Communication perception integrated self-interference reconstruction and elimination method, device and equipment
By introducing a neural network model and a least squares estimator into the integrated communication and sensing system, a self-interference transformation function is constructed, which solves the problem of self-interference cancellation technology's dependence on historical data, realizes efficient self-interference signal reconstruction and cancellation, and improves the system's real-time performance and robustness.
Patent Information
- Application Number
- CN202511067968.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-31
- Publication Date
- 2025-11-11
AI Technical Summary
Existing communication and sensing integrated systems rely excessively on historical data for self-interference cancellation technology, making it difficult to adapt to the dynamic environment of rapidly changing channel conditions and system parameters, resulting in high resource consumption and decreased cancellation performance.
We employ a neural network model based on the concept of extreme learning machine and a least squares estimator. By constructing a self-interference transformation function, we can reconstruct and eliminate self-interference signals using a small number of current samples, thereby reducing our dependence on historical data and hardware resources.
It improves the system's real-time performance and robustness, reduces resource consumption, alleviates overfitting problems, and enhances its adaptability to dynamic environments.
Smart Images
Figure CN120934648A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of integrated communication and sensing technology, and in particular to a method, apparatus and device for self-interference reconstruction and elimination in integrated communication and sensing. Background Technology
[0002] Integrated sensing and communication technology plays a crucial role in numerous applications, including intelligent transportation and the low-altitude economy. In recent years, with the development of hardware technology and the increasing convergence of communication and radar frequency bands, integrated sensing and communication design has received growing attention, and many practical examples of related joint systems have emerged. However, system integration still faces numerous challenges in close-range detection scenarios under dynamic conditions. To more accurately sense close-range targets, the device needs to receive the echo while transmitting the signal. However, since the transmitted signal and the echo share the same frequency band, the transmitted signal is inevitably misinterpreted as the echo by the receiver, leading to serious self-interference problems, especially nonlinear self-interference.
[0003] Current self-interference cancellation techniques rely too heavily on historical data. However, in integrated communication and sensing systems, time-varying channel and system characteristics are more prevalent, and some low-quality historical data can lead to a decline in cancellation performance. To achieve reliable performance, traditional deep learning methods often require a large amount of high-quality self-interference label data and long-term training, which incurs significant resource overhead, significantly increases the cost of self-interference cancellation, and makes it difficult to adapt to dynamic environments with rapidly changing channel states and system parameters. Summary of the Invention
[0004] The technical objective of this application is to provide a method, apparatus, and device for integrated communication sensing self-interference reconstruction and elimination, in order to solve the problems of current self-interference elimination relying too heavily on historical data, having high resource consumption, and being difficult to adapt to dynamic environments with rapidly changing channel states and system parameters.
[0005] To address the aforementioned technical problems, embodiments of this application provide a method for self-interference reconstruction and elimination integrating communication and sensing, comprising:
[0006] A neural network model is constructed based on the radio frequency chain of the integrated communication and sensing system;
[0007] Training data sample pairs are obtained according to a preset time window. The training data sample pairs include: received signal samples obtained from received signals in the digital domain, and transmitted signal samples obtained from digital baseband signals recorded by at least one transmitter.
[0008] The training data samples are input into the neural network model to train it, thereby obtaining the trained target neural network model;
[0009] Based on the target neural network model and the least squares estimator, a self-interference transformation function is jointly constructed.
[0010] According to the self-interference transformation function, the input complete signal pair is reconstructed and canceled to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal from the transmitter and the complete received signal in the digital domain.
[0011] Specifically, in the method described above, the step of obtaining training data sample pairs according to a preset time window includes:
[0012] In the received signal in the digital domain, obtain the received signal sample located within the preset time window;
[0013] In the digital baseband signal, a preselected signal located within the preset time window is determined;
[0014] The preselected signal and the transmitted signal within a time window of a first preset length before the preselected signal are acquired to obtain the transmitted signal sample.
[0015] The received signal samples and the transmitted signal samples are paired to obtain the training data sample pairs.
[0016] Preferably, in the method described above, the neural network model includes:
[0017] An input layer is used to input the training data sample pairs into the neural network model;
[0018] An in-phase and quadrature (IQ) imbalance factor prediction layer is used to perform unsupervised learning on the implicit IQ imbalance features in the received signal samples in the training data sample pairs to determine the imbalance factor.
[0019] An IQ imbalance compensation layer is used to perform IQ imbalance feature correction on the received signal samples based on the imbalance factor.
[0020] The nonlinear representation layer is used to learn the nonlinear effects of the power amplifier based on multiple nonlinear activation units.
[0021] A fully connected output layer is used to output the estimated self-interference signal value for each iteration;
[0022] Preferably, in the method described above, the step of jointly constructing the self-interference transformation function based on the target neural network model and the least squares estimator includes:
[0023] Based on the parameters in the target neural network model, obtain the nonlinear basis function expression;
[0024] The high-dimensional nonlinear expression obtained by mapping the transmitted signal sample through the nonlinear basis function and the received signal sample corresponding to the transmitted signal sample are input into the least squares estimator to obtain a linear mapping matrix;
[0025] The self-interference transformation function is jointly constructed based on the nonlinear basis function expression and the linear mapping matrix.
[0026] Preferably, in the method described above, the step of reconstructing and canceling self-interference in the input complete signal pair according to the self-interference transformation function to obtain the target received signal includes:
[0027] The input signal is obtained by constructing a sliding time window for each transmitted signal in the complete digital baseband transmitted signal according to a time window of the second preset length;
[0028] The input signal is reconstructed using the self-interference transformation function to obtain the reconstructed self-interference signal.
[0029] The target received signal is obtained by subtracting the self-interference signal from the complete received signal.
[0030] Another embodiment of this application also provides a control device, including:
[0031] The neural network model building module is used to build a neural network model based on the radio frequency chain of the integrated communication and sensing system.
[0032] The training data acquisition module is used to acquire training data sample pairs according to a preset time window. The training data sample pairs include: received signal samples acquired from received signals in the digital domain, and transmitted signal samples acquired from digital baseband signals recorded by at least one transmitter.
[0033] The model training module is used to input the training data samples into the neural network model to train the target neural network model and obtain the trained target neural network model.
[0034] A joint construction module is used to jointly construct a self-interference transformation function based on the target neural network model and the least squares estimator;
[0035] The self-interference reconstruction and cancellation module is used to reconstruct and cancel the self-interference of the input complete signal pair according to the self-interference transformation function to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal of the transmitter and the complete received signal in the digital domain.
[0036] Another embodiment of this application provides an electronic device including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the communication-sensing integrated self-interference reconstruction and elimination method as described above.
[0037] Another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the communication-sensing integrated self-interference reconstruction and elimination method as described above.
[0038] Another embodiment of this application provides a computer program product including computer instructions that, when executed by a processor, implement the steps of the communication-sensing integrated self-interference reconstruction and elimination method as described above.
[0039] Compared with the prior art, the communication-sensing integrated self-interference reconstruction and elimination method, apparatus and device provided in the embodiments of this application have at least the following beneficial effects:
[0040] This application introduces a neural network model based on the concept of extreme learning machine and uses a least squares estimator to achieve efficient reconstruction and suppression of self-interference signals with a small number of current samples. This effectively improves the real-time performance, robustness and resource efficiency of the system, while alleviating the overfitting problem and reducing the dependence on historical data and hardware resources. Attached Figure Description
[0041] Figure 1 This is one of the flowcharts illustrating the self-interference reconstruction and elimination method in the embodiments of this application;
[0042] Figure 2 This is a schematic diagram of the radio frequency chain structure of a single base station in an integrated active sensing scenario of communication and sensing in this application embodiment;
[0043] Figure 3 This is a schematic diagram of the neural network model in the embodiments of this application;
[0044] Figure 4 This is a second schematic flowchart of the self-interference reconstruction and elimination method in the embodiments of this application;
[0045] Figure 5 This is a schematic diagram illustrating the acquisition of training data pairs in an embodiment of this application;
[0046] Figure 6 This is the third flowchart illustrating the self-interference reconstruction and elimination method in the embodiments of this application;
[0047] Figure 7This is a schematic diagram illustrating the acquisition of the self-interference transformation function in an embodiment of this application;
[0048] Figure 8 This is the fourth flowchart illustrating the self-interference reconstruction and elimination method in the embodiments of this application;
[0049] Figure 9 This is a schematic diagram comparing the self-interference elimination performance of the technical solutions in this application with that of traditional physical polynomial model systems and linear suppression systems.
[0050] Figure 10 This is a schematic diagram of the control device in an embodiment of this application. Detailed Implementation
[0051] To make the technical problems, technical solutions, and advantages of this application clearer, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments. In the following description, specific details such as particular configurations and components are provided merely to aid in a comprehensive understanding of the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Furthermore, for clarity and brevity, descriptions of known functions and structures have been omitted.
[0052] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.
[0053] In the various embodiments of this application, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0054] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0055] In the embodiments provided in this application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.
[0056] See Figure 1 One embodiment of this application provides a method for self-interference reconstruction and elimination integrating communication sensing, comprising:
[0057] Step S101: Construct a neural network model based on the radio frequency chain of the integrated communication and sensing system;
[0058] Step S102: Obtain training data sample pairs according to a preset time window. The training data sample pairs include: received signal samples obtained from received signals in the digital domain, and transmitted signal samples obtained from digital baseband signals recorded by at least one transmitter.
[0059] Step S103: Input the training data samples into the neural network model to train it, and obtain the trained target neural network model;
[0060] Step S104: Based on the target neural network model and the least squares estimator, jointly construct the self-interference transformation function;
[0061] Step S105: According to the self-interference transformation function, perform self-interference reconstruction and self-interference cancellation on the input complete signal pair to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal from the transmitter and the complete received signal in the digital domain.
[0062] The self-interference reconstruction and elimination method for integrated communication and sensing provided in this embodiment, in terms of execution, firstly constructs a neural network model based on the radio frequency chain of the integrated communication and sensing system, wherein, see [link to documentation]. Figure 2The diagram shows the radio frequency chain structure of a single base station with a single transmitting antenna and a single receiving antenna in an integrated active sensing scenario. First, an orthogonal frequency division multiplexing (OFDM) signal x(n), i.e., the sensing signal, is generated. After being processed by a digital-to-analog converter (DAC), an IQ mixer, and a power amplifier (PA), the signal is radiated into free space by the transmitting antenna. Part of the signal is reflected by the target object to generate a target echo, which is received by the receiving antenna. The other part of the signal forms a self-interference signal through the direct coupling between the transmitting and receiving antennas and the near-field echo, which is also captured by the receiving antenna. After analog domain superposition at the receiving antenna, the two signals are amplified by a low-noise amplifier (LNA) and an IQ mixer before entering an analog-to-digital converter (ADC) to restore the signal to the digital domain y(n). Then, the self-interference cancellation module performs parameter estimation. Since the self-interference signal consists of direct coupling components and near-field echo components, its attenuation is minimal while its power is much higher than that of the target echo. This makes it difficult to eliminate the nonlinear part of the self-interference signal caused by passing through the RF chain and the self-interference channel.
[0063] Based on the above RF chain structure, this embodiment constructs a multi-layer neural network model based on the mathematical modeling of this RF chain. This model needs to consider: IQ imbalance non-ideal modeling of the IQ mixer, power amplifier non-ideal modeling, and self-interference channel modeling. Specifically, for the IQ imbalance non-ideal modeling of the IQ mixer, the amplitude imbalance and phase imbalance it generates are considered. The mixed signal is then represented as: x IQ (n)=K1x(n)+K2x * (n), where x(n) represents the transmitted signal, x * (n) represents the conjugate transform of the transmitted signal, and K1 and K2 represent the imbalance factors, with K1 >> K2. For non-ideal modeling of the power amplifier, a generalized memory polynomial model is used to capture its dynamic nonlinear response to a large-bandwidth input signal. The amplified signal is represented as: Where M represents the maximum memory depth, P represents the maximum nonlinear order, and Ψ p h p This represents the nonlinear basis functions and their corresponding gain coefficients for different nonlinear orders. For self-interference channel modeling, multipath effect modeling needs to be considered; in this case, the signal after self-interference is represented as: Taking all three considerations into account, the above formulas can be combined and put into a single equation:
[0064] Based on the above considerations, the constructed neural network model is as follows: Figure 3 As shown, it includes:
[0065] The input layer is used to input the training data sample pairs into the neural network model, and to solve the memory effect and multipath effect corresponding to the M and P parameters in the above model.
[0066] The IQ imbalance factor prediction layer, also known as the fusion layer, is used to perform unsupervised learning on the IQ imbalance features hidden in the received signal samples in the training data sample pairs to determine the imbalance factors, namely K1 and K2 mentioned above.
[0067] An IQ imbalance compensation layer is used to perform IQ imbalance feature correction on the received signal samples based on the imbalance factor, i.e., to recover x. IQ Part of;
[0068] The nonlinear representation layer is used to learn the nonlinear effects of the power amplifier, i.e., Ψ in the power amplifier, based on multiple layers of nonlinear activation units; among them, the IQ imbalance compensation layer and the nonlinear representation layer can be collectively referred to as the nonideal learning layer.
[0069] A fully connected output layer is used to output the self-interference signal estimate for each iteration, i.e., y. si .
[0070] After constructing the aforementioned neural network model, training data is needed to facilitate its training. In this embodiment, a small amount of clean self-interference signals are obtained from the received and transmitted signals within a preset time window for model training. This reduces the reliance on historical data, and training with a small amount of high-quality signals improves training efficiency and quality, thereby reducing resource consumption. Furthermore, by avoiding historical data bias caused by uneven data distribution, the stability and robustness of self-interference reconstruction and elimination are enhanced. Since this method does not rely on additional hardware overhead and comprehensively considers the changing characteristics of the actual self-interference channel, it also enhances the ability to suppress self-interference in dynamic environments.
[0071] After obtaining the training data sample pairs, model training can be performed based on them to obtain the target neural network model after training. The criteria for determining whether training is complete include, but are not limited to, those based on the number of iterations and / or training error. After obtaining the target neural network model, the parameters in the target neural network model will be frozen to avoid further parameter changes.
[0072] Furthermore, a self-interference transformation function can be jointly constructed based on the target neural network model and the least squares method. Based on this self-interference transformation function, the complete signal pairs that need to be reconstructed and eliminated for self-interference can be reconstructed and eliminated to obtain the final target received signal.
[0073] In summary, the embodiments of this application introduce a neural network model based on the idea of extreme learning machine and use a least squares estimator to achieve efficient reconstruction and suppression of self-interference signals with a small number of current samples, which effectively improves the real-time performance, robustness and resource efficiency of the system, while alleviating the overfitting problem and reducing the dependence on historical data and hardware resources.
[0074] See Figure 4 and Figure 5 Specifically, in the method described above, obtaining training data sample pairs according to a preset time window includes:
[0075] Step S401: In the received signal in the digital domain, obtain the received signal sample located within the preset time window;
[0076] Step S402: In the digital baseband signal, determine the preselected signal located within the preset time window;
[0077] Step S403: Obtain the preselected signal and the transmitted signal within a first preset time window before the preselected signal to obtain the transmitted signal sample;
[0078] Step S404: Pair the received signal sample and the transmitted signal sample to obtain the training data sample pair.
[0079] In this embodiment, the steps of obtaining training data sample pairs according to a preset time window are illustrated. First, in the digital domain received signal, a received signal sample located within the preset time window is obtained. Then, in the digital baseband signal, a preselected signal within the same time window as the received signal sample is determined. Since both the power amplifier and the self-interference channel have significant memory and multipath effects, the self-interference signal at the current moment is not only related to the transmitted signal at that moment, but also depends on the transmitted signals at multiple previous moments. Therefore, the preselected signal and the transmitted signal within a time window of a first preset length before the preselected signal are obtained as transmitted signal samples. Then, the timestamps corresponding to the transmitted signal samples and the timestamps corresponding to the received signals are paired to obtain the required training data sample pairs.
[0080] It should be noted that the training data sample pairs can be represented as: Among them, n represents the current time n, and P represents the preset length. When n < P, it means that the preamble signal samples at the current time are insufficient. At this time, to ensure the neat format of the transmitted signal samples, zero-padding is used to complete each transmitted signal sample.
[0081] See Figure 6 and Figure 7 Preferably, for the method as described above, the joint construction of the self-interference transformation function according to the target neural network model and the least squares estimator includes:
[0082] Step S601, obtain the non-linear basis function expression according to the parameters in the target neural network model;
[0083] Step S602, input the high-dimensional non-linear expression obtained by mapping the transmitted signal samples through the non-linear basis function and the received signal samples corresponding to the transmitted signal samples into the least squares estimator to obtain a linear mapping matrix;
[0084] Step S603, jointly construct the self-interference transformation function according to the non-linear basis function expression and the linear mapping matrix.
[0085] In this embodiment, after obtaining the target neural network, when jointly constructing the self-interference transformation function by combining the least squares estimator, first, according to the parameters in the target neural network model, obtain the non-linear basis function expression in the target neural network model. The non-linear basis function expression is the aforementioned Ψ, denoted as f here. Furthermore, the transmitted signal samples can be mapped to their high-dimensional non-linear expressions in the high-dimensional features through the non-linear basis function expression, that is Furthermore, by inputting the high-dimensional non-linear expression and the received signal samples corresponding to the transmitted signal samples into the least squares estimator, a linear mapping matrix is obtained. Among them, the linear mapping matrix is expressed as:
[0086] Furthermore, by performing a dot product on the non-linear basis function expression and the linear mapping matrix, the required self-interference transformation function can be obtained, that is, H·f.
[0087] It should be noted that to ensure the accuracy of the obtained non-linear basis function, after obtaining the target neural network, the parameters of the IQ imbalance compensation layer and the non-linear representation layer of the neural network will be frozen.
[0088] See Figure 8 Preferably, for the method as described above, the self-interference reconstruction and self-interference cancellation of the input complete signal pair according to the self-interference transformation function to obtain the target received signal includes:
[0089] Step S801: Construct a sliding time window for each transmitted signal in the complete digital baseband transmitted signal according to a second preset time window to obtain the input signal;
[0090] Step S802: Perform self-interference reconstruction on the input signal according to the self-interference transformation function to obtain the reconstructed self-interference signal;
[0091] Step S803: Subtract the self-interference signal from the complete received signal to obtain the target received signal.
[0092] In this embodiment, an example is given of self-interference reconstruction and self-interference elimination based on the self-interference transformation function obtained above. First, the complete digital baseband transmission signal is preprocessed to obtain the input signal. Specifically, a sliding time window is constructed based on the second preset length time window before each transmission signal. When the time length corresponding to the signal before a transmission signal is less than the second preset length, zero padding is used to fill it. It should be noted that the second preset length in this embodiment is the first preset length plus one mentioned above.
[0093] After obtaining the input signal, the self-interference transformation function can be used to reconstruct the self-interference signal, resulting in the reconstructed self-interference signal. Then, by subtracting the aforementioned self-interference signal from the complete received signal, the target received signal can be obtained, i.e. It should be noted that the reconstructed self-interference signal is output time-by-time and subtracted point by point from the complete received signal to obtain the target received signal, which is then used for subsequent parameter estimation, and can significantly improve the near-range sensing performance of the integrated communication and sensing system.
[0094] See Figure 9 The diagram shows a comparison of the self-interference cancellation performance of the technical solution provided in this application with that of traditional physical polynomial model systems and linear suppression systems. The X-axis represents the number of available learning samples, and the Y-axis represents the interference cancellation ratio (ICR) in dB, specifically the ratio between the effective signal power and the residual interference signal power.
[0095] It is evident that, with a small sample size, the integrated communication and sensing system in this application exhibits a higher suppression ratio compared to traditional physical multinomial model systems and linear suppression systems. Furthermore, when the sample size allows the multinomial model to achieve high performance, this application's embodiments still achieve similar self-interference suppression performance.
[0096] See Figure 10Another embodiment of this application also provides a control device, including:
[0097] Neural network model building module 1001 is used to build a neural network model based on the radio frequency chain of the integrated communication and sensing system.
[0098] The training data acquisition module 1002 is used to acquire training data sample pairs according to a preset time window. The training data sample pairs include: received signal samples acquired from received signals in the digital domain, and transmitted signal samples acquired from digital baseband signals recorded by at least one transmitter.
[0099] The model training module 1003 is used to input the training data samples into the neural network model to train it, and obtain the trained target neural network model.
[0100] The joint construction module 1004 is used to jointly construct the self-interference transformation function based on the target neural network model and the least squares estimator;
[0101] The self-interference reconstruction and cancellation module 1005 is used to reconstruct and cancel the self-interference of the input complete signal pair according to the self-interference transformation function to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal of the transmitter and the complete received signal in the digital domain.
[0102] Specifically, in the apparatus described above, the training data acquisition module includes:
[0103] The first acquisition unit is used to acquire the received signal sample located within the preset time window from the received signal in the digital domain;
[0104] The first determining unit is used to determine a preselected signal located within the preset time window in the digital baseband signal;
[0105] The second acquisition unit is used to acquire the preselected signal and the transmitted signal within a time window of a first preset length before the preselected signal, and to obtain the transmitted signal sample.
[0106] A pairing unit is used to pair the received signal samples and the transmitted signal samples to obtain the training data sample pairs.
[0107] Preferably, in the apparatus described above, the neural network model includes:
[0108] An input layer is used to input the training data sample pairs into the neural network model;
[0109] The IQ imbalance factor prediction layer is used to perform unsupervised learning on the IQ imbalance features hidden in the received signal samples in the training data sample pairs to determine the imbalance factor.
[0110] An IQ imbalance compensation layer is used to perform IQ imbalance feature correction on the received signal samples based on the imbalance factor.
[0111] The nonlinear representation layer is used to learn the nonlinear effects of the power amplifier based on multiple nonlinear activation units.
[0112] A fully connected output layer is used to output the estimated self-interference signal value for each iteration;
[0113] Preferably, in the apparatus described above, the joint construction module includes:
[0114] The third acquisition unit is used to acquire the nonlinear basis function expression based on the parameters in the target neural network model;
[0115] The fourth acquisition unit is used to input the high-dimensional nonlinear expression obtained by mapping the transmitted signal sample through the nonlinear basis function and the received signal sample corresponding to the transmitted signal sample into the least squares estimator to obtain a linear mapping matrix;
[0116] The function construction unit is used to jointly construct the self-interference transformation function based on the nonlinear basis function expression and the linear mapping matrix.
[0117] Preferably, in the apparatus described above, the self-interference reconstruction and elimination module includes:
[0118] The data preprocessing unit is used to construct a sliding time window for each transmitted signal in the complete digital baseband transmitted signal according to a second preset time window to obtain the input signal;
[0119] The self-interference reconstruction unit is used to reconstruct the input signal based on the self-interference transformation function to obtain the reconstructed self-interference signal.
[0120] The self-interference cancellation unit is used to subtract the self-interference signal from the complete received signal to obtain the target received signal.
[0121] The apparatus embodiments of this application are apparatuses corresponding to the embodiments of the methods described above. All implementation means in the method embodiments described above are applicable to the apparatus embodiments and can achieve the same technical effects. The apparatus embodiments provided in this application can implement all the method steps implemented in the method embodiments described above and can achieve the same technical effects. Therefore, the parts that are the same as those in the method embodiments and their beneficial effects will not be described in detail here.
[0122] Another embodiment of this application provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the self-interference reconstruction and elimination method of communication-sensing integration as described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0123] Another embodiment of this application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the self-interference reconstruction and elimination method of communication-sensing integration as described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0124] Another embodiment of this application provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the communication-sensing integrated self-interference reconstruction and elimination method as described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0125] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.
[0126] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.
[0127] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for self-interference reconstruction and elimination integrating communication and sensing, characterized in that, include: A neural network model is constructed based on the radio frequency chain of the integrated communication and sensing system; Training data sample pairs are obtained according to a preset time window. The training data sample pairs include: received signal samples obtained from received signals in the digital domain, and transmitted signal samples obtained from digital baseband signals recorded by at least one transmitter. The training data samples are input into the neural network model to train it, thereby obtaining the trained target neural network model; Based on the target neural network model and the least squares estimator, a self-interference transformation function is jointly constructed. According to the self-interference transformation function, the input complete signal pair is reconstructed and canceled to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal from the transmitter and the complete received signal in the digital domain.
2. The method according to claim 1, characterized in that, The step of obtaining training data sample pairs according to a preset time window includes: In the received signal in the digital domain, obtain the received signal sample located within the preset time window; In the digital baseband signal, a preselected signal located within the preset time window is determined; The preselected signal and the transmitted signal within a time window of a first preset length before the preselected signal are acquired to obtain the transmitted signal sample. The received signal samples and the transmitted signal samples are paired to obtain the training data sample pairs.
3. The method according to claim 1, characterized in that, The neural network model includes: An input layer is used to input the training data sample pairs into the neural network model; The in-phase orthogonal IQ imbalance factor prediction layer is used to perform unsupervised learning on the IQ imbalance features hidden in the received signal samples in the training data sample pairs to determine the imbalance factor. An IQ imbalance compensation layer is used to perform IQ imbalance feature correction on the received signal samples based on the imbalance factor. The nonlinear representation layer is used to learn the nonlinear effects of the power amplifier based on multiple nonlinear activation units. The fully connected output layer is used to output the self-interference signal estimate for each iteration.
4. The method according to claim 1, characterized in that, The step of jointly constructing the self-interference transformation function based on the target neural network model and the least squares estimator includes: Based on the parameters in the target neural network model, obtain the nonlinear basis function expression; The high-dimensional nonlinear expression obtained by mapping the transmitted signal sample through the nonlinear basis function and the received signal sample corresponding to the transmitted signal sample are input into the least squares estimator to obtain a linear mapping matrix; The self-interference transformation function is jointly constructed based on the nonlinear basis function expression and the linear mapping matrix.
5. The method according to claim 1, characterized in that, The step of reconstructing and canceling self-interference in the input complete signal pair according to the self-interference transformation function to obtain the target received signal includes: The input signal is obtained by constructing a sliding time window for each transmitted signal in the complete digital baseband transmitted signal according to a time window of the second preset length; The input signal is reconstructed using the self-interference transformation function to obtain the reconstructed self-interference signal. The target received signal is obtained by subtracting the self-interference signal from the complete received signal.
6. A control device, characterized in that, include: The neural network model building module is used to build a neural network model based on the radio frequency chain of the integrated communication and sensing system. The training data acquisition module is used to acquire training data sample pairs according to a preset time window. The training data sample pairs include: received signal samples acquired from received signals in the digital domain, and transmitted signal samples acquired from digital baseband signals recorded by at least one transmitter. The model training module is used to input the training data samples into the neural network model to train the target neural network model and obtain the trained target neural network model. A joint construction module is used to jointly construct a self-interference transformation function based on the target neural network model and the least squares estimator; The self-interference reconstruction and cancellation module is used to reconstruct and cancel the self-interference of the input complete signal pair according to the self-interference transformation function to obtain the target received signal. The complete signal pair includes: the complete digital baseband transmitted signal of the transmitter and the complete received signal in the digital domain.
7. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the communication-sensing integration self-interference reconstruction and elimination method as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the method for self-interference reconstruction and elimination of communication-sensing integration as described in any one of claims 1 to 5.
9. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the communication-aware integrated self-interference reconstruction and elimination method as described in any one of claims 1 to 5.