networks are jointly trained. The synthetic high-frequency signal is generated by the TIMEGAN network, and the original low-frequency signal and the synthetic high-frequency signal are reconstructed. The generated signal is evaluated according to the index, and the cycle slip signal is generated, so as to enrich the data set. It should be noted that the present application is not only suitable for the field of satellite signal processing, but also can be used for other fields if it is necessary to generate abnormal signals. According to the idea of the present application, the original signal is decomposed by wavelet, the singular value interpolation is performed on the wavelet scale, and then the wavelet reconstruction is performed, so as to accurately add abnormal values to the epoch of the original signal.

[0132] In the embodiment of the present application, the vector sequence in the original data is marked as 0.9, and the vector sequence of the synthetic data is marked as 0. The reason why the original data is not marked as 1 in the discriminator is to smooth the label and not to encourage the discriminator to be too confident in the execution of the classification task. For the TIMEGAN model, if the generated data is equivalent to the original data, the discrimination accuracy of the model should be low. That is, if the model can generate data that is almost indistinguishable from the original data, the values of each performance indicator should be close to 0, not close to 0.9. Therefore, the lower the value of all indicators, the better the performance of the TIMEGAN model. Finally, the discriminator score, the prediction score, the accuracy, the precision, the recall and the F1 score are used to evaluate the performance of the model. The discriminator score and the prediction score describe the MAE value of the original signal and the generated cycle slip signal, and the accuracy, the precision, the recall and the F1 score describe the similarity index between the true signal and the generated signal. The step-by-step training of the network introduced by TIMEGAN can significantly improve the performance.

[0133] With reference to Figure 7 The embodiment of the present application also provides a cycle slip signal generation system, comprising:

[0134] A first module 701 is configured to acquire carrier phase data.

[0135] A second module 702 is configured to pre-process the carrier phase data to obtain pre-processed data.

[0136] A third module 703 is configured to perform wavelet decomposition processing on the pre-processed data to obtain a low-frequency signal and a first high-frequency signal.

[0137] A fourth module 704 is configured to perform singular value interpolation processing on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal.

[0138] A fifth module 705 is configured to perform construction processing on the second high-frequency signal according to a time-varying sequence generative adversarial network to obtain a third high-frequency signal.

[0139] A sixth module 706 is configured to perform synthetic reconstruction processing on the third high-frequency signal and the low-frequency signal to generate a cycle slip signal.

[0140] It can be understood that the contents in the above method embodiments are all applicable to the present system embodiment, the present system embodiment specifically realizes the same functions as the above method embodiments, and achieves the same beneficial effects as the above method embodiments.

[0141] With reference to Figure 8The embodiment of the present application also provides an electronic device, comprising a processor 801 and a memory 802; the memory 802 is used for storing a program; and the processor 801 executes the program to realize the method as described above.

[0142] Corresponding to the method of Figure 1 The embodiment of the present application also provides a computer readable storage medium, which stores a program, and the program is executed by a processor to realize the method as described above.

[0143] The embodiment of the present application also discloses a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method as shown in the figure. Figure 1

[0144] To sum up, the embodiment of the present application has the following advantages: the embodiment of the present application utilizes the high positioning accuracy of wavelet transform, and after selecting a specific wavelet, the synthesis of cycle slip signals can be completed, and the robustness is strong. In addition, the embodiment of the present application utilizes the ability of the adversarial network to learn data distribution, and after selecting network parameters, a network structure suitable for cycle slip signal generation, an optimizer and the like are designed, a TIMEGAN-based adversarial neural network model is constructed and trained, a synthetic signal similar to the real cycle slip signal can be generated, the data set of the cycle slip detection field based on deep learning can be effectively expanded, and great convenience is brought to researchers in the field.

[0145] In some alternative embodiments, the functions / operations mentioned in the block diagram can not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two blocks shown in succession can actually be executed substantially simultaneously or the blocks can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flowcharts of the present application are provided by way of example, and the purpose is to provide a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and in which sub-operations described as part of larger operations are independently executed.

[0146] ​Furthermore, although the present application is described in the context of functional modules, it is to be understood that one or more of the described functions and / or features can be integrated in a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It will also be appreciated that detailed discussion of the actual implementation of each module is not necessary to an understanding of the application. Rather, the actual implementation of the modules, in combination with their attributes, functions, and internal relationships, are to be understood within the context of the devices disclosed herein. Thus, those skilled in the art with access to patents, scientific journals, and other public sources known by those skilled in the art will be able to practice the application as set forth in the claims without undue experimentation, using ordinary skill in the art along with the present disclosure. It is also to be understood that the specific concepts disclosed are merely illustrative and that the scope of the present application is to be determined by the entire scope of the claims, along with all equivalents of the claims and their equivalents.

[0147] If the functions are implemented in software, the functions can be stored in or implemented as one or more computer program products, which can be incorporated into a computer-readable medium for use by or in connection with an apparatus, method or system as described herein. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, or a computer-readable signal. The computer-readable medium can be, for example, but is not limited to, volatile or non-volatile memory, a floppy diskette, a compact disc read-only memory (CD-ROM), a magnetic tape, a flash drive, a hard disk drive, a programmable read-only memory (PROM), a random access memory (RAM), a reprogrammable read-only memory (REPROM) or a similar medium. The computer-readable medium can be paper, on which the program is printed. The computer-readable medium can be a medium, such as RAM, ROM, PROM, a tape, a hard disk or a compact disc, which can store data for use by or in connection with an application program. The computer-readable medium can also be a medium that carries such data also, or a medium that merges, for example, for use in a transmission, such as an electrical, optical or electromagnetic transmission. The computer-readable medium can also be a medium that can store data for use by or in connection with a computer program executing on a computer. The computer-readable medium can also be paper, on which the program is printed.

[0148] The logic and / or steps represented in the flowcharts and / or otherwise described herein, for example, can be considered as a sequence of instructions executed in an order to achieve a logic function, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device.

[0149] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0150] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following techniques can be used to implement the hardware used in the described embodiments: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and so forth.

[0151] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples.

[0152] Although embodiments of the present application have been shown and described, it would be recognized by those of ordinary skill in the art that various changes, modifications, substitutions, and alterations can be made hereto without departing from the principles and spirit of the application, and it is intended that the scope of the application be limited only by the claims and equivalents thereof.

[0153] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the described embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the present application, and these equivalent modifications or replacements are all included in the scope defined by the claims of the present application.

Claims

1. A method for generating cycle slip signals, characterized in that, The method includes: Acquire carrier phase data; The carrier phase data is preprocessed to obtain preprocessed data; The preprocessed data is subjected to wavelet decomposition to obtain a low-frequency signal and a first high-frequency signal; The first high-frequency signal is subjected to singular value interpolation based on a preset cycle slip value to obtain the second high-frequency signal; The second high-frequency signal is processed using a time-varying sequence generative adversarial network to obtain the third high-frequency signal; The third high-frequency signal and the low-frequency signal are combined and reconstructed to generate a cycle slip signal; The step of performing singular value interpolation on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal includes: Get the first epoch of the week jump to be added; The value of the second epoch in the first high-frequency signal is adjusted according to the preset cycle slip value to obtain the second high-frequency signal, wherein the value of the second epoch is half of the value of the first epoch. The process of constructing and processing the second high-frequency signal based on a time-varying sequence generative adversarial network to obtain a third high-frequency signal includes: The second high-frequency signal is extracted and processed to obtain a training dataset; The training dataset is input into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal; the time-varying sequence generative adversarial network includes a generator, a discriminator, an embedding network, and a recovery network; The extraction and processing of the second high-frequency signal to obtain the training dataset includes: The second high-frequency signal is processed to extract its components, resulting in the first dataset; The first dataset is subjected to time-series reversal processing and sorted in reverse chronological order to obtain the second dataset; The second dataset is normalized to obtain the third dataset; The data in the third dataset is segmented according to a preset sequence length, and then the segmented data is randomly arranged to obtain the training dataset. The step of inputting the training dataset into the time-varying sequence generative adversarial network for data generation processing to obtain a third high-frequency signal includes: Obtain the random noise vector; The training dataset is input into the embedding network and the recovery network for training. The random noise vector is input into the generator and the discriminator for training. The time-varying sequence generative adversarial network is trained based on a global optimization loss function to generate a third high-frequency signal.

2. The method according to claim 1, characterized in that, The preprocessing of the carrier phase data to obtain preprocessed data includes: The carrier phase data is filtered according to a preset threshold to obtain filtered data; The filtered data is subjected to cycle slip repair processing to obtain repaired data; The repaired data is subjected to consistency processing to obtain preprocessed data.

3. The method according to claim 1, characterized in that, The step of performing wavelet decomposition on the preprocessed data to obtain a low-frequency signal and a first high-frequency signal includes: Based on the preprocessed data, wavelet functions are selected and processed to obtain orthogonal biorthogonal wavelet functions; The preprocessed data is decomposed according to the positive bi-orthogonal wavelet function to obtain a low-frequency signal and a first high-frequency signal.

4. A cycle slip signal generation system, said system being applied to the cycle slip signal generation method as described in claim 1, characterized in that, The system includes: The first module is used to acquire carrier phase data; The second module is used to preprocess the carrier phase data to obtain preprocessed data; The third module is used to perform wavelet decomposition on the preprocessed data to obtain a low-frequency signal and a first high-frequency signal; The fourth module is used to perform singular value interpolation on the first high-frequency signal according to a preset cycle slip value to obtain a second high-frequency signal; The fifth module is used to construct and process the second high-frequency signal based on the time-varying sequence generative adversarial network to obtain the third high-frequency signal; The sixth module is used to synthesize and reconstruct the third high-frequency signal and the low-frequency signal to generate a cycle slip signal.

5. An electronic device, characterized in that, The electronic device includes a memory and a processor; The memory is used to store programs; The processor executes the program to implement the method of any one of claims 1 to 3.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 3.

Images