Multi-modal positioning method and system based on intelligent wearable device

Abstract

The invention relates to the technical field of equipment positioning, and discloses a multi-mode positioning method and system based on intelligent wearable equipment, and the method comprises the steps: receiving a Beidou satellite signal and a GPS satellite signal, and executing the sparse channel estimation, and obtaining a channel parameter; identifying and suppressing multipath interference according to the channel parameter to obtain a purified signal; separating the purification signal into Beidou observation data and GPS observation data, and carrying out cross correction calculation to obtain position information; and performing channel state prediction based on the position information and the channel parameters, and compensating positioning calculation at the next moment to obtain a target positioning result and a communication transmission scheme. According to the invention, the complementary advantages of two sets of Beidou and GPS systems are fully utilized, the GPS clock error offset is corrected by using the time service precision of Beidou, the positioning precision of the intelligent wearable device is improved, and the power consumption is reduced.

Description

Technical Field

[0001] This invention relates to the field of device positioning technology, and in particular to a multimodal positioning method and system based on smart wearable devices. Background Technology

[0002] Due to physical constraints such as small size, limited power consumption, and finite antenna size, smart wearable devices face technical challenges including weak signal reception, low positioning accuracy, and short battery life. Traditional single-satellite positioning systems are easily affected by factors such as signal blockage, multipath interference, and ionospheric delay in complex environments. They are significantly inadequate in handling the channel attenuation, multipath propagation, and dynamic blockage issues unique to wearable devices, lacking targeted channel modeling and adaptive processing mechanisms. Furthermore, existing technologies separate positioning and communication functions, failing to dynamically adjust communication strategies based on positioning accuracy and channel quality, resulting in low energy efficiency. Summary of the Invention

[0003] This invention provides a multimodal positioning method and system based on smart wearable devices. This invention makes full use of the complementary advantages of the BeiDou and GPS systems, using the timing accuracy of BeiDou to correct the clock bias of GPS, thereby improving the positioning accuracy of smart wearable devices and reducing power consumption.

[0004] In a first aspect, the present invention provides a multimodal positioning method based on a smart wearable device, the multimodal positioning method based on a smart wearable device comprising: The system receives BeiDou and GPS satellite signals and performs sparse channel estimation to obtain channel parameters. Based on the channel parameters, multipath interference is identified and suppressed to obtain a purified signal; The purified signal is separated into BeiDou observation data and GPS observation data, and cross-correction calculation is performed to obtain the location information; Based on the location information and the channel parameters, channel state prediction is performed and the positioning calculation for the next time step is compensated to obtain the target positioning result and communication transmission scheme.

[0005] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, the step of receiving BeiDou satellite signals and GPS satellite signals and performing sparse channel estimation to obtain channel parameters includes: Receive BeiDou satellite signals and GPS satellite signals, and calculate the BeiDou signal attenuation coefficient and the GPS signal attenuation coefficient respectively; The BeiDou satellite signal is attenuated based on the BeiDou signal attenuation coefficient to obtain the BeiDou received signal, and the GPS satellite signal is attenuated based on the GPS signal attenuation coefficient to obtain the GPS received signal. The BeiDou received signal and the GPS received signal are combined to obtain a combined energy signal; The motion intensity index is calculated based on the three-axis acceleration data of the smart wearable device. When the motion intensity index exceeds the intensity index threshold, the Beidou propagation delay and GPS propagation delay in the merged energy signal are compensated respectively to obtain the dual constellation signal. Sparse channel estimation is performed on the dual-constellation signal to obtain the channel parameters.

[0006] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, the step of performing sparse channel estimation on the dual-constellation signal to obtain channel parameters includes: Compressed observation data for the smart wearable device is constructed based on the dual-constellation signals. The noise variance is calculated based on the compressed observation data and a noise tolerance threshold is set. At the same time, the regularization intensity coefficient is calculated based on the current signal-to-noise ratio. A solution model is established based on the noise variance, the noise tolerance threshold, and the regularization intensity coefficient. Calculate the sparse channel impulse response solution based on the solution model; The sparse channel impulse response solution is weighted and fused with the channel impulse response at the previous time step to obtain the channel parameters.

[0007] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, the step of calculating the sparse channel impulse response solution based on the solution model includes: The iteration step size and the current solution vector are initialized based on the solution model, and the iteration step size and the current solution vector are used as the initial solution. At the same time, the convergence threshold of the smart wearable device is set. Substitute the current solution vector into the objective function in the solution model to calculate the gradient direction, and update the current solution vector along the negative gradient direction according to the iteration step size to obtain the gradient-updated solution. A soft threshold shrinkage operation is performed on each component of the gradient update solution to obtain the shrunken solution vector; Calculate the normalized difference between the shrunken solution vector and the solution vector of the previous iteration. When the normalized difference is less than the convergence threshold, terminate the iteration and output the sparse channel impulse response solution. Otherwise, use the shrunken solution vector as the new current solution to continue the iteration.

[0008] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, the step of identifying and suppressing multipath interference based on the channel parameters to obtain a cleaned signal includes: Based on the channel parameters, a satellite visibility adjacency matrix and a degree matrix are constructed. The satellite visibility adjacency matrix is ​​used to determine the connection weights according to the spatial geometric distribution between satellites, and the degree matrix reflects the spatial association strength of each satellite node. A satellite spatial relationship graph structure is created based on the connection weights and the spatial association strength; The satellite spatial relationship graph structure is input into the graph convolutional layer for spatial feature extraction to obtain spatial channel features. At the same time, the channel parameters are input into the gated temporal convolutional layer for temporal feature extraction to obtain temporal channel features. The spatial channel features and the temporal channel features are fused to obtain a fused feature vector; The multipath interference probability is calculated based on the fused feature vector. When the multipath interference probability exceeds a preset probability threshold, multipath component suppression is performed on the channel parameters to obtain a clean signal.

[0009] In conjunction with the first aspect, in the fifth implementation of the first aspect of the present invention, the step of separating the purified signal into BeiDou observation data and GPS observation data and performing cross-correction calculation to obtain location information includes: BeiDou observation data is extracted from the purified signal, and GPS observation data is also extracted from the purified signal. The timing accuracy advantage of the BeiDou system is used to correct the receiver clock offset in the GPS observation data, and the geometric configuration advantage of the GPS system is used to correct the geometric dilution accuracy of the BeiDou observation data, thus establishing a set of observation equations. The adaptive factor is calculated based on the residual change magnitude in the observation equation set, and an adaptive filter parameter set is generated based on the adaptive factor. Based on the adaptive filtering parameter set, a prediction and update loop of capacitive Kalman filtering is performed to finally calculate the location information.

[0010] In conjunction with the first aspect, in the sixth implementation of the first aspect of the present invention, the step of performing a prediction and update loop of capacitive Kalman filtering based on the adaptive filtering parameter set to finally calculate the location information includes: Based on the set of adaptive filtering parameters, volumetric sampling points are generated and propagated to the prediction time through state transition equations that include uniform motion model and random walk clock difference model, forming volumetric point distribution data. Based on the volume point distribution data, the mean of state prediction is statistically calculated, and the adaptive factor is adjusted by the hyperbolic tangent function according to the degree of residual exceeding the limit to obtain the prior state statistical parameters. Substitute the prior state statistical parameters into the observation equation set to calculate the theoretical observation values ​​corresponding to each volume sampling point, and dynamically update the observation noise covariance matrix according to the signal quality to form the observation space statistical parameters. The optimal Kalman gain is calculated using the observed spatial statistical parameters, and the prior state statistical parameters are corrected based on the optimal Kalman gain to solve for the location information.

[0011] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, the step of performing channel state prediction and compensating for the positioning calculation at the next moment based on the location information and the channel parameters to obtain the target positioning result and the communication transmission scheme includes: The channel change rate is calculated based on the time-domain changes of the channel parameters. At the same time, the change in the geometric precision dilution factor is extracted from the location information, and the remaining battery power ratio of the smart wearable device is obtained. The channel change rate, the change in the geometric precision dilution factor, and the remaining battery power ratio are weighted and summed to obtain a weighted result. When the weighted result exceeds the trigger threshold, the transmission decision function value is calculated. The channel adaptation coefficient and power adaptation coefficient of the base transmit power are adjusted according to the transmission decision function value, and the dynamic transmit power is determined based on the channel adaptation coefficient and the power adaptation coefficient. The time interval for intermittent communication is calculated based on the reciprocal relationship between the dynamic transmission power and the transmission decision function, thereby obtaining the transmission power parameter and the communication interval parameter. The transmission power parameter and the communication interval parameter are then combined to generate the transmission parameter. Based on the transmission parameters, channel state prediction is performed and the positioning calculation for the next moment is compensated to obtain the target positioning result and communication transmission scheme.

[0012] In conjunction with the first aspect, in the eighth implementation of the first aspect of the present invention, the step of performing channel state prediction and compensating for the positioning calculation at the next moment based on the transmission parameters to obtain the target positioning result and the communication transmission scheme includes: The communication interval parameter in the transmission parameters is used as the prediction step size, and the channel parameters are extrapolated and predicted using a second-order autoregressive integral moving average algorithm to obtain the predicted channel state at the next time step. Based on the predicted channel state, the cumulative deviation of the prediction error is corrected to obtain the corrected predicted channel parameters. At the same time, the spatial position of the smart wearable device at the next moment is calculated based on the location information to obtain the predicted trajectory position. Based on the corrected predicted channel parameters, the BeiDou observation data and the GPS observation data are weighted and re-executed with capacitive Kalman filtering to obtain the target positioning result. The optimal communication timing for the next cycle is calculated based on the predicted trajectory position and the transmit power parameter in the transmission parameters, and the communication transmission scheme is determined by combining the predicted channel state and the optimal communication timing.

[0013] Secondly, the present invention provides a multimodal positioning system based on a smart wearable device, the multimodal positioning system based on a smart wearable device comprising: The sparse channel estimation module is used to receive BeiDou satellite signals and GPS satellite signals and perform sparse channel estimation to obtain channel parameters. A multipath interference suppression module is used to identify and suppress multipath interference based on the channel parameters to obtain a clean signal; The cross-correction calculation module is used to separate the purified signal into BeiDou observation data and GPS observation data and perform cross-correction calculation to obtain location information; The channel state prediction module is used to perform channel state prediction based on the location information and the channel parameters and compensate for the positioning calculation at the next moment to obtain the target positioning result and communication transmission scheme.

[0014] The technical solution provided by this invention significantly reduces the computational complexity of channel estimation by using a compressed sensing algorithm specifically designed for wearable devices, controlling the sampling rate to one-quarter of the original signal length, thus ensuring estimation accuracy while meeting the power consumption constraints of the device. The dual-channel attention-enhanced graph convolutional network can simultaneously capture satellite spatial geometric information and channel temporal changes, achieving accurate identification and suppression of multipath interference. The network parameters are reduced by more than one-third compared to traditional methods, resulting in a significant improvement in processing speed. The BeiDou-GPS cross-correction mechanism fully utilizes the complementary advantages of the two systems, using BeiDou's timing accuracy to correct GPS clock offsets, while leveraging GPS's geometric configuration advantages to improve BeiDou positioning, overcoming the performance limitations of a single system. The adaptive capacitive Kalman filter can automatically adjust the covariance parameter based on changes in the observation residuals, adapting to the complex and ever-changing motion environment of wearable devices. The channel prediction compensation mechanism uses an ARIMA model to proactively predict future channel states and combines this with device trajectory prediction to optimize the positioning solution for the next moment, improving the system's continuous positioning capability. This invention integrates positioning accuracy, channel quality, and remaining power into the transmission decision, achieving deep integration of positioning and communication. Through intelligent power adjustment and intermittent communication strategies, it significantly reduces power consumption while ensuring positioning performance.

[0015] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of one embodiment of the multimodal positioning method based on a smart wearable device according to the present invention; Figure 2 This is a schematic diagram of one embodiment of a multimodal positioning system based on a smart wearable device according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0020] To facilitate understanding of this embodiment, a multimodal positioning method based on a smart wearable device, as disclosed in this embodiment of the invention, will first be described in detail. For example... Figure 1 As shown, this method includes the following steps: 101. Receive BeiDou satellite signals and GPS satellite signals and perform sparse channel estimation to obtain channel parameters; Specifically, a dual-constellation signal reception and processing model is established in the smart wearable device. The device simultaneously receives BeiDou and GPS satellite signals through a miniaturized antenna. The signal strength is detected and quantified using front-end circuitry in the receiving link. Then, the attenuation coefficients of the BeiDou and GPS signals are calculated based on the environmental noise levels of different frequency bands, thus modeling the actual signal attenuation after human body obstruction and antenna size reduction. Attenuation compensation processing is performed on the original BeiDou satellite signal based on the BeiDou signal attenuation coefficient. The processed BeiDou received signal reflects the loss characteristics in the actual propagation path. Similarly, attenuation compensation is performed on the original GPS satellite signal based on the GPS signal attenuation coefficient to obtain the corresponding GPS received signal. The two independently corrected signals are fused at the energy level by a merging module to obtain a merged energy signal. Since the smart wearable device is in motion, motion intensity indicators are calculated in real time using data from a three-axis accelerometer. When the motion intensity indicator exceeds a preset intensity threshold, it indicates that the device is in a state of vigorous motion. Dynamic compensation is then performed on the BeiDou and GPS propagation delays in the merged energy signal to eliminate propagation path delay deviations caused by motion, resulting in a corrected dual-constellation signal. Based on the dual-constellation signal, a sparse channel estimation algorithm is executed. High-quality channel impulse response is reconstructed under finite sampling conditions using compressed sensing and sliding window iteration mechanism, and complete channel parameters containing multipath information, noise features and path gain are extracted.

[0021] 102. Based on channel parameters, identify and suppress multipath interference to obtain a cleaned signal; Specifically, a graph model reflecting the spatial distribution of satellites is constructed based on channel parameters. Based on the geometrical relationships between different satellites, connection weights are determined according to the correlation between spatial angle differences and signal strength, and these weights are filled into the satellite visibility adjacency matrix. Simultaneously, a degree matrix is ​​generated based on the cumulative connection weights of each satellite with other satellites, quantifying the association strength of each satellite node in the spatial geometry. Using the adjacency matrix and degree matrix, a satellite spatial relationship graph structure is created, allowing the entire satellite system to be modeled as a network topology capable of graph operations within the device. The satellite spatial relationship graph structure is input into a graph convolutional layer. During convolution, the spatial features between satellite nodes are extracted through weighted aggregation of the adjacency matrix, yielding spatial channel features that reflect the geometrical distribution of satellites. Simultaneously, channel parameters are input into a gated temporal convolutional layer. The combination of temporal convolution and gating mechanisms captures the dynamic characteristics of channel evolution over time, outputting temporal channel features that reflect the signal's temporal variation. A feature fusion mechanism combines spatial and temporal channel features in a unified vector space to obtain a fused feature vector. The multipath interference probability is calculated based on the fused feature vector. When the multipath interference probability value exceeds the preset probability threshold, it indicates that there is a strong multipath component in the currently received signal. Then, the corresponding multipath component in the channel parameters is suppressed to remove redundant paths and reflected signals, so that the remaining signal can be closer to the real direct wave propagation state and a clean signal is obtained.

[0022] 103. The purified signal is separated into BeiDou observation data and GPS observation data, and cross-correction calculation is performed to obtain the location information; Specifically, the signal processing module separates the purified signal, extracting pseudorange and Doppler observations belonging to the BeiDou system as BeiDou observation data, and extracting pseudorange and Doppler observations belonging to the GPS system as GPS observation data. Leveraging the BeiDou system's advantage in timing accuracy, the system corrects for receiver clock bias commonly found in GPS observation data, ensuring higher reliability and consistency in the time dimension. Simultaneously, utilizing the wide coverage of GPS satellites in terms of geometric distribution, the system corrects for the geometric dilution accuracy issues that easily occur in BeiDou observation data. Based on the corrected data, a set of observation equations is established, including pseudorange and Doppler equations, reflecting both time correction and spatial optimization constraints. During the solution process, the residual variation of the equation set is monitored. Significant residual variations indicate the presence of dynamic errors in the data. Therefore, an adaptive factor is calculated through residual analysis, and a set of adaptive filter parameters is generated to dynamically adjust the state noise covariance and observation noise covariance of the filter. Within the capacitive Kalman filter framework, an cyclical calculation process of prediction and update is performed using an adaptive filter parameter set. The state prior estimate is obtained through the prediction stage, and the state correction is achieved by combining new observations through the update stage. The process gradually converges and outputs position information in multiple iterations.

[0023] 104. Based on location information and channel parameters, perform channel state prediction and compensate for the positioning solution at the next moment to obtain the target positioning result and communication transmission scheme.

[0024] Specifically, the channel change rate is calculated based on the channel response difference between two consecutive observation periods, and the change in the geometric precision dilution factor is derived from the location information. Simultaneously, the remaining battery power ratio of the smart wearable device is read from the power management module. The channel change rate, geometric dilution change, and battery power ratio are normalized and weighted to form a trigger value. When the trigger value exceeds a trigger threshold, adaptive transmission decision calculation is initiated to obtain the transmission decision function value. At the power control end, based on the transmission decision function, the channel adaptation coefficient and battery power adaptation coefficient corresponding to the base transmit power are adjusted. The channel adaptation coefficient periodically amplifies or suppresses with channel changes to match fast fading, while the battery power adaptation coefficient monotonically shrinks with remaining battery power to protect the battery, thus determining the dynamic transmit power in real time. At the time-domain scheduling end, the reciprocal relationship of the transmission decision function is used to adaptively stretch the time interval of intermittent communication. When the channel is stable and energy is sufficient, the interval is shortened to improve data freshness, while it is appropriately extended when the channel fluctuates rapidly or the battery power is low to reduce the duty cycle. Simultaneously, transmit power parameters and communication interval parameters are obtained, and these are combined into transmittable parameters. Based on the latest transmission parameters, the channel state predictor performs short-term extrapolation of the channel gain and phase drift at the next moment, and transforms the extrapolation result into adaptive compensation for the positioning observation noise and weight matrix. This enables the cross-positioning and filtering update at the next moment to obtain more stable residuals and faster convergence speed without increasing the computational burden. The compensated target positioning result and the matching communication transmission scheme are then output.

[0025] In one specific embodiment, the process of performing step 101 may specifically include the following steps: Receive BeiDou satellite signals and GPS satellite signals, and calculate the BeiDou signal attenuation coefficient and the GPS signal attenuation coefficient respectively; The BeiDou satellite signal is attenuated based on the BeiDou signal attenuation coefficient to obtain the BeiDou received signal, and the GPS satellite signal is attenuated based on the GPS signal attenuation coefficient to obtain the GPS received signal. The BeiDou and GPS received signals are combined to obtain a combined energy signal. The motion intensity index is calculated based on the triaxial acceleration data of smart wearable devices. When the motion intensity index exceeds the intensity index threshold, the propagation delay of Beidou and GPS in the merged energy signal is compensated to obtain the dual constellation signal. Sparse channel estimation is performed on the dual-constellation signal to obtain the channel parameters.

[0026] Specifically, during operation, smart wearable devices simultaneously receive BeiDou and GPS satellite signals via a miniaturized antenna array, and perform basic signal processing operations such as bandpass filtering, low-noise amplification, and down-conversion through front-end RF circuitry. Due to the small size of wearable devices, limited battery capacity, and the fact that they are often obstructed by the wearer's body, the received signals experience power attenuation and multipath loss during propagation. The attenuation coefficients for BeiDou and GPS signals are calculated separately. Based on the ratio of the received signal strength to the reference transmitted power, and combined with antenna gain, propagation path loss models, and environmental noise levels, dynamic estimation is performed to obtain the attenuation coefficients. Attenuation compensation is applied to the original BeiDou satellite signal based on the BeiDou signal attenuation coefficient, ensuring that the compensated BeiDou received signal reflects the effective propagation characteristics from the satellite to the terminal. Attenuation processing is also applied to the original GPS satellite signal based on the GPS signal attenuation coefficient, resulting in a GPS received signal with equivalent compensation. The BeiDou and GPS received signals are then fused at the energy level. A merged energy signal is generated by summing the squares of the amplitudes of the BeiDou and GPS received signals and combining this with a time synchronization strategy. Motion intensity indicators are calculated based on data from the built-in three-axis accelerometer. Motion features are extracted by performing modulus calculations on the rate of change of triaxial acceleration and combining the mean and standard deviation of a sliding window. When the motion intensity index exceeds a preset intensity threshold, it indicates that the device is currently in a state of violent motion, introducing additional propagation delay errors into the propagation path. Therefore, a dynamic compensation mechanism is automatically triggered to correct the BeiDou and GPS propagation delays in the merged energy signal. The compensation method adopts a strategy combining prediction and correction. For example, it combines short-term prediction and pseudorange correction algorithms for inertial navigation to adjust the time delay in the dual-constellation signal in real time, thus obtaining the dual-constellation signal. Sparse channel estimation is performed on the dual-constellation signal. Due to the limited computing resources of smart wearable devices, large-scale least squares methods or complex batch processing are not suitable. Therefore, a sparse reconstruction method based on compressed sensing theory is adopted. A downsampling measurement matrix is ​​constructed to reduce computation and energy consumption by reducing the sampling rate. Then, the L1 norm minimization method is used to recover the channel impulse response under the residual constraint. At the same time, the results are updated with a sliding window using historical channel state prediction weights to ensure that the estimation process can converge to a relatively stable solution under limited computing power. After sparse channel estimation, the device outputs channel parameters that include multipath components, channel gain, and noise statistics.

[0027] In one specific embodiment, the process of performing sparse channel estimation on the dual-constellation signal to obtain channel parameters may specifically include the following steps: Compressed observation data for smart wearable devices is constructed based on dual-constellation signals; The noise variance is calculated based on the compressed observation data and the noise tolerance threshold is set. At the same time, the regularization intensity coefficient is calculated based on the current signal-to-noise ratio. A solution model is established based on noise variance, noise tolerance threshold, and regularization intensity coefficient. Calculate the sparse channel impulse response solution based on the solution model; The sparse channel impulse response solution is weighted and fused with the channel impulse response at the previous time step to obtain the channel parameters.

[0028] Specifically, the device receives pre-processed and delay-compensated BeiDou and GPS dual-constellation signals via an antenna array, containing propagation characteristics of different frequency bands and satellite paths. Due to the limited computing and storage resources of smart wearable devices, the concept of compressed observation is used to map the high-dimensional original signal into a low-dimensional observation vector through a downsampling matrix, obtaining compressed observation data. Noise statistical analysis is performed on the compressed observation data. The noise variance is obtained by performing sliding window statistics on the residual terms of the compressed observation sequence, and then the average energy deviation is calculated by combining it with the detected signal strength distribution, resulting in a quantitative characterization of the actual noise level. A noise tolerance threshold is set, representing the maximum allowable noise interference level during signal recovery, so that the optimization solution does not overfit the noise components. The regularization strength coefficient is dynamically adjusted according to the current signal-to-noise ratio (SNR). When the SNR is high, indicating good signal quality, the regularization strength is reduced to retain more detailed information; when the SNR is low, the regularization strength is increased to enhance sparsity constraints and suppress the influence of noise. A solution model is established based on noise variance, noise tolerance threshold, and regularization intensity coefficient. The core of the model is a constrained sparse optimization problem, aiming to minimize the L1 norm of the channel impulse response in the sparse domain while satisfying the noise tolerance threshold. The solution model ensures that the recovered channel impulse response solution is as sparse as possible, thus highlighting the direct path and a small number of effective multipath components, and effectively controlling errors under constraints. Numerical algorithms such as iterative threshold shrinkage and alternating direction multiplier methods are employed during the optimization process. These algorithms have relatively low computational cost, making them suitable for operation in resource-constrained smart wearable devices, and they can converge to a relatively optimal solution within a finite number of iterations. After the sparse optimization is completed, a sparse channel impulse response solution is obtained, which in the time domain represents several high-energy impulse points, corresponding to the positions and gains of the direct signal and the main multipath components. To avoid instability in the estimation results due to random disturbances at a single moment, a time-series fusion mechanism is introduced. The sparse channel impulse response solution at the current moment is weighted and fused with the channel impulse response at previous moments. The weight allocation is dynamically adjusted based on the residual size, signal-to-noise ratio, or motion state. When the device is in a stable signal state, the weight of the current moment estimate is increased to improve response sensitivity; when the device is in a state of violent motion or increased noise interference, the weight of the previous moment is appropriately increased to improve smoothness and robustness. Channel parameters are obtained through a weighted fusion strategy.

[0029] In one specific embodiment, the process of calculating the sparse channel impulse response solution based on the solution model can specifically include the following steps: The iteration step size and the current solution vector are initialized based on the solution model, and the iteration step size and the current solution vector are used as the initial solution. At the same time, the convergence threshold of the smart wearable device is set. Substitute the current solution vector into the objective function in the solution model to calculate the gradient direction, and update the current solution vector along the negative gradient direction according to the iteration step size to obtain the gradient-updated solution. Perform a soft threshold shrinkage operation on each component of the gradient update solution to obtain the shrunken solution vector; Calculate the normalized difference between the shrunken solution vector and the solution vector of the previous iteration. When the normalized difference is less than the convergence threshold, terminate the iteration and output the sparse channel impulse response solution. Otherwise, use the shrunken solution vector as the new current solution and continue iterating.

[0030] Specifically, the iteration step size and the current solution vector are initialized based on the preset solution model. The iteration step size determines the movement magnitude during each update along the negative gradient direction. If the step size is too large, the solution may oscillate around the optimum or even diverge, while if the step size is too small, the convergence speed will be reduced. Therefore, an adaptive initial step size is selected by considering historical noise levels, signal-to-noise ratio, and hardware computing power. The current solution vector is initialized as an all-zero vector or a small random perturbation vector to ensure that the sparsity assumption holds in the initial stage. At the same time, a convergence threshold is set, representing the minimum allowable change during the iteration process. When the difference between two consecutive iterations is less than the convergence threshold, the solution is considered to have converged to an acceptable sparse channel impulse response. The current solution vector is substituted into the objective function of the solution model, and the gradient direction of the objective function with respect to the current solution is calculated. The gradient direction indicates the direction of the fastest descent of the objective function. Along the negative gradient direction, the current solution vector is updated according to the set iteration step size to obtain a new gradient-updated solution. A soft-threshold shrinkage operation is performed on each component of the gradient update solution. Components with smaller amplitudes are attenuated or directly compressed to zero to eliminate weak disturbances caused by noise, while retaining the principal components with larger amplitudes to highlight the true direct paths and a small number of effective multipath components in the channel impulse response. If the amplitude of a component is less than the threshold, the component is set to zero; if the amplitude is greater than the threshold, the threshold value is subtracted from its original value to obtain the shrunken solution vector. The shrunken solution vector is compared with the solution vector of the previous iteration, and the normalized difference is calculated. The normalized difference is an important indicator of the change in the result between two iterations. If the difference is still large, it means that the solution has not yet converged and iteration needs to continue; if the difference is less than the convergence threshold set at initialization, it means that the change in the solution is small enough and the channel impulse response has stabilized near the current solution. At this point, the iteration is terminated, and the current shrunken solution vector is output as the final sparse channel impulse response solution. If the normalized difference is greater than the convergence threshold, the shrunken solution vector is substituted back into the objective function as the new current solution to continue iteration.

[0031] In one specific embodiment, the process of performing step 102 may specifically include the following steps: The satellite visibility adjacency matrix and degree matrix are constructed based on channel parameters. The satellite visibility adjacency matrix is ​​used to determine the connection weights according to the spatial geometric distribution between satellites, and the degree matrix reflects the spatial association strength of each satellite node. A satellite spatial relationship graph structure is created based on connection weights and spatial correlation strength; The satellite spatial relationship graph structure is input into the graph convolutional layer for spatial feature extraction to obtain spatial channel features. At the same time, the channel parameters are input into the gated temporal convolutional layer for temporal feature extraction to obtain temporal channel features. Spatial channel features and temporal channel features are fused to obtain a fused feature vector; The multipath interference probability is calculated based on the fused feature vector. When the multipath interference probability exceeds the preset probability threshold, the channel parameters are suppressed to obtain a clean signal.

[0032] Specifically, a satellite visibility adjacency matrix and a degree matrix are constructed based on channel parameters. Using the geometric coordinates between the satellites and the smart wearable device receiver, the relative spatial angles and distances between each pair of satellites are calculated, and a connection weight is assigned to each pair. The value of the connection weight is inversely proportional to the square of the spatial distance and is adjusted according to the correlation index in the channel parameters to ensure that the weight accurately reflects the coupling degree of satellite signals in spatial distribution. The weights are filled into a matrix structure according to satellite numbers to obtain the satellite visibility adjacency matrix. To measure the overall correlation between each satellite and other satellites, the weights in each row are summed to generate a degree matrix. The diagonal elements of the matrix reflect the centrality and spatial correlation strength of a single satellite in the network topology. The adjacency matrix and the degree matrix are used together to construct a satellite spatial relationship graph structure. The satellite spatial relationship graph structure uses satellites as nodes and connection weights as edge attributes to reflect the geometric relationships between satellites, and strengthens the network topology characteristics through node degrees. The satellite spatial relationship graph structure is then input into a graph convolutional layer for spatial feature extraction. In graph convolution operations, a high-dimensional feature representation of each satellite node under the influence of its neighboring nodes is obtained through weighted aggregation of the adjacency matrix and the feature matrix. This reflects the complex relationship between spatial coupling and observation geometry between satellites, forming spatial channel features. Simultaneously, the original channel parameters are directly input into a gated temporal convolutional layer, and dynamic features of channel evolution over time are extracted using one-dimensional convolution operations and a gating mechanism. The introduction of the gating structure effectively captures sudden changes and periodic patterns in non-stationary channels, and the resulting temporal channel features can characterize the temporal correlation and fading modes of multipath signals. Spatial and temporal channel features are fused, and key features are highlighted through attention weighting or linear combination, enabling the fused vector to reflect the joint characteristics of channel state and satellite geometric distribution. Based on the fused feature vector, the multipath interference probability is calculated. A fully connected layer and a sigmoid function output are used to map the fused vector to a value between 0 and 1, representing the probability that the currently received signal is interfered with by multipath effects. When the calculated multipath interference probability exceeds a preset probability threshold, a significant multipath component is determined to exist in the channel parameters, and a multipath component suppression mechanism is activated. The suppression operation involves constructing a filter matrix to attenuate or remove components in the channel impulse response that have large delays or whose energy does not conform to the characteristics of a direct wave, thereby obtaining a purified signal.

[0033] In one specific embodiment, the process of performing step 103 may specifically include the following steps: Extract BeiDou observation data from the purified signal, and simultaneously extract GPS observation data from the purified signal; By leveraging the timing accuracy advantage of the BeiDou system to correct receiver clock bias in GPS observation data, and by utilizing the geometric configuration advantage of the GPS system to correct the geometric dilution accuracy of BeiDou observation data, a set of observation equations is established. The adaptive factor is calculated based on the magnitude of the residual change in the observation equation set, and an adaptive filter parameter set is generated based on the adaptive factor. Based on the adaptive filtering parameter set, a prediction and update loop of capacitive Kalman filtering is performed to finally calculate the location information.

[0034] Specifically, data extraction is performed based on the purified signal. Pseudorange observations, Doppler observations, and related signal measurements belonging to the BeiDou system are used as BeiDou observation data, while corresponding observation data belonging to the GPS system are extracted from the purified signal. Cross-correction is performed using the complementary characteristics of BeiDou and GPS. The BeiDou system has an advantage in timing accuracy; its high-precision timing capability provides a more accurate time reference for the receiver. Therefore, the receiver clock offset in GPS observation data is corrected using BeiDou data, allowing GPS pseudorange and Doppler data to be interpreted on a unified time scale. Meanwhile, the GPS system, due to its wide satellite distribution and superior geometric configuration, has an advantage in geometric accuracy dilution factor. The geometric constraints of GPS observation data are used to correct the geometric dilution accuracy of BeiDou data, making the BeiDou solution more reliable in the spatial dimension. After clock offset correction and geometric accuracy optimization, the two types of data are jointly used to establish an observation equation set. An adaptive factor is calculated based on the residual variation amplitude in the observation equation set. The variation amplitude of the residual in the observation equation set is calculated in real time. When the residual changes are small, it indicates that the observed data is stable and reliable, and the adaptive factor is set to a smaller value to maintain the stability of the filter. Conversely, when the residual changes drastically, it indicates the presence of noise or anomalies in the observations, and the adaptive factor is set to a larger value to enhance the filter's robustness to noise. Based on the adaptive factor, a set of adaptive filtering parameters is generated to dynamically adjust the process noise covariance and the observation noise covariance. A capacitive Kalman filter prediction and update loop is executed based on the adaptive filtering parameter set. In the prediction phase, the device position, velocity, and clock error are extrapolated using a state transition model to obtain prior estimates. In the update phase, newly acquired observation data is compared with the prior estimates, and the nonlinear function is effectively approximated through capacitive sampling point propagation. The state estimates are then corrected using the adaptively adjusted covariance parameters. Prediction and update are performed alternately, the residual gradually decreases, the state vector gradually converges, and the position information is output.

[0035] In one specific embodiment, the process of performing a prediction and update loop of capacitive Kalman filtering based on an adaptive filtering parameter set to finally calculate the location information can specifically include the following steps: Based on the adaptive filter parameter set, volumetric sampling points are generated and propagated to the prediction time through the state transition equation containing the uniform motion model and the random walk clock difference model, forming volumetric point distribution data. The mean of state prediction is calculated based on the volume point distribution data. At the same time, the adaptive factor is adjusted by the hyperbolic tangent function according to the degree of residual exceedance to obtain the prior state statistical parameters. Substitute the prior state statistical parameters into the observation equation set, calculate the theoretical observation values ​​corresponding to each volume sampling point, and dynamically update the observation noise covariance matrix according to the signal quality to form the observation space statistical parameters. The optimal Kalman gain is calculated by observing spatial statistical parameters, and the prior state statistical parameters are corrected based on the optimal Kalman gain to solve for the location information.

[0036] Specifically, within the filter's operating cycle, a set of volumetric sampling points is generated based on the previous round's filter output and the adaptive filter parameter set for the current observation cycle. The construction of these volumetric sampling points follows the principle of unscented transformation, constructing a symmetrical distribution according to the current state mean and state covariance matrix. Positive and negative perturbations are introduced into each state dimension to fully cover the main trends in the state space under the Gaussian assumption. Each volumetric sampling point is used as a state variable input to the state transition equation. The equation contains two sub-models: a uniform motion model describes the stationary displacement changes of the device within a short time, while a random walk clock bias model is introduced to dynamically model the internal time drift of the receiver. This ensures that the state variables, after propagation to the prediction time, accurately reflect the continuous evolution of the device in the spatiotemporal coordinates and the local time reference. All volumetric sampling points, after propagation, collectively constitute the volumetric sampling point distribution data. Statistical analysis is performed on all propagated volumetric sampling points to calculate the state prediction mean, which serves as the central estimate of the prior state. Then, the prior state covariance matrix is ​​constructed based on the distribution characteristics of each volumetric sampling point's deviation from the prediction mean, improving the accuracy of the representation of state uncertainty. A residual feedback adjustment mechanism is introduced, using the squared residuals and variances of the previous observation period to form a residual exceedance index. This index is then mapped to an adaptive factor adjustment quantity via a hyperbolic tangent function. This factor dynamically expands or contracts the diagonal term of the current state covariance, forming prior state statistical parameters. These prior state statistical parameters are substituted into the observation equations, which consist of pseudorange and Doppler frequency shift functions derived from satellite geometry, corresponding to position and velocity measurement constraints, respectively. The theoretical observation value in the observation space is calculated for each volumetric sampling point, forming the observation prediction distribution. Simultaneously, the observation noise covariance matrix is ​​dynamically updated based on signal quality indicators (such as signal strength, signal-to-noise ratio, and number of available satellites) reflected in the current channel parameters. The weights of observation components from low signal-to-noise ratio satellites are increased or redundant and anomalous observations are directly masked, forming the observation space statistical parameters. After obtaining the prior state prediction distribution and observation spatial statistical parameters, statistical matching operations are performed based on the unscented Kalman filter theory. The Kalman gain is calculated using the cross-covariance between the prior state covariance and the observation prediction covariance. The Kalman gain is a crucial control variable in the filter used to balance prediction and observation. The prior state mean is updated using the optimal Kalman gain, and corrections are made based on the observation residuals. The optimal estimated state at the current moment is then output, including the device's three-dimensional position, three-dimensional velocity, and receiver clock error.

[0037] In one specific embodiment, the process of performing step 104 may specifically include the following steps: The channel change rate is calculated based on the time-domain variation of channel parameters. At the same time, the change of geometric precision dilution factor is extracted from the location information, and the remaining battery power ratio of the smart wearable device is obtained. The channel change rate, the change in geometric precision dilution factor, and the proportion of remaining battery power are weighted and summed to obtain a weighted result. When the weighted result exceeds the trigger threshold, the transmission decision function value is calculated. The channel adaptation coefficient and power adaptation coefficient of the base transmit power are adjusted according to the transmission decision function value, and the dynamic transmit power is determined based on the channel adaptation coefficient and power adaptation coefficient. The time interval for intermittent communication is calculated based on the reciprocal relationship between dynamic transmit power and transmission decision function, and the transmit power parameter and communication interval parameter are obtained. The transmit power parameter and communication interval parameter are then combined to generate transmission parameters. Based on the transmission parameters, channel state prediction is performed and the positioning calculation for the next moment is compensated to obtain the target positioning result and communication transmission scheme.

[0038] Specifically, during continuous operation, the changing trends of channel parameters over time are continuously monitored. The channel impulse response or channel gain matrix of two adjacent sampling periods is selected as input. The difference between the two period channel matrices is normalized by introducing the Frobenius norm, and the channel change rate is calculated. This effectively reflects the communication channel instability caused by multipath structures, obstruction, or frequency-selective fading in the environment. Simultaneously, the corresponding geometric precision dilution factor is extracted from the current positioning result and compared with the geometric precision dilution factor of the previous period to calculate the change in the geometric precision dilution factor, reflecting the impact of the current satellite configuration on positioning accuracy. The power management module of the smart wearable device is accessed to obtain the current remaining battery power. The remaining battery power is then ratioized to the device's design capacity to obtain the remaining battery power ratio. The channel change rate, geometric dilution change, and remaining battery power ratio are weighted and summed to construct a normalized weighted result. The weight allocation is determined based on long-term measured data of the device, ensuring that channel stability, positioning reliability, and energy consumption are prioritized in a comprehensive consideration. When the weighted result exceeds a preset trigger threshold, the device is determined to be in a sensitive region where communication and positioning quality changes significantly or energy status is critical. At this point, an adaptive transmission scheduling mechanism is triggered, and a transmission decision function value is calculated. Based on the transmission decision function value, the communication power is adjusted. This adjustment process is based on two factors: a channel adaptation coefficient, which decreases when the channel is stable and increases when the channel is unstable, based on the channel rate of change component in the decision function, to ensure sufficient communication link quality under high interference conditions; and a battery adaptation coefficient, a monotonic function of the remaining battery power percentage, which increases when the remaining power is high to support frequent transmissions and decreases when the power is low to limit transmission energy consumption. The dynamic transmission power for the current period is determined jointly based on the channel adaptation coefficient, the battery adaptation coefficient, and the base transmission power, serving as the power setting value for the device's next data uplink or status broadcast. The time interval for the next communication, i.e., the intermittent communication scheduling period, is calculated based on the reciprocal relationship of the transmission decision function value. When the system is stable, the decision function value is small, and its reciprocal is large, automatically extending the communication cycle to reduce resource consumption. Conversely, when the system state changes drastically or channel uncertainty increases, the decision function value rises, its reciprocal decreases, and the communication interval shortens accordingly to improve response capability. The transmit power parameter and communication interval parameter together form the transmission parameter packet for the next cycle, used to control the power modulation and data scheduling behavior of the communication module. Using short-time series prediction methods such as ARIMA or RLS, based on previous channel change rates and power scheduling results, the channel parameter trend for the next cycle is estimated, and the prediction error is fed back into the location calculation process as an observation noise covariance adjustment factor, improving the adaptive capability of the next cycle's positioning calculation to channel non-stationarity.After completing communication power regulation, scheduling cycle adjustment and positioning parameter compensation, the equipment simultaneously outputs the predicted and corrected target positioning results and feasible communication transmission schemes, achieving optimal positioning and communication coordination under the condition of minimizing resources.

[0039] In one specific embodiment, the process of performing channel state prediction and compensating for the positioning calculation at the next moment based on the transmission parameters to obtain the target positioning result and the communication transmission scheme can specifically include the following steps: The communication interval parameter in the transmission parameters is used as the prediction step size, and the second-order autoregressive integral moving average algorithm is used to extrapolate and predict the channel parameters to obtain the predicted channel state at the next time step. Based on the prediction channel state, the cumulative deviation of the prediction error is corrected to obtain the corrected prediction channel parameters. At the same time, the spatial position of the smart wearable device at the next moment is calculated based on the location information to obtain the predicted trajectory position. Based on the corrected predicted channel parameters, the BeiDou observation data and GPS observation data are weighted and re-executed to obtain the target positioning result. The optimal communication timing for the next cycle is calculated based on the predicted trajectory position and the transmit power parameter in the transmission parameters, and the communication transmission scheme is determined by combining the predicted channel state and the optimal communication timing.

[0040] Specifically, communication interval parameters are extracted from the transmission scheduling module and used as the time step input to the prediction module for the channel state prediction model. The communication interval parameter is equivalent to the prediction step size, defining the time span from the current state to the next predicted state. A second-order autoregressive integral moving average (ARIMA(2,1,1)) model is used to perform sequence modeling and state extrapolation of continuously acquired channel parameters. The ARIMA model, by constructing a difference equation based on two time lag terms and one moving average term, captures the linear trend and residual dynamics of channel gain or multipath indicators, thus structurally possessing the ability to simultaneously predict linear changes and sudden fluctuations. Using the ARIMA model, the current and the channel parameters from the previous two periods are used as input, and convolution estimation is performed in conjunction with the residual sequence to obtain the predicted channel state for the next time step. After completing the channel extrapolation, a bias correction process is performed based on historical prediction errors. By accumulating the errors between the predicted and actual observations over several previous cycles using a moving average, a residual cumulative deviation vector is constructed. This residual cumulative deviation vector is then weighted and fed back to the predicted channel state to obtain the corrected predicted channel parameters. Simultaneously, the position and velocity estimates from the previous cycle are invoked, and combined with a uniform acceleration motion model or motion state recognition algorithm, the device's three-dimensional spatial predicted trajectory position at the next moment is calculated based on the predicted step size, providing spatial prior reference for subsequent communication scheduling. Based on the corrected channel prediction parameters, weighting coefficients are assigned to BeiDou and GPS observation data. The weights are dynamically adjusted according to the gain stability, signal-to-noise ratio, and obstruction probability of each satellite system's corresponding channel under the predicted state. For example, when the GPS channel prediction is stable and the number of available satellites is high, its weight increases; conversely, when the BeiDou system prediction shows high-frequency channel disturbances or obstruction tendencies, its weight decreases. The weighted and fused observation data is substituted into a capacitive Kalman filter to reinitialize the capacitive sampling points. A state propagation and observation update process is constructed based on the current state covariance. Within the unscented transformation framework, prediction and correction are performed through state transitions and observation equations, outputting the target location result. This ensures that structural disturbances are incorporated into the location estimation before channel changes. The predicted trajectory location and transmit power parameters are input into the communication window calculation module. The optimal communication timing is calculated based on the device's potential geographical location and wireless environment characteristics at the next moment. The optimal communication timing considers not only the reachability of the predicted trajectory within the communication area but also the expected rate of change of channel state and channel load interference at the geographical location. For example, communication is prioritized at the edge moment before the predicted trajectory crosses a channel obstruction area to improve success rate and reduce retransmission overhead. The predicted channel state and optimal communication timing are jointly input into the communication control module. Based on the timing window constraints, the channel access strategy, scheduling sequence, and power adjustment parameters are adjusted to form a communication transmission scheme.

[0041] The above describes the multimodal positioning method based on a smart wearable device in the embodiments of the present invention. The following describes the multimodal positioning system based on a smart wearable device in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the multimodal positioning system based on a smart wearable device in this invention includes: The sparse channel estimation module 201 is used to receive BeiDou satellite signals and GPS satellite signals and perform sparse channel estimation to obtain channel parameters. The multipath interference suppression module 202 is used to identify and suppress multipath interference based on channel parameters to obtain a clean signal. The cross-correction solution module 203 is used to separate the purified signal into BeiDou observation data and GPS observation data and perform cross-correction solution to obtain location information; The channel state prediction module 204 is used to perform channel state prediction based on location information and channel parameters and compensate for the positioning calculation at the next moment to obtain the target positioning result and communication transmission scheme.

[0042] Through the synergistic cooperation of the above-mentioned components, beneficial effects can be achieved. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0043] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0044] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multimodal positioning method based on a smart wearable device, characterized in that, include: The system receives BeiDou and GPS satellite signals and performs sparse channel estimation to obtain channel parameters. Based on the channel parameters, multipath interference is identified and suppressed to obtain a purified signal; The purified signal is separated into BeiDou observation data and GPS observation data, and cross-correction calculation is performed to obtain the location information; Based on the location information and the channel parameters, channel state prediction is performed and the positioning calculation for the next time step is compensated to obtain the target positioning result and communication transmission scheme.

2. The multimodal positioning method based on a smart wearable device according to claim 1, characterized in that, The process of receiving BeiDou satellite signals and GPS satellite signals and performing sparse channel estimation to obtain channel parameters includes: Receive BeiDou satellite signals and GPS satellite signals, and calculate the BeiDou signal attenuation coefficient and the GPS signal attenuation coefficient respectively; The BeiDou satellite signal is attenuated based on the BeiDou signal attenuation coefficient to obtain the BeiDou received signal, and the GPS satellite signal is attenuated based on the GPS signal attenuation coefficient to obtain the GPS received signal. The BeiDou received signal and the GPS received signal are combined to obtain a combined energy signal; The motion intensity index is calculated based on the three-axis acceleration data of the smart wearable device. When the motion intensity index exceeds the intensity index threshold, the Beidou propagation delay and GPS propagation delay in the merged energy signal are compensated respectively to obtain the dual constellation signal. Sparse channel estimation is performed on the dual-constellation signal to obtain the channel parameters.

3. The multimodal positioning method based on a smart wearable device according to claim 2, characterized in that, The sparse channel estimation performed on the dual-constellation signal to obtain channel parameters includes: Compressed observation data for the smart wearable device is constructed based on the dual-constellation signals. The noise variance is calculated based on the compressed observation data and a noise tolerance threshold is set. At the same time, the regularization intensity coefficient is calculated based on the current signal-to-noise ratio. A solution model is established based on the noise variance, the noise tolerance threshold, and the regularization intensity coefficient. Calculate the sparse channel impulse response solution based on the solution model; The sparse channel impulse response solution is weighted and fused with the channel impulse response at the previous time step to obtain the channel parameters.

4. The multimodal positioning method based on a smart wearable device according to claim 3, characterized in that, The calculation of the sparse channel impulse response solution based on the solution model includes: The iteration step size and the current solution vector are initialized based on the solution model, and the iteration step size and the current solution vector are used as the initial solution. At the same time, the convergence threshold of the smart wearable device is set. Substitute the current solution vector into the objective function in the solution model to calculate the gradient direction, and update the current solution vector along the negative gradient direction according to the iteration step size to obtain the gradient-updated solution. A soft threshold shrinkage operation is performed on each component of the gradient update solution to obtain the shrunken solution vector; Calculate the normalized difference between the shrunken solution vector and the solution vector of the previous iteration. When the normalized difference is less than the convergence threshold, terminate the iteration and output the sparse channel impulse response solution. Otherwise, use the shrunken solution vector as the new current solution to continue the iteration.

5. The multimodal positioning method based on a smart wearable device according to claim 1, characterized in that, The step of identifying and suppressing multipath interference based on the channel parameters to obtain a cleaned signal includes: Based on the channel parameters, a satellite visibility adjacency matrix and a degree matrix are constructed. The satellite visibility adjacency matrix is ​​used to determine the connection weights according to the spatial geometric distribution between satellites, and the degree matrix reflects the spatial association strength of each satellite node. A satellite spatial relationship graph structure is created based on the connection weights and the spatial association strength; The satellite spatial relationship graph structure is input into the graph convolutional layer for spatial feature extraction to obtain spatial channel features. At the same time, the channel parameters are input into the gated temporal convolutional layer for temporal feature extraction to obtain temporal channel features. The spatial channel features and the temporal channel features are fused to obtain a fused feature vector; The multipath interference probability is calculated based on the fused feature vector. When the multipath interference probability exceeds a preset probability threshold, multipath component suppression is performed on the channel parameters to obtain a clean signal.

6. The multimodal positioning method based on a smart wearable device according to claim 1, characterized in that, The process of separating the purified signal into BeiDou observation data and GPS observation data and performing cross-correction calculations to obtain location information includes: BeiDou observation data is extracted from the purified signal, and GPS observation data is also extracted from the purified signal. The timing accuracy advantage of the BeiDou system is used to correct the receiver clock offset in the GPS observation data, and the geometric configuration advantage of the GPS system is used to correct the geometric dilution accuracy of the BeiDou observation data, thus establishing a set of observation equations. The adaptive factor is calculated based on the residual change magnitude in the observation equation set, and an adaptive filter parameter set is generated based on the adaptive factor. Based on the adaptive filtering parameter set, a prediction and update loop of capacitive Kalman filtering is performed to finally calculate the location information.

7. The multimodal positioning method based on a smart wearable device according to claim 6, characterized in that, The prediction and update loop based on the adaptive filtering parameter set, performing capacitive Kalman filtering, ultimately calculates the location information, including: Based on the set of adaptive filtering parameters, volumetric sampling points are generated and propagated to the prediction time through state transition equations that include uniform motion model and random walk clock difference model, forming volumetric point distribution data. Based on the volume point distribution data, the mean of the state prediction is statistically calculated, and the adaptive factor is adjusted by the hyperbolic tangent function according to the degree of residual exceeding the limit to obtain the prior state statistical parameters. Substitute the prior state statistical parameters into the observation equation set to calculate the theoretical observation values ​​corresponding to each volume sampling point, and dynamically update the observation noise covariance matrix according to the signal quality to form the observation space statistical parameters. The optimal Kalman gain is calculated using the observed spatial statistical parameters, and the prior state statistical parameters are corrected based on the optimal Kalman gain to solve for the location information.

8. The multimodal positioning method based on a smart wearable device according to claim 1, characterized in that, The step of performing channel state prediction and compensating for the next time step based on the location information and channel parameters to obtain the target positioning result and communication transmission scheme includes: The channel change rate is calculated based on the time-domain changes of the channel parameters. At the same time, the change in the geometric precision dilution factor is extracted from the location information, and the remaining battery power ratio of the smart wearable device is obtained. The channel change rate, the change in the geometric precision dilution factor, and the remaining battery power ratio are weighted and summed to obtain a weighted result. When the weighted result exceeds the trigger threshold, the transmission decision function value is calculated. The channel adaptation coefficient and power adaptation coefficient of the base transmit power are adjusted according to the transmission decision function value, and the dynamic transmit power is determined based on the channel adaptation coefficient and the power adaptation coefficient. The time interval for intermittent communication is calculated based on the reciprocal relationship between the dynamic transmission power and the transmission decision function, thereby obtaining the transmission power parameter and the communication interval parameter. The transmission power parameter and the communication interval parameter are then combined to generate the transmission parameter. Based on the transmission parameters, channel state prediction is performed and the positioning calculation for the next moment is compensated to obtain the target positioning result and communication transmission scheme.

9. The multimodal positioning method based on a smart wearable device according to claim 8, characterized in that, The step of performing channel state prediction and compensating for the positioning calculation at the next moment based on the transmission parameters to obtain the target positioning result and communication transmission scheme includes: The communication interval parameter in the transmission parameters is used as the prediction step size, and the channel parameters are extrapolated and predicted using a second-order autoregressive integral moving average algorithm to obtain the predicted channel state at the next time step. Based on the predicted channel state, the cumulative deviation of the prediction error is corrected to obtain the corrected predicted channel parameters. At the same time, the spatial position of the smart wearable device at the next moment is calculated based on the location information to obtain the predicted trajectory position. Based on the corrected predicted channel parameters, the BeiDou observation data and the GPS observation data are weighted and re-executed with capacitive Kalman filtering to obtain the target positioning result. The optimal communication timing for the next cycle is calculated based on the predicted trajectory position and the transmit power parameter in the transmission parameters, and the communication transmission scheme is determined by combining the predicted channel state and the optimal communication timing.

10. A multimodal positioning system based on a smart wearable device, characterized in that, A method for performing a multimodal positioning based on a smart wearable device as described in any one of claims 1-9, comprising: The sparse channel estimation module is used to receive BeiDou satellite signals and GPS satellite signals and perform sparse channel estimation to obtain channel parameters. A multipath interference suppression module is used to identify and suppress multipath interference based on the channel parameters to obtain a clean signal; The cross-correction calculation module is used to separate the purified signal into BeiDou observation data and GPS observation data and perform cross-correction calculation to obtain location information; The channel state prediction module is used to perform channel state prediction based on the location information and the channel parameters and compensate for the positioning calculation at the next moment to obtain the target positioning result and communication transmission scheme.

Images