An uplink sensing and communication integrated system signal processing method and device and storage medium

Abstract

The application provides an uplink sensing integrated system signal processing method and device and a storage medium, comprising: performing coarse synchronization on a first signal received by a plurality of antennas to obtain a first time offset, a first carrier frequency offset and a second signal after coarse synchronization; performing preliminary channel estimation according to the second signal to obtain an estimated first channel state information tensor; performing deep learning based on a hybrid neural network according to the first channel state information tensor to obtain an estimated second channel state information tensor, a second time offset and a second carrier frequency offset; performing tensor decomposition according to the second channel state information tensor, and acquiring sensing information according to the first time offset, the first carrier frequency offset, the second time offset and / or the second carrier frequency offset. The application realizes more accurate synchronization error estimation and channel state information denoising, guarantees more reliable communication performance, and realizes high-precision positioning and speed measurement of sensing.

Description

Technical Field

[0001] This application relates to the field of integrated sensing and communication (ISAC), and in particular to a signal processing method, apparatus and storage medium for an integrated sensing and communication system. Background Technology

[0002] In uplink ISAC research, two main challenges are faced: first, eliminating synchronization errors (Time Offset (TO) and Carrier Frequency Offset (CFO)) to ensure highly reliable communication performance; second, the synchronization accuracy of existing communication systems is typically at the level of hundreds of nanoseconds and hundreds of hertz, which can lead to meter-level sensing errors, making it difficult to support high-precision target sensing. Current communication technologies primarily eliminate synchronization errors through synchronization error estimation schemes; uplink ISAC synchronization error elimination and sensing schemes aim to improve sensing accuracy.

[0003] Currently, synchronization error estimation schemes in uplink communication systems are divided into model-based methods and deep learning-based methods. While model-based methods offer higher robustness and lower computational complexity, their estimation accuracy is significantly limited by noise and available pilot resources. Deep learning-based methods make an ideal time-invariant assumption about synchronization error and do not consider the spatial domain advantages of multi-input multi-output (MIMO) systems.

[0004] Regarding uplink ISAC synchronization error cancellation and sensing schemes, existing methods include reference path-based synchronization, cross-antenna cross-correlation, Kalman filtering, and complex convolutional neural network-based methods. Reference path-based synchronization heavily relies on the direct path, which often disappears. Furthermore, the dynamic fluctuations of the wireless channel reduce the accuracy of reference path identification algorithms. Cross-antenna cross-correlation amplifies noise through cross-correlation and requires the transmitter and receiver to remain stationary, making it unsuitable for mobile user scenarios. Kalman filtering-based methods struggle to eliminate the Channel State Information (CSI) error (CFO). Complex convolutional neural network-based methods are ineffective at eliminating both the Time of Occurrence (TO) and CFO.

[0005] Therefore, how to eliminate synchronization errors and improve sensing and communication accuracy remains an important research direction for those skilled in the art. Summary of the Invention

[0006] The technical objective of this application is to provide a signal processing method, apparatus, and storage medium for an uplink integrated sensing system, in order to solve the problem that uplink ISAC synchronization error is difficult to eliminate in the prior art, resulting in low sensing accuracy and communication accuracy.

[0007] To address the aforementioned technical problems, embodiments of this application provide a signal processing method for an integrated uplink sensing system, comprising:

[0008] Coarsely synchronize the first signal received by multiple antennas to obtain the first time offset, the first carrier frequency offset, and the coarsely synchronized second signal;

[0009] Based on the second signal, a preliminary channel estimation is performed to obtain the estimated first channel state information tensor;

[0010] Based on the first channel state information tensor, deep learning based on a hybrid neural network is performed to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset.

[0011] Tensor decomposition is performed based on the second channel state information tensor, and sensing information is obtained based on the first time offset, the first carrier frequency offset, the second time offset, and / or the second carrier frequency offset.

[0012] Specifically, in the uplink sensing integrated system signal processing method described above, the step of coarsely synchronizing the first signal received by multiple antennas to obtain a first time offset, a first carrier frequency offset, and a coarsely synchronized second signal includes:

[0013] The first signal is preprocessed and frequency domain transformed to obtain a frequency domain signal with a preset preamble;

[0014] Sliding correlation is performed based on the frequency domain signal and the pre-stored preamble information to obtain the first time offset and the first carrier frequency offset.

[0015] The first signal is compensated based on the first time offset and the first carrier frequency offset to obtain the second signal.

[0016] Furthermore, in the uplink sensing integrated system signal processing method described above, the preprocessing and frequency domain transformation of the first signal to obtain a frequency domain signal with a preset preamble includes:

[0017] The first signal is downsampled by matched filtering and the cyclic prefix is ​​removed to obtain the preprocessed first signal;

[0018] The preprocessed first signal is transformed to the frequency domain by Fourier transform to obtain the frequency domain signal.

[0019] Preferably, in the uplink sensing integrated system signal processing method described above, the step of performing deep learning based on a hybrid neural network according to the first channel state information tensor to obtain an estimated second channel state information tensor, a second time offset, and a second carrier frequency offset includes:

[0020] The first channel state information tensor is denoised to obtain the denoised third channel state information tensor.

[0021] The first channel state information tensor and the third channel state information tensor are fused according to the learnable dynamic gating weights to obtain the second channel state information tensor.

[0022] The second carrier frequency offset is obtained by estimating the carrier frequency offset based on the cross-converter.

[0023] The second time offset is obtained by estimating the time offset based on the attention mechanism.

[0024] Preferably, in the uplink sensing integrated system signal processing method described above, the step of performing tensor decomposition based on the second channel state information tensor and obtaining sensing information based on the first time offset, the first carrier frequency offset, the second time offset, and the second carrier frequency offset includes:

[0025] Based on the tensor decomposition algorithm, the second channel state information tensor is decomposed to obtain the factor matrix, time offset vector and carrier frequency offset vector.

[0026] Based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix, the user perception information in the perception information is obtained;

[0027] Based on the factor matrix, the time offset vector, and the carrier frequency offset vector, obtain the scatterer sensing information corresponding to each scatterer in the sensing information.

[0028] Specifically, in the uplink sensing integrated system signal processing method described above, the step of obtaining user perception information from the perception information based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix includes:

[0029] Based on the first carrier frequency offset and the second carrier frequency offset, obtain the first radial velocity estimate of the user in the user perception information;

[0030] Based on the first time offset and the second time offset, the first estimated distance between the user and the base station is obtained;

[0031] An angle estimate is determined by performing angle estimation based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm;

[0032] Based on the first angle estimate, the first estimated distance, and the location information of the base station, the first estimated location of the user in the user perception information is obtained.

[0033] Specifically, in the uplink sensing integrated system signal processing method described above, the step of determining the first angle estimate value by performing angle estimation based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm includes:

[0034] Based on the factor matrix, obtain the first angle vector of the user's direct path and the covariance matrix of the first angle vector;

[0035] The covariance matrix is ​​decomposed into eigenvalues ​​to obtain a descending diagonal matrix of eigenvalues ​​and an orthogonal eigenvector matrix.

[0036] Based on the orthogonal eigenvector matrix, the noise subspace is extracted;

[0037] Based on the noise subspace, determine the spatial power function;

[0038] The first angle estimate is determined by searching based on the spatial power function.

[0039] Specifically, in the uplink sensing integrated system signal processing method described above, the step of obtaining the scatterer sensing information corresponding to each scatterer in the sensing information based on the factor matrix, the time offset vector, and the carrier frequency offset vector includes:

[0040] Based on the factor matrix, obtain the second angle vector corresponding to the target scatterer, where the target scatterer is any scatterer in the scene;

[0041] The second angle vector is estimated according to the multi-signal classification algorithm to obtain the second angle estimate value;

[0042] Based on the geometric relationship between the user, the target scatterer, and the base station, and the first estimated distance, a second estimated distance from the target scatterer to the base station is determined;

[0043] Based on the location information of the base station, the second estimated distance, and the second estimated angle value, the reference position estimate of the target scatterer is determined;

[0044] The reference position estimate is optimized according to a preset optimization algorithm to obtain the second estimated position of the target scatterer.

[0045] Preferably, the uplink sensing integrated system signal processing method described above further includes:

[0046] The second signal is finely synchronized based on the minimum mean square error, the second channel state information tensor, the second time offset, and the second carrier frequency offset to obtain the fourth signal.

[0047] The fourth signal is demodulated to determine the target symbol, and the Euclidean distance between the target symbol and the fourth signal is minimized.

[0048] Data recovery is performed on the target symbol to obtain the transmitted data.

[0049] Another embodiment of this application also provides a control device, including:

[0050] The coarse synchronization module is used to coarsely synchronize the first signal received by multiple antennas to obtain the first time offset, the first carrier frequency offset, and the coarsely synchronized second signal.

[0051] The first channel estimation module is used to perform preliminary channel estimation based on the second signal to obtain the estimated first channel state information tensor.

[0052] The second channel estimation module is used to perform deep learning based on a hybrid neural network based on the first channel state information tensor to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset.

[0053] The sensing information acquisition module is used to perform tensor decomposition based on the second channel state information tensor, and to acquire sensing information based on the first time offset, the first carrier frequency offset, the second time offset and / or the second carrier frequency offset.

[0054] 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 uplink sensing integrated system signal processing method as described above.

[0055] 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 uplink sensing integrated system signal processing method as described above.

[0056] Another embodiment of this application provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the uplink sensing integrated system signal processing method as described above.

[0057] Compared with the prior art, the uplink sensing integrated system signal processing method, apparatus and storage medium provided in the embodiments of this application have at least the following beneficial effects:

[0058] This application effectively reduces computational load by employing coarse synchronization and secondary synchronization based on a hybrid neural network. Simultaneously, it considers more realistic time-varying synchronization error characteristics and leverages the denoising, feature extraction, and time-varying capture capabilities of the hybrid neural network architecture to achieve higher-precision synchronization error estimation and channel state information tensor denoising, thereby ensuring more reliable communication performance. Furthermore, by combining the efficient channel state information tensor cleansing effect of the hybrid model and utilizing the space-time-frequency multidimensional matching advantages of tensor decomposition, it can effectively extract high-order channel features and maintain robustness in complex propagation environments, achieving high-precision multipath separation. Ultimately, it enables high-precision positioning and velocity measurement for both user and multi-scatterer sensing. Attached Figure Description

[0059] Figure 1 This is one of the flowcharts illustrating the uplink integrated sensing system signal processing method of this application;

[0060] Figure 2 This is the second flowchart illustrating the signal processing method of the uplink integrated sensing system of this application;

[0061] Figure 3 This is the third flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0062] Figure 4 This is the fourth flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0063] Figure 5 This is the fifth flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0064] Figure 6 This is the sixth flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0065] Figure 7 This is the seventh flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0066] Figure 8 This is the eighth flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0067] Figure 9 This is the ninth flowchart illustrating the uplink integrated sensing system signal processing method of this application;

[0068] Figure 10 This is a schematic diagram of the control device of this application. Detailed Implementation

[0069] 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.

[0070] 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.

[0071] 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.

[0072] 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, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0073] 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 that B is determined solely based on A, but can also be determined based on A and / or other information.

[0074] To facilitate understanding by those skilled in the art, the scenarios and sensing signal models used in the following description will be explained first when describing the embodiments of this application.

[0075] Perception Scenario: This application primarily considers scenarios involving dynamic uplink users and multiple dynamic scatterers in an uplink ISAC system. For example... Figure 3As shown, the user transmits uplink ISAC signals to the base station for uplink communication, and a portion of the uplink ISAC signals is reflected by a scatterer and received by the base station. The base station receives the direct path signal from the user and the indirect path signal reflected by the scatterer, and processes them to achieve uplink communication, user positioning, and velocity measurement of multiple dynamic scatterers. The location of the base station is... , pairing A uniform linear antenna is used for receiving. User pairing. A uniform linear antenna with an antenna spacing of . .

[0076] System model parameter settings:

[0077] 1) The MIMO (Orthogonal Frequency Division Multiplexing) signals transmitted by the user occupy a total of One subcarrier, M OFDM symbols and The space-time-frequency domain resources of each antenna, some of which are known pilot signals and some of which are unknown communication data;

[0078] 2) Without considering the influence of obstacles, there is always a direct path between the user and the base station. Multipath is also considered to be reflected from dynamic scatterers, and the power of the direct path is significantly greater than that of the multipath.

[0079] Sensing signal model

[0080] User's The up-converted OFDM signal transmitted on the antenna can be represented as:

[0081]

[0082] in, It is the transmission power; and These represent the index values ​​of the OFDM symbol and the subcarrier, respectively. This indicates the transmission of known pilot symbols or communication data. and These represent the carrier frequency and the total duration of the OFDM symbol, respectively. Indicates the subcarrier spacing; Represents a rectangular window function.

[0083] The transmitted OFDM signal passes through The reflection from the scatterer is received by the base station, among which the received scatterer is the scatterer. The OFDM symbol, the first The down-converted OFDM signal on each subcarrier can be represented as:

[0084]

[0085] in, , This represents the total attenuation of the direct beam. This represents the total attenuation of the non-direct path. Represents the radar scattering coefficient. Indicates the distance from the user to the base station. Indicates the user has reached the [number]th [number]. The distance between the scatterers Indicates the first The distance from each scatterer to the base station; Indicates the transmit beamforming gain. Indicates the first The transmit power of each diameter; Indicates wavelength. Represents the speed of light; This represents the Doppler frequency shift from the user to the base station, where, This represents the user's radial velocity relative to the base station; Indicates the user has reached the [number]th [number]. Doppler frequency shift from the scatterer to the base station; This indicates the latency between the base station and the user. Indicates the user has reached the [number]th [number]. The time delay between each scatterer and the base station; This represents the angle of arrival (AoA) from the user to the base station. Indicates the first AoA from each scatterer to the base station; This represents the sum of additive white Gaussian noise (AWGM) and interference. This indicates the receiving guide vector.

[0086]

[0087] The timing reference for uplink synchronization is determined by the earliest detectable path identified during the coarse synchronization phase. In scenarios where a detectable direct path exists, this path is equivalent to the physical direct path. Therefore, there will be synchronization errors after coarse synchronization, which we call residual TO and CFO.

[0088] The specific solutions of this application will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0089] See Figure 1 One embodiment of this application provides a signal processing method for an uplink integrated sensing system, comprising:

[0090] Step S101: Perform coarse synchronization on the first signal received by the multiple antennas to obtain the first time offset, the first carrier frequency offset, and the coarsely synchronized second signal;

[0091] Step S102: Perform preliminary channel estimation based on the second signal to obtain the estimated first channel state information tensor;

[0092] Step S103: Perform deep learning based on a hybrid neural network based on the first channel state information tensor to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset.

[0093] Step S104: Perform tensor decomposition based on the second channel state information tensor, and obtain sensing information based on the first time offset, the first carrier frequency offset, the second time offset, and / or the second carrier frequency offset.

[0094] In this embodiment, when processing the uplink sensing system signal, the first signal received by the multiple antennas, namely ISAC, is first coarsely synchronized, and the first time offset, the first carrier frequency offset, and the second signal after coarse synchronization are obtained. By coarsely synchronizing the first signal, significant synchronization error components can be quickly located and removed with low computational complexity, providing a pre-alignment basis for subsequent steps and effectively reducing the negative interference of noise on high-precision estimation.

[0095] Furthermore, a preliminary channel estimation is performed on the second signal, from which significant synchronization error components have been removed, to obtain a first channel state information tensor. This significantly reduces the interference of synchronization error on channel parameter extraction. Especially in cases of low signal-to-noise ratio, this step, as a preprocessing layer, effectively filters out the frequency offset effect caused by synchronization error, providing a "cleaner" initial value for channel estimation in subsequent stages. This pre-alignment mechanism accelerates the convergence of subsequent neural networks and improves stability under dynamic channel fluctuations. Simultaneously, based on the receiving characteristics of multi-antenna systems, the spatial gain of the antenna array can be fully utilized to enhance the detection capability of weak reflection paths and complex paths, effectively solving the problem of direct path disappearance.

[0096] Subsequently, deep learning based on a hybrid neural network is performed on the first channel state information tensor to achieve secondary refinement, resulting in an estimated second channel state information tensor, second time offset, and second carrier frequency offset. This approach overcomes the limitations of existing deep learning methods that only assume time-invariant signals. Through the hybrid network structure, the non-static characteristics of time offset and carrier frequency offset, as well as the spatial domain correlation characteristics of MIMO systems, can be learned simultaneously. It considers more realistic time-varying synchronization error characteristics and utilizes the denoising, feature extraction, and time-varying capture capabilities of the proposed hybrid neural network architecture to achieve higher-precision synchronization error estimation and channel state information tensor denoising, thereby ensuring more reliable communication performance.

[0097] Finally, by performing tensor decomposition on the second channel state information tensor and obtaining the required sensing information based on the obtained time offset and carrier frequency offset, the desired sensing information is acquired. This method combines the efficient channel state information tensor cleansing results of the hybrid model with the spatio-temporal-frequency multi-dimensional matching advantages of tensor decomposition, effectively extracting high-order channel features and maintaining robustness in complex propagation environments. This achieves high-precision multipath separation, ultimately enabling high-precision positioning and velocity measurement for both users and multiple scatterers.

[0098] In summary, this application effectively reduces computational load through coarse synchronization and secondary synchronization based on hybrid neural networks. Simultaneously, it considers more realistic time-varying synchronization error characteristics and utilizes the denoising, feature extraction, and time-varying capture capabilities of the hybrid neural network architecture to achieve higher-precision synchronization error estimation and channel state information tensor denoising, thereby ensuring more reliable communication performance. Furthermore, combining the efficient channel state information tensor cleansing effect of the hybrid model and leveraging the space-time-frequency multidimensional matching advantages of tensor decomposition, it can effectively extract high-order channel features and maintain robustness in complex propagation environments, achieving high-precision multipath separation and ultimately realizing high-precision positioning and velocity measurement for user and multi-scatterer perception.

[0099] See Figure 2 Specifically, in the uplink sensing integrated system signal processing method described above, the step of coarsely synchronizing the first signal received by multiple antennas to obtain a first time offset, a first carrier frequency offset, and a coarsely synchronized second signal includes:

[0100] Step S201: Preprocess and frequency domain transform the first signal to obtain a frequency domain signal with a preset preamble;

[0101] Step S202: Perform sliding correlation based on the frequency domain signal and the pre-stored preamble information to obtain the first time offset and the first carrier frequency offset.

[0102] Step S203: Compensate the first signal according to the first time offset and the first carrier frequency offset to obtain the second signal.

[0103] In this embodiment, during coarse synchronization, the first signal is preprocessed and frequency-domain transformed to obtain a frequency-domain signal with a preset preamble. This frequency-domain representation converts the time-domain synchronization features into precisely extractable frequency-domain phase difference information. Compared to methods that directly process the time domain, frequency-domain transformation effectively suppresses noise interference and enhances the peak characteristic recognition capability of the preamble sequence. The pre-stored preamble, as a known reference signal, forms stable phase-amplitude correlation features in the frequency domain, enabling subsequent correlation operations to maintain a high synchronization flag detection rate even in low signal-to-noise ratio environments, providing a reliable "anchor" positioning basis for subsequent high-precision parameter estimation.

[0104] The first time offset and the first carrier frequency offset are obtained using a sliding correlation mechanism, which estimates the time delay and Doppler of the direct path, respectively: and ,in, and This refers to the first time offset and the first carrier frequency offset estimated based on the preamble. It can be seen that the delay and Doppler of the direct path are equivalent to the first time offset and the first carrier frequency offset, respectively. Therefore, the delay and the first time offset of the direct path will be collectively referred to as the first time offset, and the Doppler of the direct path and the first carrier frequency offset will be collectively referred to as the first carrier frequency offset. and These represent the direct trajectory estimation errors, which consist of the residual synchronization errors after coarse synchronization. and These are the residual time offset vector and the carrier frequency offset vector, respectively, whose elements can be represented as:

[0105]

[0106]

[0107] in, and These represent the correlation coefficients between the current offset value and the offset value of the previous symbol, respectively. and These represent the random fluctuations in residual time offset and carrier frequency offset, respectively.

[0108] As can be seen from the above, the residual time offset vector and carrier frequency offset are modeled as a first-order autoregressive model to characterize their time-varying characteristics caused by oscillator and phase noise.

[0109] Through the above steps, a dual joint search of the time and frequency axes is achieved, enabling the simultaneous capture of the joint distribution characteristics of the signal in the two-dimensional time-frequency domain, and effectively identifying the exact position of the preamble sequence at the receiver. Pre-stored preamble matching gives this method inherent noise suppression capabilities, making it particularly suitable for complex channel environments where direct paths disappear and reflection paths dominate in highly dynamic moving scenarios. It can achieve coarse synchronization accuracy at the nanosecond and 100Hz levels even with limited pilot resources, providing accurate initial estimates for subsequent fine-tuning.

[0110] Furthermore, the first signal can be compensated based on the first time offset and the first carrier frequency offset to obtain the second signal after coarse synchronization. Specifically, it can be expressed as: .

[0111] Based on this, the above steps for preliminary channel estimation based on the second signal can be performed by first using known pilot symbols. The second signal is processed to obtain the first channel state information vector. Specifically, it is expressed as: And the first channel state information tensor under the complete space-time-frequency resource is constructed by cubic spline interpolation, represented as: .exist In the process, the residual time offset undergoes a linear phase rotation along the subcarrier axis, which is difficult to separate from the time delay information. The residual carrier frequency offset undergoes a nonlinear phase rotation along the OFDM axis, which is also difficult to separate from the Doppler information. Therefore, further synchronization error elimination is required on the first channel state information tensor.

[0112] See Figure 3 Furthermore, in the uplink sensing integrated system signal processing method described above, the preprocessing and frequency domain transformation of the first signal to obtain a frequency domain signal with a preset preamble includes:

[0113] Step S301: The first signal is downsampled by matched filtering and the cyclic prefix is ​​removed to obtain the preprocessed first signal;

[0114] Step S302: The preprocessed first signal is transformed to the frequency domain by Fourier transform to obtain the frequency domain signal.

[0115] In this embodiment, the steps of preprocessing and frequency domain transformation of the first signal described above are also illustrated. Specifically, this includes downsampling through matched filtering, which effectively avoids spectral aliasing interference, suppresses background noise, extracts target signal features to the maximum extent, and avoids losing important signal features; then, removing the cyclic prefix effectively avoids inter-symbol interference (ISI) and inter-carrier interference (ICI), and helps reduce the amount of data, thereby reducing the burden of subsequent processing. Furthermore, the preprocessed first signal can be transformed to the frequency domain through Fourier transform to obtain the desired frequency domain signal.

[0116] See Figure 4 Preferably, in the uplink sensing integrated system signal processing method described above, the step of performing deep learning based on a hybrid neural network according to the first channel state information tensor to obtain the estimated second channel state information tensor, the second time offset, and the second carrier frequency offset includes:

[0117] Step S401: Denoise the first channel state information tensor to obtain a denoised third channel state information tensor.

[0118] Step S402: The first channel state information tensor and the third channel state information tensor are fused according to the learnable dynamic gating weights to obtain the second channel state information tensor.

[0119] Step S403: Estimate the carrier frequency offset based on the cross-connector to obtain the second carrier frequency offset;

[0120] Step S404: Time offset estimation is performed based on the attention mechanism to obtain the second time offset.

[0121] In this embodiment, the steps for obtaining the estimated second channel state information tensor, second time offset, and second carrier frequency offset based on deep learning using a hybrid neural network are illustrated. Firstly, the first channel state information tensor is normalized to obtain the normalized first channel state information tensor. ,in, The F-norm is used to ensure that the data input to the hybrid neural network has a standard format, making it easier for the hybrid neural network to process.

[0122] In specific processing, the first channel state information tensor is denoised. The denoising module that performs the denoising process is mainly composed of residual blocks, channel attention mechanisms (e.g., Squeeze-and-Excitation, SE), and skip connections. It is used to suppress the noise of the channel state information while preserving the phase structure carrying the key information of synchronization and perception to the greatest extent and avoiding phase distortion.

[0123] Then, the first channel state information tensor and the third channel state information tensor are fused according to the learnable dynamic gating weights to obtain the second channel state information tensor. While significantly improving the signal-to-noise ratio, it helps ensure that the phase information in the original signal, which is extremely sensitive to synchronization, is not excessively smoothed or lost.

[0124] Subsequently, the carrier frequency offset is estimated based on the crossformer to obtain the second carrier frequency offset. Compared with the standard self-attention mechanism (Transformer), the crossformer fully exploits the translation invariance of the carrier frequency offset in the spatial dimension and the global linear phase slope characteristics in the frequency domain dimension through the cross-dimensional attention mechanism, thereby capturing the time-varying law of the carrier frequency offset more accurately and significantly improving the frequency synchronization accuracy.

[0125] Subsequently, time offset estimation is performed based on the attention mechanism to obtain the second time offset. Preferably, a lightweight self-attention mechanism is used for time offset estimation. In order to avoid interference from the linear phase growth in the frequency domain to the learning of time domain features, this application retains the powerful representation capability of the standard Transformer and introduces a denormalization structure to significantly reduce the amount of computation and memory overhead while maintaining accuracy, thereby achieving higher inference efficiency.

[0126] To reduce the overfitting risk of the hybrid neural network while improving its generalization ability, a multi-task loss function is used to guide the network. This function aims to simultaneously optimize two key objectives: 1) reconstructing "clean" channel state information without residual synchronization errors, and 2) accurately estimating residual time offset and carrier frequency offset. To effectively balance these different objectives, an adaptive weighting scheme based on homoscedastic uncertainty is preferred, enabling the automatic learning of the relative contribution of each task. The overall loss function is expressed as follows:

[0127]

[0128] in, , and This represents the loss of a single task. , , Learnable parameters, representing the uncertainty of each task, are optimized along with the network parameters to dynamically weigh each loss component, with the final logarithmic term serving as a regularizer.

[0129] See Figure 5 Preferably, in the uplink sensing integrated system signal processing method described above, the step of performing tensor decomposition based on the second channel state information tensor and obtaining sensing information based on the first time offset, the first carrier frequency offset, the second time offset, and the second carrier frequency offset includes:

[0130] Step S501: Based on the tensor decomposition algorithm, the second channel state information tensor is decomposed to obtain the factor matrix, time offset vector and carrier frequency offset vector.

[0131] Step S502: Based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix, obtain the user perception information in the perception information;

[0132] Step S503: Based on the factor matrix, the time offset vector, and the carrier frequency offset vector, obtain the scatterer sensing information corresponding to each scatterer in the sensing information.

[0133] This embodiment illustrates how to acquire sensing information, which includes user sensing information and scatterer sensing information. Since the second channel state information tensor contains multipath information, when acquiring sensing information, the multi-target sensing problem is first transformed into an independent single-target sensing problem, i.e., estimating the physical parameter information (time delay, angle, and Doppler) of each path. Specifically, in this embodiment, multipath information is separated using a tensor decomposition method based on Regularized Alternating Least Squares (RALS), ensuring a one-to-one correspondence between time delay, angle, and Doppler information. The second channel state information tensor... It can be represented as: ,in, Represents the weight vector. Represents the factor matrix, Represented as inner product symbol, Let AoA be a normalized vector, and It is represented as an error vector. and These are the normalized time delay vector and Doppler vector, which are expressed as follows:

[0134]

[0135]

[0136] Therefore, the multipath signal separation problem can be expressed as:

[0137]

[0138] This multipath signal separation problem can be solved using the RALS algorithm, which decomposes the problem into four coupled subproblems:

[0139]

[0140]

[0141]

[0142]

[0143] in , and It represents the matrixing of three different ways of expanding a tensor. This represents the regularization hyperparameter. After multiple iterations, the final output is the estimated value. , , and The columns of these matrices with the same index collectively represent a multipath component. Therefore, the method described above effectively decouples different multipath signals while maintaining tight coupling of physical parameters within each path.

[0144] Then, user perception information can be obtained from the perception information based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix. This information is derived by using four offset parameters to correspond to the differences in time delay and frequency offset for each user, and by using the factor matrix to map multi-antenna measurements to the user feature space. This effectively suppresses multi-user interference and ensures the independence of each user's perception information during the estimation process through the orthogonalization condition matrix.

[0145] Based on the factor matrix, time offset vector, and carrier frequency offset vector, the scatterer sensing information corresponding to each scatterer in the sensing information is obtained. Among them, the time offset vector and carrier frequency offset vector encode the distance and radial velocity information of the scatterer, respectively, and the geometric relationship constraints of multipath propagation provided by the factor matrix can ensure the accuracy of the obtained scatterer sensing.

[0146] See Figure 6Specifically, in the uplink sensing integrated system signal processing method described above, the step of obtaining user perception information from the perception information based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix includes:

[0147] Step S601: Based on the first carrier frequency offset and the second carrier frequency offset, obtain the first radial velocity estimate of the user in the user perception information;

[0148] Step S602: Obtain the first estimated distance between the user and the base station based on the first time offset and the second time offset;

[0149] Step S603: Based on the first estimated distance, the factor matrix, and the multiple signal classification algorithm, perform angle estimation to determine the first angle estimate value;

[0150] Step S604: Based on the first angle estimate, the first estimated distance, and the location information of the base station, obtain the first estimated location of the user in the user perception information.

[0151] This embodiment illustrates the specific steps for obtaining user-perceived information. Firstly, based on the first carrier frequency offset and the second carrier frequency offset, the estimated first radial velocity of the user in the user-perceived information is obtained, which can be specifically expressed as:

[0152]

[0153] in, This represents the estimated first radial velocity.

[0154] Then, based on the first time offset and the second time offset, the first estimated distance between the user and the base station is obtained, which can be specifically expressed as: ,in, This indicates the first estimated distance.

[0155] Furthermore, based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm, angle estimation is performed to determine the first angle estimate. By using the MUSIC angle estimation scheme based on spatial dimensionality reduction using the factor matrix, the accuracy of angle measurement can be guaranteed while reducing computational overhead.

[0156] Finally, based on the first angle estimate, the first estimated distance, and the base station's location information, the user's first estimated location in the user's perceived information can be obtained through simple calculation, which can be specifically expressed as: .

[0157] See Figure 7 Specifically, in the uplink sensing integrated system signal processing method described above, the step of determining the first angle estimate value by performing angle estimation based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm includes:

[0158] Step S701: Based on the factor matrix, obtain the first angle vector of the user's direct path and the covariance matrix of the first angle vector;

[0159] Step S702: Perform eigenvalue decomposition on the covariance matrix to obtain a descending eigenvalue diagonal matrix and an orthogonal eigenvector matrix;

[0160] Step S703: Extract the noise subspace based on the orthogonal eigenvector matrix;

[0161] Step S704: Determine the spatial power function based on the noise subspace;

[0162] Step S705: Search according to the spatial power function to determine the first angle estimate.

[0163] This embodiment illustrates the angle estimation steps described above. First, the first angle vector in the factor matrix is ​​extracted for angle estimation. Since the direct path has the strongest power, in this embodiment, the index value corresponding to the maximum value in the weight vector is set to... Then the first angle vector of the direct trajectory can be used This indicates that the covariance matrix of the first angle vector can be further obtained, i.e. ,in, This represents the conjugate transpose. This covariance matrix not only preserves the cross-correlation relationships between the elements but also effectively suppresses the effects of received noise.

[0164] Then, by performing eigenvalue decomposition on the covariance matrix above, we can obtain the descending eigenvalue diagonal matrix and the orthogonal eigenvector matrix, which can be specifically represented as: ,in, This represents a diagonal matrix of eigenvalues ​​in descending order. This represents an orthogonal eigenvector matrix. Through eigenvalue decomposition, the intrinsic properties of the covariance matrix are explicitly mapped to the eigenvalue and eigenvector domains, allowing subsequent processing to be performed directly on the signal subspace, avoiding performance waste and computational redundancy caused by blind computation across the entire space.

[0165] Furthermore, based on the aforementioned orthogonal eigenvector matrix, a noise projection space, i.e., a noise subspace, is constructed that is strictly orthogonal to the signal direction. In this noise subspace, the feature vectors are orthogonal to the real steering vectors, so that the subsequent search function can be constructed based on this orthogonality, which can ensure that the peak position of the angle spectrum corresponds to the real target.

[0166] Based on the noise subspace, the spatial power function is determined, which can be specifically expressed as: ,in, ; By constructing a spectral function, namely a spatial power function, that is highly sensitive to the true direction of arrival of the wave and minimizes other directions, it becomes possible to distinguish multiple closely spaced target sources in space, thereby ensuring the accuracy of subsequent searches.

[0167] The first angle estimate is determined by searching based on the spatial power function. On the one hand, it combines the orthogonal constraints of the aforementioned noise subspace to achieve high-precision angle resolution; on the other hand, it completes the discretization index mapping from the continuous function domain through spectral peak localization, which facilitates subsequent system calls. Furthermore, since the peak positions of the MUSIC spectrum have good robustness, the reliability of the final estimate is guaranteed.

[0168] See Figure 8 Specifically, in the uplink sensing integrated system signal processing method described above, obtaining the scatterer sensing information corresponding to each scatterer in the sensing information based on the factor matrix, the time offset vector, and the carrier frequency offset vector includes:

[0169] Step S801: Obtain the second angle vector corresponding to the target scatterer according to the factor matrix, wherein the target scatterer is any scatterer in the scene;

[0170] Step S802: Estimate the second angle vector according to the multiple signal classification algorithm to obtain the second angle estimate value;

[0171] Step S803: Determine the second estimated distance from the target scatterer to the base station based on the geometric relationship between the user, the target scatterer, and the base station and the first estimated distance;

[0172] Step S804: Determine the reference position estimate of the target scatterer based on the location information of the base station, the second estimated distance, and the second estimated angle value;

[0173] Step S805: Optimize the reference position estimate according to a preset optimization algorithm to obtain the second estimated position of the target scatterer.

[0174] In this embodiment, since the decoupling of multi-target perception has already been achieved through tensor decomposition, an example of the specific steps for obtaining scatterer perception information is given by taking any scatterer as the target scatterer.

[0175] The angle vector, time offset vector, and carrier frequency offset vector corresponding to the target scatterer can be expressed as follows: , and Then, following the steps described above for obtaining the user angle vector based on the factor matrix and the MUSIC method, the second angle estimate corresponding to the target scatterer can be obtained. Time offset estimate and carrier frequency offset estimate And because This is the difference between the time delay from the user to the scatterer and then to the base station and the time delay from the user to the base station. Therefore, the distance from the user to the scatterer and then to the base station can be expressed as: At this point, based on the geometric relationship between the user, the base station, and the target scatterer, a second estimated distance from the target scatterer to the base station can be determined. Specifically, it can be expressed as:

[0176]

[0177] Based on this, the reference position estimate of the target scatterer can be determined according to the location information of the base station, the second estimated distance, and the second angle estimate. Specifically, it can be expressed as:

[0178]

[0179] Finally, by optimizing the above-obtained reference position estimate using a preset optimization algorithm, the second estimated position of the target scatterer can be obtained.

[0180] In one embodiment, the optimization function in a pre-defined optimization algorithm (e.g., the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method in quasi-Newton algorithms) can be expressed as:

[0181]

[0182] By using the reference position obtained above as the starting point for the search, the final second estimated position can be obtained.

[0183] See Figure 9 Preferably, the uplink sensing integrated system signal processing method described above further includes:

[0184] Step S901: Based on the minimum mean square error, the second channel state information tensor, the second time offset, and the second carrier frequency offset, fine synchronization is performed on the second signal to obtain the fourth signal;

[0185] Step S902: Demodulate the fourth signal to determine the target symbol, wherein the Euclidean distance between the target symbol and the fourth signal is minimized;

[0186] Step S903: Perform data recovery on the target symbol to obtain the transmitted data.

[0187] In this embodiment, the application of the second time offset and the second carrier frequency offset obtained above in communication performance is also illustrated. Firstly, the second signal is compensated based on the second time offset and the second carrier frequency offset to obtain the compensated fifth signal, which can be specifically expressed as follows:

[0188]

[0189] in, This indicates the fifth signal after compensation. It is the second channel state information tensor Fibers It is a noise vector.

[0190] Then, by further compensation based on the weight matrix of the Minimum Mean Square Error (MMSE) and the fifth signal, the final fourth signal can be obtained. In this embodiment, the weight matrix of the MMSE can be expressed as:

[0191]

[0192] in, It is an identity matrix; It is a regularization factor; It is noise power, which is expressed as Finally, the equalized fourth signal can be expressed as: .

[0193] Then, in this embodiment, it is assumed that the residual noise after equalization is additive white Gaussian noise (AWGN), and the maximum likelihood (ML) detection problem in data demodulation is simplified to finding constellation symbols. ,in, To modulate the set of constellation points (e.g., Quadrature Phase-Shift Keying (QPSK), 16-QAM), such that the probability... Minimization, which is equivalent to finding the target symbol with the minimum Euclidean distance to the fourth signal, can be calculated as follows: .

[0194] Finally, the estimated target symbol By mapping the data to the corresponding binary sequence and recovering the data, the transmitted data can be generated. By performing communication demodulation based on the aforementioned "clean" second channel state information, second time offset, and second carrier frequency offset, highly reliable communication performance can be obtained.

[0195] join Figure 10 Another embodiment of this application also provides a control device, including:

[0196] The coarse synchronization module 1001 is used to coarsely synchronize the first signal received by multiple antennas to obtain a first time offset, a first carrier frequency offset, and a coarsely synchronized second signal.

[0197] The first channel estimation module 1002 is used to perform preliminary channel estimation based on the second signal to obtain an estimated first channel state information tensor.

[0198] The second channel estimation module 1003 is used to perform deep learning based on a hybrid neural network based on the first channel state information tensor to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset.

[0199] The perception information acquisition module 1004 is used to perform tensor decomposition based on the second channel state information tensor, and acquire perception information based on the first time offset, the first carrier frequency offset, the second time offset and / or the second carrier frequency offset.

[0200] Specifically, in the control device described above, the coarse synchronization module includes:

[0201] The first processing unit is used to preprocess and frequency domain transform the first signal to obtain a frequency domain signal with a preset preamble.

[0202] The second processing unit is used to perform sliding correlation based on the frequency domain signal and the pre-stored preamble information to obtain the first time offset and the first carrier frequency offset.

[0203] The third processing unit is used to compensate the first signal according to the first time offset and the first carrier frequency offset to obtain the second signal.

[0204] Furthermore, in the control device described above, the first sub-processing module includes:

[0205] The first processing unit is used to downsample the first signal by matched filtering and remove the cyclic prefix to obtain the preprocessed first signal;

[0206] The second processing unit is used to transform the preprocessed first signal to the frequency domain through Fourier transform to obtain the frequency domain signal.

[0207] Preferably, in the control device described above, the second channel estimation module includes:

[0208] The fourth sub-processing module is used to denoise the first channel state information tensor to obtain a denoised third channel state information tensor.

[0209] The fifth sub-processing module is used to fuse the first channel state information tensor and the third channel state information tensor according to the learnable dynamic gating weights to obtain the second channel state information tensor.

[0210] The sixth sub-processing module is used to estimate the carrier frequency offset based on the cross-converter to obtain the second carrier frequency offset;

[0211] The seventh sub-processing module is used to estimate the time offset based on the attention mechanism to obtain the second time offset.

[0212] Preferably, in the control device described above, the sensing information acquisition module includes:

[0213] The eighth sub-processing module is used to decompose the second channel state information tensor based on the tensor decomposition algorithm to obtain the factor matrix, time offset vector and carrier frequency offset vector.

[0214] The ninth sub-processing module is used to obtain user perception information from the perception information based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix.

[0215] The tenth sub-processing module is used to obtain the scatterer sensing information corresponding to each scatterer in the sensing information based on the factor matrix, the time offset vector, and the carrier frequency offset vector.

[0216] Specifically, in the control device described above, the ninth sub-processing module includes:

[0217] The third processing unit is used to obtain the first radial velocity estimate of the user in the user perception information based on the first carrier frequency offset and the second carrier frequency offset.

[0218] The fourth processing unit is used to obtain a first estimated distance between the user and the base station based on the first time offset and the second time offset;

[0219] The fifth processing unit is used to perform angle estimation based on the first estimated distance, the factor matrix, and the multiple signal classification algorithm, and determine the first angle estimate value.

[0220] The sixth processing unit is used to obtain the first estimated position of the user in the user perception information based on the first angle estimate, the first estimated distance, and the location information of the base station.

[0221] Specifically, in the control device described above, the fifth processing unit includes:

[0222] The first sub-processing unit is used to obtain the first angle vector of the user's direct path and the covariance matrix of the first angle vector according to the factor matrix.

[0223] The second sub-processing unit is used to perform eigenvalue decomposition on the covariance matrix to obtain a descending eigenvalue diagonal matrix and an orthogonal eigenvector matrix.

[0224] The third sub-processing unit is used to extract the noise subspace based on the orthogonal eigenvector matrix;

[0225] The fourth sub-processing unit is used to determine the spatial power function based on the noise subspace;

[0226] The fifth sub-processing unit is used to search based on the spatial power function to determine the first angle estimate.

[0227] Specifically, in the control device described above, the tenth sub-processing module includes:

[0228] The seventh processing unit is used to obtain the second angle vector corresponding to the target scatterer according to the factor matrix, wherein the target scatterer is any scatterer in the scene;

[0229] The eighth processing unit is used to estimate the second angle vector according to the multi-signal classification algorithm to obtain the second angle estimate value;

[0230] The ninth processing unit is configured to determine a second estimated distance from the target scatterer to the base station based on the geometric relationship between the user, the target scatterer, and the base station and the first estimated distance;

[0231] The tenth processing unit is used to determine the reference position estimate of the target scatterer based on the location information of the base station, the second estimated distance, and the second estimated angle value;

[0232] The eleventh processing unit is used to optimize the reference position estimate according to a preset optimization algorithm to obtain the second estimated position of the target scatterer.

[0233] Preferably, the control device described above further includes:

[0234] The first processing module is used to perform fine synchronization on the second signal based on the minimum mean square error, the second channel state information tensor, the second time offset, and the second carrier frequency offset to obtain the fourth signal.

[0235] The second processing module is used to demodulate the fourth signal and determine the target symbol, wherein the Euclidean distance between the target symbol and the fourth signal is minimized.

[0236] The third processing module is used to recover data from the target symbol to obtain the transmitted data.

[0237] 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 provided in this application embodiments can implement all the method steps implemented in the method embodiments described above and can achieve the same technical effects. Therefore, the parts and beneficial effects that are the same as those in the method embodiments in this embodiment will not be described in detail here.

[0238] 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 uplink sensing integrated system signal processing method described above and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0239] Another embodiment of this application provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the uplink sensing integrated system signal processing method described above, achieving the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0240] Another embodiment of this application provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the uplink sensing integrated system signal processing method described above, and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0241] 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.

[0242] 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.

[0243] 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 signal processing method for an integrated uplink sensing system, characterized in that, include: Coarsely synchronize the first signal received by multiple antennas to obtain the first time offset, the first carrier frequency offset, and the coarsely synchronized second signal; Based on the second signal, a preliminary channel estimation is performed to obtain the estimated first channel state information tensor; Based on the first channel state information tensor, deep learning based on a hybrid neural network is performed to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset. Tensor decomposition is performed based on the second channel state information tensor, and sensing information is obtained based on the first time offset, the first carrier frequency offset, the second time offset, and / or the second carrier frequency offset.

2. The uplink sensing integrated system signal processing method according to claim 1, characterized in that, The step of coarsely synchronizing the first signal received by multiple antennas to obtain a first time offset, a first carrier frequency offset, and a coarsely synchronized second signal includes: The first signal is preprocessed and frequency domain transformed to obtain a frequency domain signal with a preset preamble; Sliding correlation is performed based on the frequency domain signal and the pre-stored preamble information to obtain the first time offset and the first carrier frequency offset. The first signal is compensated based on the first time offset and the first carrier frequency offset to obtain the second signal.

3. The uplink sensing integrated system signal processing method according to claim 2, characterized in that, The step of preprocessing and frequency domain transformation of the first signal to obtain a frequency domain signal with a preset preamble includes: The first signal is downsampled by matched filtering and the cyclic prefix is ​​removed to obtain the preprocessed first signal; The preprocessed first signal is transformed to the frequency domain by Fourier transform to obtain the frequency domain signal.

4. The uplink sensing integrated system signal processing method according to claim 1, characterized in that, The step of performing deep learning based on a hybrid neural network on the first channel state information tensor to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset includes: The first channel state information tensor is denoised to obtain the denoised third channel state information tensor. The first channel state information tensor and the third channel state information tensor are fused according to the learnable dynamic gating weights to obtain the second channel state information tensor. The second carrier frequency offset is obtained by estimating the carrier frequency offset based on the cross-converter. The second time offset is obtained by estimating the time offset based on the attention mechanism.

5. The uplink sensing integrated system signal processing method according to claim 1, characterized in that, The step of performing tensor decomposition based on the second channel state information tensor and obtaining sensing information based on the first time offset, the first carrier frequency offset, the second time offset, and the second carrier frequency offset includes: Based on the tensor decomposition algorithm, the second channel state information tensor is decomposed to obtain the factor matrix, time offset vector and carrier frequency offset vector. Based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix, the user perception information in the perception information is obtained; Based on the factor matrix, the time offset vector, and the carrier frequency offset vector, obtain the scatterer sensing information corresponding to each scatterer in the sensing information.

6. The uplink sensing integrated system signal processing method according to claim 5, characterized in that, The step of obtaining user perception information from the perception information based on the first time offset, the first carrier frequency offset, the second time offset, the second carrier frequency offset, and the factor matrix includes: Based on the first carrier frequency offset and the second carrier frequency offset, obtain the first radial velocity estimate of the user in the user perception information; Based on the first time offset and the second time offset, the first estimated distance between the user and the base station is obtained; An angle estimate is determined by performing angle estimation based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm; Based on the first angle estimate, the first estimated distance, and the location information of the base station, the first estimated location of the user in the user perception information is obtained.

7. The uplink sensing integrated system signal processing method according to claim 6, characterized in that, The step of estimating the angle based on the first estimated distance, the factor matrix, and the multi-signal classification algorithm to determine the first angle estimate includes: Based on the factor matrix, obtain the first angle vector of the user's direct path and the covariance matrix of the first angle vector; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain a descending diagonal matrix of eigenvalues ​​and an orthogonal eigenvector matrix. Based on the orthogonal eigenvector matrix, the noise subspace is extracted; Based on the noise subspace, determine the spatial power function; The first angle estimate is determined by searching based on the spatial power function.

8. The uplink sensing integrated system signal processing method according to claim 6, characterized in that, The step of obtaining scatterer sensing information corresponding to each scatterer in the sensing information based on the factor matrix, the time offset vector, and the carrier frequency offset vector includes: Based on the factor matrix, obtain the second angle vector corresponding to the target scatterer, where the target scatterer is any scatterer in the scene; The second angle vector is estimated according to the multi-signal classification algorithm to obtain the second angle estimate value; Based on the geometric relationship between the user, the target scatterer, and the base station, and the first estimated distance, a second estimated distance from the target scatterer to the base station is determined; Based on the location information of the base station, the second estimated distance, and the second estimated angle value, the reference position estimate of the target scatterer is determined; The reference position estimate is optimized according to a preset optimization algorithm to obtain the second estimated position of the target scatterer.

9. The uplink sensing integrated system signal processing method according to claim 1, characterized in that, Also includes: The second signal is finely synchronized based on the minimum mean square error, the second channel state information tensor, the second time offset, and the second carrier frequency offset to obtain the fourth signal. The fourth signal is demodulated to determine the target symbol, and the Euclidean distance between the target symbol and the fourth signal is minimized. Data recovery is performed on the target symbol to obtain the transmitted data.

10. A control device, characterized in that, include: The coarse synchronization module is used to coarsely synchronize the first signal received by multiple antennas to obtain the first time offset, the first carrier frequency offset, and the coarsely synchronized second signal. The first channel estimation module is used to perform preliminary channel estimation based on the second signal to obtain the estimated first channel state information tensor. The second channel estimation module is used to perform deep learning based on a hybrid neural network based on the first channel state information tensor to obtain the estimated second channel state information tensor, second time offset, and second carrier frequency offset. The sensing information acquisition module is used to perform tensor decomposition based on the second channel state information tensor, and to acquire sensing information based on the first time offset, the first carrier frequency offset, the second time offset and / or the second carrier frequency offset.

11. 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 uplink sensing integrated system signal processing method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when executed by a processor, the computer program implements the steps of the uplink sensing integrated system signal processing method as described in any one of claims 1 to 9.

13. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the uplink sensing integrated system signal processing method as described in any one of claims 1 to 9.

Images