Satellite same-frequency interference cancellation method based on deep neural network

Abstract

The invention belongs to the technical field of navigation signal processing, and particularly discloses an on-satellite same-frequency interference cancellation method based on a deep neural network, and the method comprises the steps: obtaining a preprocessed input signal sequence which comprises an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; inputting the input signal sequence into a DNN-BiLSTM model, and reconstructing a target interference signal in the combined signal; the target interference signal is a part of signal overlapped with the receiving frequency band of the GNSS receiver in the ideal enhanced signal transmitted by the low-orbit satellite transmitter; and carrying out interference cancellation on the combined signal received by the GNSS receiver and the reconstructed target interference signal to obtain a received signal of the GNSS receiver after interference cancellation. According to the invention, the ideal enhanced signal transmitted by the low-orbit satellite transmitter is used as the prior information, the channel characteristics are subjected to blind estimation by using the DNN-BiLSTM model in combination with the combined signal received by the GNSS receiver, and the same-frequency interference elimination is realized.

Description

Technical Field

[0001] The present application belongs to the field of navigation signal processing technology, and more specifically, to a method for canceling co-frequency interference on a satellite based on a deep neural network. Background Art

[0002] The Global Navigation Satellite System (GNSS) can provide positioning, navigation and timing services worldwide. The current GNSS system has limitations such as weak landing signal strength and low information transmission rate. To address these limitations, low-orbit satellites can complement the advantages of medium- and high-orbit satellites, and as an enhancement and supplement to the GNSS system, improve the accuracy, availability and integrity of satellite navigation positioning.

[0003] However, some frequency bands of the transmitting end of the low-orbit navigation augmentation system overlap with the GNSS frequency band, which will interfere with the received GNSS signal and affect the signal reception accuracy. Therefore, it is necessary to introduce interference elimination technology to effectively suppress or eliminate the interference of the self-transmitted signal through signal processing algorithms or antenna design to ensure that low-orbit satellites can effectively use GNSS signals.

[0004] Digital domain interference cancellation technology has been widely used in the field of co-band interference suppression. However, when dealing with signal interference in high dynamic environments, traditional digital domain interference cancellation technology still has shortcomings in real-time and nonlinear interference modeling. Not only is the implementation method complex, but the anti-co-band interference performance is also limited. Summary of the invention

[0005] In view of the above-mentioned defects in the prior art, the present application provides an on-board co-frequency interference cancellation method based on deep neural network, aiming to solve the problem of co-frequency interference suppression in low-orbit satellite navigation systems.

[0006] In a first aspect, the present application provides a method for canceling co-frequency interference on a satellite based on a deep neural network, comprising: Acquire a preprocessed input signal sequence, where the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; The input signal sequence is input into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; the target interference signal is the part of the ideal enhanced signal emitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver; The combined signal received by the GNSS receiver is subtracted from the reconstructed target interference signal to perform interference cancellation, and the GNSS receiver receiving signal after interference cancellation is obtained.

[0007] In a second aspect, the present application also provides an on-board co-frequency interference cancellation device based on a deep neural network, comprising: An acquisition module is used to acquire a preprocessed input signal sequence, where the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; The reconstruction module is used to input the input signal sequence into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; the target interference signal is the part of the ideal enhanced signal emitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver; The cancellation module is used to perform interference cancellation by subtracting the combined signal received by the GNSS receiver from the reconstructed target interference signal to obtain the GNSS receiver receiving signal after the interference cancellation.

[0008] In a third aspect, the present application also provides an electronic device, comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0009] In a fourth aspect, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0010] In a fifth aspect, the present application further provides a computer program product, which, when executed on a processor, enables the processor to execute the method described in the first aspect or any possible implementation manner of the first aspect.

[0011] The on-board co-channel interference cancellation method based on deep neural network provided in the present application uses the ideal enhanced signal transmitted by the low-orbit satellite transmitter as prior information, combines the combined signal received by the GNSS receiver, and uses the DNN-BiLSTM model to blindly estimate the channel characteristics, thereby reconstructing the target interference signal in the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver, thereby realizing the co-channel interference cancellation under rapidly changing and complex low-orbit satellite navigation signal environment conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or related technologies, the following is a brief introduction to the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 It is a flowchart of a method for canceling co-frequency interference on a satellite based on a deep neural network provided in an embodiment of the present application; Figure 2 is a structural diagram of a sliding window input model provided in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of the DNN-BiLSTM model provided in the embodiment of the present application; Figure 4 This is a diagram of an enhanced co-frequency transceiver model of a low-orbit satellite navigation platform provided in an embodiment of the present application; Figure 5 is a diagram of the radio frequency channel baseband equivalent model provided in an embodiment of the present application; Figure 6 It is a structural schematic diagram of an on-board co-frequency interference cancellation device based on a deep neural network provided in an embodiment of the present application; Figure 7 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0015] A deep neural network (DNN) is a multi-layer neural network consisting of multiple hidden layers, each containing multiple neurons; DNN can extract deep features of input data. Bi-directional long short-term memory (BiLSTM) is a special recurrent neural network that combines the characteristics of long short-term memory networks with the ability to process bidirectional data, and has significant advantages in processing sequence data.

[0016] Based on this, the embodiment of the present application proposes a DNN-BiLSTM model, which combines the advantages of DNN and BiLSTM, has powerful feature extraction and timing processing capabilities, can effectively realize the on-board co-frequency interference cancellation of the low-orbit satellite navigation system, and improve the system's anti-interference capability and positioning accuracy.

[0017] Figure 1 is a flow chart of a method for canceling co-frequency interference on a satellite based on a deep neural network provided in an embodiment of the present application, such as Figure 1 As shown, the method comprises at least the following steps: S101. Acquire a preprocessed input signal sequence, where the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver.

[0018] Specifically, the ideal enhanced signal sequence transmitted by the preprocessed low-orbit satellite transmitter and the combined signal sequence received by the GNSS receiver are obtained as the input signal sequence of the DNN-BiLSTM model.

[0019] Optionally, the preprocessing method includes distinguishing the real part and the imaginary part, denoising, filtering, etc., to improve the quality and reliability of the signal.

[0020] Optionally, S101 specifically includes: The ideal enhanced signal transmitted by the low-orbit satellite transmitter is sampled in two ways, I / Q, to obtain the real component sequence and imaginary component sequence of the ideal enhanced signal; the combined signal received by the GNSS receiver is sampled in two ways, I / Q, to obtain the real component sequence and imaginary component sequence of the combined signal.

[0021] Specifically, signals are generally represented in complex form, with the real and imaginary parts corresponding to the amplitude and phase information of the signal, respectively. When using neural networks to process signals, it is necessary to extract useful features from the input signal and input the real and imaginary parts of the signal into the neural network, respectively. The two parts can be processed independently to improve the precision and accuracy of signal processing; at the same time, due to the different characteristics of the real and imaginary parts of the signal in the frequency domain, merging them into real signals and directly inputting them into the neural network will lead to information redundancy and confusion, affecting the performance of the neural network.

[0022] Therefore, when obtaining the input signal sequence, the ideal enhanced signal transmitted by the low-orbit satellite transmitter and the combined signal received by the GNSS receiver are sampled in two ways, I / Q, to distinguish the real component signal and the imaginary component signal, and use them as the input data of the DNN-BiLSTM model at the same time. That is, the DNN-BiLSTM model contains 4 groups of signal component sequences as input at the same time.

[0023] S102, input the input signal sequence into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; the target interference signal is a portion of the ideal enhanced signal emitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver.

[0024] Specifically, the input signal sequence is input into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal. The target interference signal is the part of the ideal enhanced signal emitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver, so it can also be called a self-interference signal.

[0025] The DNN-BiLSTM model uses the ideal enhanced signal transmitted by the input low-orbit satellite transmitter as prior information, and combines the combined signal received by the GNSS receiver to blindly estimate the channel characteristics to achieve reconstruction of the target interference signal. It is adaptable to rapidly changing and complex signal environments, and is particularly suitable for application scenarios where channel parameters are unknown. The model output is the reconstructed target interference signal sequence, that is, the component estimation value of the part of the signal coupled to the GNSS receiver in the ideal enhanced signal transmitted by the low-orbit satellite transmitter, including the component estimation value of the I / Q path.

[0026] Optionally, in S102, inputting the input signal sequence into the DNN-BiLSTM model includes: The input signal sequence is obtained by segmenting a sliding window of a fixed size and input into the DNN-BiLSTM model in the form of a short signal sequence.

[0027] Figure 2 is a structural diagram of a sliding window input model provided in an embodiment of the present application, such as Figure 2 As shown in the figure, a fixed-size sliding window is used to segment the acquired input signal sequence, and the segmentation into short signal sequences is used as a DNN-BiLSTM model, which can enhance the model's extraction of local signal features.

[0028] Specifically, by adjusting the window size and step size, the real component sequence and imaginary component sequence of the ideal enhanced signal transmitted by the low-orbit satellite transmitter, as well as the real signal component sequence and imaginary component sequence of the combined signal received by the GNSS receiver, are divided into four groups of short signal sequences of the same length, and then input into the DNN-BiLSTM model.

[0029] The selection of sliding window size requires a trade-off between model complexity and performance. A window that is too small may not capture enough time series dependencies, while a window that is too large will increase the amount of computation and the risk of overfitting. Optionally, the signal sequence is normalized and standardized to be converted into a format suitable for model input before being input into the model.

[0030] S103: Subtract the combined signal received by the GNSS receiver from the reconstructed target interference signal to perform interference cancellation, and obtain a GNSS receiver receiving signal after the interference cancellation.

[0031] Specifically, the interference cancellation can be achieved by separating the target interference signal reconstructed by the DNN-BiLSTM model from the combined signal received by the GNSS receiver. and reconstructed target jammer signal Subtract, satisfy:

[0032] in, This is the signal received by the GNSS receiver after interference cancellation.

[0033] The on-board co-channel interference cancellation method based on deep neural network provided in the embodiment of the present application uses the ideal enhanced signal transmitted by the low-orbit satellite transmitter as prior information, combined with the combined signal received by the GNSS receiver, and uses the DNN-BiLSTM model to blindly estimate the channel characteristics, and reconstruct the target interference signal in the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver, thereby realizing the co-channel interference cancellation under rapidly changing and complex low-orbit satellite navigation signal environment conditions.

[0034] It can be imagined that the embodiment of the present application only performs interference cancellation on part of the interference signals in the combined signal received by the GNSS receiver, that is, part of the signal in the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver. The GNSS receiver received signal after interference cancellation obtained by the technical solution provided in the embodiment of the present application may also contain other interference signals and noise signals, etc. The interference cancellation method for this part of the interference signal is not limited in the embodiment of the present application.

[0035] The structure of the DNN-BiLSTM model is introduced below.

[0036] Figure 3 : is a schematic diagram of the structure of the DNN-BiLSTM model provided in the embodiment of the present application, such as Figure 3 As shown in the figure, the DNN-BiLSTM model includes an input layer, a BiLSTM module, a DNN module and an output layer.

[0037] The input layer has a total of 2N nodes, N=256. The sliding window technology is used to slide the input signal sequence. Each time the input takes the N / 2 sampling points before the current time t and the N / 2 sampling points after the current time t to form the input signal sequence at the current time t. The input signal sequence contains the real component sequence and the imaginary component sequence of the signal. This allows the model to focus on a certain number of data points before and after the current time when processing the ideal enhanced signal transmitted by the low-orbit satellite transmitter and the combined signal received by the GNSS receiver. This helps the model obtain the historical information and future trends of the signal and understand the dynamic changes of the signal.

[0038] The BiLSTM module includes a forward LSTM layer (LSTM L ) and the backward LSTM layer (LSTM R), each LSTM layer includes 128 neurons. The forward LSTM layer processes the input signal sequence forward, from the first time step to the last time step, and the backward LSTM layer processes the input signal sequence backward, from the last time step to the first time step, to capture the temporal dependencies in the input signal sequence. The outputs of the forward LSTM layer and the backward LSTM layer are concatenated together at each time step to form a hidden state vector containing complete context information, which is then passed to the next layer. This bidirectional processing can make full use of the historical information and future trends of the input signal sequence, and improve the model's ability to process complex time series signals.

[0039] The DNN module includes a Dropout layer and two fully connected (FC) layers. The BiLSTM module has a large number of parameters and is prone to overfitting. The Dropout layer uses regularization technology to randomly set the neuron output of the BiLSTM module to 0 with a probability of 0.1, so that the model will not rely on certain specific neurons during training, thereby enhancing the generalization ability of the model and preventing overfitting.

[0040] The number of neurons in the two fully connected layers is 500 and 1 respectively. The fully connected layers extract the signal features processed in the BiLSTM layer through weighted and activated functions, and map the extracted features to outputs. Optionally, the activation function used in the two fully connected layers is the Leaky ReLU function to enhance the ability of the deep learning network to eliminate nonlinear interference of the input signal.

[0041] The number of nodes in the output layer is 2, and its output is the component estimate of the partial signal coupled to the GNSS receiver in the ideal enhanced signal transmitted by the low-orbit satellite transmitter. The reconstructed target interference signal is then eliminated from the combined signal to complete the self-interference cancellation process of the entire signal.

[0042] The DNN-BiLSTM model constructed in the embodiment of the present application can adapt to various types of channels through blind estimation of the channel, adopts sliding window input technology, and uses BiLSTM to assist DNN in completing the task of signal interference elimination, which can not only ensure the excellent feature extraction capability of DNN, but also bring the ability to capture the forward and backward features of the signal more comprehensively, and improve the accuracy and robustness of signal prediction. This network structure can more easily process complex long signal data, enhance the understanding of the dynamic changes of the signal, facilitate the extraction of self-signal and its interference features, and enhance the model's ability to capture time series features to facilitate the model to handle long sequence signal interference.

[0043] The following introduces the training process of the DNN-BiLSTM model.

[0044] Optionally, the training process of the DNN-BiLSTM model is as follows: Simulate the ideal enhanced signal emitted by the low-orbit satellite transmitter, the combined signal received by the GNSS receiver, and the interference signal coupled into the combined signal from the ideal enhanced signal, construct training samples and divide them into training set and validation set; Initialize the weights and biases of the DNN-BiLSTM model, input the training set into the DNN-BiLSTM model, calculate the network output through the forward propagation process, and update the weights and biases through the back propagation process of the error to realize the training of the neural network; Iterate forward propagation and back propagation until the model converges and obtains the desired weights and biases; The trained DNN-BiLSTM model is verified through the validation set to obtain the trained DNN-BiLSTM model.

[0045] Specifically, the training samples are constructed by simulating the scenario of co-frequency transmission and reception of the low-orbit satellite navigation enhancement platform, simulating the ideal enhanced signal transmitted by the low-orbit satellite transmitter, the combined signal received by the GNSS receiver, and the interference signal coupled into the combined signal from the ideal enhanced signal, and divided into training set and verification set.

[0046] Furthermore, the weights and biases of the DNN-BiLSTM model are initialized, the training set is input into the DNN-BiLSTM model, the network output is calculated through forward propagation, and the weights and biases are updated through the back propagation process of the error to realize the training of the neural network.

[0047] For the forward propagation process: The forward propagation process refers to the process of passing the input data through the network layer by layer until the output layer. In order to calculate the actual output of the DNN-BiLSTM model, it is carried out from input to output. Among them, the input of the i-th node of the hidden layer (BiLSTM module and DNN module in the DNN-BiLSTM model) and output satisfy:

[0048]

[0049] in, For the input signal components; From the input layer to the hidden layer The weight of each node; For the hidden layer The bias of the node.

[0050] Output layer Node input And the output result It can be calculated based on the output of the hidden layer, which satisfies:

[0051]

[0052] in, The output layer The bias of each node; is the hidden layer Nodes to the output layer The weight of the node. So the output result of the kth node in the output layer It can be further expressed as:

[0053] The activation function used by the DNN module is the Leaky ReLU function. The expression of the Leaky ReLU function is as follows:

[0054] Where a is a constant, ).

[0055] The Leaky ReLU function is used in the DNN module to activate neurons from input to output, which can introduce nonlinear factors into neurons, allowing the DNN network to approximate any nonlinear function, helping the model to learn the effective features of the input signal and enhance its ability to reconstruct nonlinear interference.

[0056] Compare the actual output of the output layer with the target output (the interference signal coupled into the combined signal in the ideal enhanced signal in the training sample) and calculate the error:

[0057] in, The output layer If the expectation cannot be met, the error will be propagated backwards.

[0058] For the back propagation process: the back propagation process refers to calculating the error and calculating the adjustment amount of the weights and thresholds of each layer of the neural network according to the error criterion function combined with the adaptive gradient algorithm to achieve the correction of the weights and bias of the neural network. Optionally, the back propagation process of the error adopts the RMSProp algorithm, and the loss function adopts the MSE function.

[0059] The core of the RMSProp algorithm is to use different learning rates for each parameter. These learning rates are adaptively adjusted according to the size of the most recent gradient of the parameter to avoid the problems caused by the global learning rate. Specifically, the correction amount of the output layer weights during the back propagation process is , the correction amount of the output layer bias , the correction value of hidden layer weight , the correction amount of hidden layer bias Satisfy respectively:

[0060]

[0061]

[0062]

[0063] in, is the learning rate.

[0064] By continuously iterating forward propagation and back propagation until the model converges (reaching the preset number of training times or loss convergence), stop training and obtain the expected weights and biases; then use the validation set to verify the trained DNN-BiLSTM model. If the interference cancellation result of the validation set can reach the level of the training set, it is confirmed that the trained DNN-BiLSTM model has the ability to cancel interference and save the model.

[0065] The following introduces the simulation process of obtaining the ideal enhanced signal transmitted by the low-orbit satellite transmitter, the combined signal received by the GNSS receiver, and the interference signal coupled into the combined signal from the ideal enhanced signal.

[0066] Figure 4 This is an enhanced co-frequency transceiver model diagram of a low-orbit satellite navigation platform provided in an embodiment of the present application, such as Figure 4 As shown in the figure, the model simulates the whole process of a low-orbit satellite receiving a self-interference signal in an actual scenario. The whole process can be divided into a RF channel model and a channel transmission model.

[0067] The RF channel model mainly simulates the generation of enhanced signals transmitted by the antenna end of the low-orbit satellite, and simulates the linear distortion caused by the filter, output multiplexer and antenna during the transmission process, as well as the nonlinear distortion caused by the high power amplifier. The channel transmission model mainly simulates the enhanced signal propagating through the non-ideal channel to the GNSS receiving antenna, including the multipath effect and other effects. The specific process is as follows: Step a, simulating an ideal signal transmitted by a low-orbit satellite transmitter.

[0068] The low-orbit satellite transmitter transmits a standardized constant envelope multiplexed signal, i.e., an ideal signal. In order to reduce the distortion of the combined signal when it passes through the high power amplifier, the Phase Optimized Constant Envelope Transmission (POCET) constant envelope multiplexing technology is used to generate a standardized constant envelope multiplexing signal in the low-orbit satellite transmitter.

[0069] Taking Beidou B1 frequency as an example, the B1I signal expression is:

[0070] in, is the B1I signal amplitude, is the B1I signal carrier frequency, It is the initial phase of B1I signal carrier.

[0071] The B1C signal expression is:

[0072] in, is the B1C signal amplitude, is the B1C signal carrier frequency, It is the initial phase of B1C signal carrier.

[0073] The structure of the B1A signal is similar to that of the B1C signal, consisting of a data component and a pilot component. The B1A signal expression is:

[0074] in, is the B1A signal amplitude, is the B1A signal carrier frequency, It is the initial phase of the B1A signal carrier.

[0075] The constant envelope multiplexing signal at the B1 frequency point, that is, the ideal signal expression for transmission is:

[0076] in, and are the transmission power and transmission phase of the i-th signal respectively, C, It is the intermodulation component.

[0077] Step b, simulating an ideal enhanced signal transmitted by a low-orbit satellite transmitter.

[0078] Ideal signal Generate an enhanced signal transmitted by the RF end through the RF channel model , Figure 5 is a diagram of the radio frequency channel baseband equivalent model provided in an embodiment of the present application, such as Figure 5As shown, the RF channel model includes the following components: Power amplifier pre-filter: A power amplifier pre-filter is used to characterize the linear distortion behavior of the signal generation unit and the frequency generation and up-conversion unit in the low-orbit satellite navigation system.

[0079] High power amplifier: Using the Saleh model, the baseband equivalent output amplitude of the signal after being processed by the high power amplifier is and baseband equivalent output phase offset They are:

[0080]

[0081] in, , , and The four parameters together determine the degree of amplitude and phase distortion. is the complex envelope amplitude of the input signal.

[0082] Power Amplifier Post Filter: A power amplifier post filter is used to characterize the linear distortion behavior of the output multiplexer and antenna in a low-orbit satellite navigation system.

[0083] The ideal signal obtained in step a The signal is enhanced by the power amplifier pre-filter, Saleh power amplifier, and power amplifier post-filter in sequence. By properly setting the filter and power amplifier, the ideal enhanced signal emitted by the low-orbit satellite transmitter can be simulated. , providing accurate input for subsequent signal generation.

[0084] Step c, simulating the interference signal coupled into the combined signal from the combined signal and the ideal enhanced signal received by the GNSS receiver.

[0085] Ideal enhanced signal The interference signal coupled into the GNSS receiver is obtained through the channel transmission model :

[0086] in, is the number of multipaths, For the The amplitude attenuation of the path, Indicates The channel transmission model can accurately simulate the various distortions of the signal in the process of propagating through non-ideal channels to the GNSS receiving antenna, including the multipath effect.

[0087] The process of generating a simulated GNSS signal is similar to that of generating an enhanced signal, but its power is much smaller than that of an enhanced signal.

[0088] Finally, the combined signal is received at the GNSS receiver :

[0089] in, is the GNSS signal received by the GNSS receiver, is the interference signal coupled into the GNSS receiver, For noise.

[0090] The training sample construction is completed by simulating each step of signal generation.

[0091] The on-board co-frequency interference cancellation method based on deep neural network provided in the embodiment of the present application, combined with low-orbit satellite technology and co-frequency transceiver mechanism, models the entire process of a low-orbit satellite receiving a self-interference signal, and can efficiently and accurately generate an ideal enhanced signal at the low-orbit satellite transmitter and a received signal at the GNSS receiver, providing training and verification samples for the modeling of interference cancellation.

[0092] The on-board co-frequency interference cancellation device based on a deep neural network provided in the present application is described below. The on-board co-frequency interference cancellation device based on a deep neural network described below and the on-board co-frequency interference cancellation method based on a deep neural network described above can be referenced to each other.

[0093] Figure 6 is a schematic diagram of the structure of an on-board co-frequency interference cancellation device based on a deep neural network provided in an embodiment of the present application, such as Figure 6 As shown, the device at least includes: An acquisition module 601 is used to acquire a preprocessed input signal sequence, where the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; The reconstruction module 602 is used to input the input signal sequence into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; the target interference signal is a portion of the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver; The cancellation module 603 is used to perform interference cancellation by subtracting the combined signal received by the GNSS receiver from the reconstructed target interference signal to obtain a GNSS receiver receiving signal after the interference cancellation.

[0094] Optionally, the DNN-BiLSTM model includes an input layer, a BiLSTM module, a DNN module and an output layer; the BiLSTM module includes a forward LSTM layer and a backward LSTM layer; and the DNN module includes a Dropout layer and two fully connected layers.

[0095] Optionally, the activation functions of the two fully connected layers in the DNN module are Leaky ReLU functions.

[0096] Optionally, the training process of the DNN-BiLSTM model is as follows: Simulate the ideal enhanced signal emitted by the low-orbit satellite transmitter, the combined signal received by the GNSS receiver, and the interference signal coupled into the combined signal from the ideal enhanced signal, construct training samples and divide them into training set and validation set; Initialize the weights and biases of the DNN-BiLSTM model, input the training set into the DNN-BiLSTM model, calculate the network output through the forward propagation process, and update the weights and biases through the back propagation process of the error to realize the training of the neural network; Iterate the forward propagation process and the back propagation process until the model converges and obtains the desired weights and biases; The trained DNN-BiLSTM model is verified through the validation set to obtain the trained DNN-BiLSTM model.

[0097] Optionally, the error back propagation process uses the RMSProp algorithm.

[0098] Optionally, simulating an ideal enhanced signal transmitted by a low-orbit satellite transmitter, a combined signal received by a GNSS receiver, and an interference signal coupled into the combined signal from the ideal enhanced signal includes: POCET constant envelope multiplexing technology is used to simulate the ideal signal transmitted by the low-orbit satellite transmitter; The ideal signal is input into the RF channel model for enhancement to simulate the ideal enhanced signal transmitted by the low-orbit satellite transmitter; the RF channel model includes a power amplifier pre-filter, a Saleh power amplifier and a power amplifier post-filter which are set in sequence; The ideal enhanced signal is input into the channel transmission model to simulate the interference signal in the ideal enhanced signal coupled into the combined signal received by the GNSS receiver, and the combined signal received by the GNSS receiver.

[0099] Optionally, the reconstruction module 602 is specifically configured to: The input signal sequence is obtained by segmenting a sliding window of a fixed size and input into the DNN-BiLSTM model in the form of a short signal sequence.

[0100] Optionally, the acquisition module 601 is specifically used for: The ideal enhanced signal transmitted by the low-orbit satellite transmitter is sampled in two ways, I / Q, to obtain the real component sequence and imaginary component sequence of the ideal enhanced signal; the combined signal received by the GNSS receiver is sampled in two ways, I / Q, to obtain the real component sequence and imaginary component sequence of the combined signal.

[0101] It can be understood that the detailed functional implementation of each of the above-mentioned units / modules can be found in the introduction of the aforementioned method embodiment, and will not be repeated here.

[0102] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method, which will not be repeated here.

[0103] Based on the method in the above embodiment, an embodiment of the present application provides an electronic device. The device may include: at least one memory for storing programs and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the above embodiment.

[0104] Figure 7 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 7 As shown, the electronic device may include: a processor (Processor) 701, a communication interface (Communications Interface) 702, a memory (Memory) 703 and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704. The processor 701 may call the software instructions in the memory 703 to execute the method described in the above embodiment.

[0105] In addition, the logic instructions in the above-mentioned memory 703 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the 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 for 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 methods of each embodiment of the present application.

[0106] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0107] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0108] It is understandable that the processor in the embodiment of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0109] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0110] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0111] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0112] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for canceling co-frequency interference on a satellite based on a deep neural network, characterized in that: include: Acquire a preprocessed input signal sequence, wherein the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; Input the input signal sequence into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; the target interference signal is a portion of the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver; The combined signal received by the GNSS receiver is subtracted from the reconstructed target interference signal to perform interference cancellation, so as to obtain a GNSS receiver receiving signal after the interference cancellation.

2. The method for cancelling co-channel interference on a satellite according to claim 1, characterized in that: The DNN-BiLSTM model includes an input layer, a BiLSTM module, a DNN module and an output layer; the BiLSTM module includes a forward LSTM layer and a backward LSTM layer; the DNN module includes a Dropout layer and two fully connected layers.

3. The method for cancelling co-channel interference on a satellite according to claim 2, characterized in that: The activation function of the two fully connected layers in the DNN module is the Leaky ReLU function.

4. The method for cancelling co-channel interference on a satellite according to claim 2, characterized in that: The training process of the DNN-BiLSTM model is as follows: Simulate the ideal enhanced signal emitted by the low-orbit satellite transmitter, the combined signal received by the GNSS receiver, and the interference signal coupled into the combined signal from the ideal enhanced signal, construct training samples and divide them into training set and validation set; Initializing the weights and biases of the DNN-BiLSTM model, inputting the training set into the DNN-BiLSTM model, calculating the network output through a forward propagation process, and updating the weights and biases through a back propagation process of the error to implement neural network training; Iterate the forward propagation process and the back propagation process until the model converges and obtains the desired weights and biases; The trained DNN-BiLSTM model is verified by using the verification set to obtain the trained DNN-BiLSTM model.

5. The method for cancelling co-channel interference on a satellite according to claim 4, characterized in that: The error back propagation process adopts the RMSProp algorithm.

6. The method for cancelling co-channel interference on a satellite according to claim 4, characterized in that: The method of simulating an ideal enhanced signal transmitted by a low-orbit satellite transmitter, a combined signal received by a GNSS receiver, and an interference signal coupled into the combined signal from the ideal enhanced signal includes: POCET constant envelope multiplexing technology is used to simulate the ideal signal transmitted by the low-orbit satellite transmitter; The ideal signal is input into the radio frequency channel model for enhancement to simulate the ideal enhanced signal transmitted by the low-orbit satellite transmitter; the radio frequency channel model includes a power amplifier pre-filter, a Saleh power amplifier and a power amplifier post-filter arranged in sequence; The ideal enhanced signal is input into a channel transmission model to simulate an interference signal in the ideal enhanced signal coupled into a combined signal received by a GNSS receiver, and the combined signal received by the GNSS receiver.

7. The on-board co-channel interference cancellation method according to claim 1, characterized in that: The step of inputting the input signal sequence into the DNN-BiLSTM model comprises: The input signal sequence obtained by segmenting with a sliding window of a fixed size is input into the DNN-BiLSTM model in the form of a short signal sequence.

8. The on-board co-channel interference cancellation method according to claim 1 or 7, characterized in that: The step of obtaining the preprocessed input signal sequence comprises: The ideal enhanced signal transmitted by the low-orbit satellite transmitter is sampled in two ways, I / Q, to obtain a real component sequence and an imaginary component sequence of the ideal enhanced signal; the combined signal received by the GNSS receiver is sampled in two ways, I / Q, to obtain a real signal component sequence and an imaginary component sequence of the combined signal.

9. A satellite co-frequency interference cancellation device based on deep neural network, characterized in that: include: An acquisition module, used to acquire a preprocessed input signal sequence, wherein the input signal sequence includes an ideal enhanced signal sequence transmitted by a low-orbit satellite transmitter and a combined signal sequence received by a GNSS receiver; A reconstruction module, used for inputting the input signal sequence into the DNN-BiLSTM model to reconstruct the target interference signal in the combined signal; The target interference signal is a portion of the ideal enhanced signal transmitted by the low-orbit satellite transmitter that overlaps with the receiving frequency band of the GNSS receiver; The cancellation module is used to perform interference cancellation by subtracting the combined signal received by the GNSS receiver from the reconstructed target interference signal to obtain the GNSS receiver receiving signal after the interference cancellation.

10. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 8.

Images