High-precision indoor positioning system based on smart bracelet

Abstract

The invention discloses a high-precision indoor positioning system based on a smart bracelet, and the system comprises a data processing module which is used for obtaining and preprocessing multi-source sensing data collected by the smart bracelet; the gait reckoning module is used for executing gait event detection and pedestrian dead reckoning processing; the wireless observation module is used for carrying out preprocessing and credibility evaluation on indoor wireless positioning observation data; the fusion positioning module is used for carrying out credibility weighted fusion; the trajectory constraint module is used for applying trajectory rationality constraint and correcting abnormal position points; the error correction module is used for constructing a positioning error dynamic correction model and executing online correction processing; and the track output module is used for carrying out smoothing and time resampling processing and extracting a human body position determination result set. The method is based on the multi-source perception fusion and federated Kalman dynamic correction method, achieves the indoor high-precision positioning of the smart bracelet, and has the advantages of continuous track, controllable error and high environmental adaptability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent positioning, and in particular to a high-precision indoor positioning system based on a smart bracelet. BACKGROUND

[0002] Existing indoor positioning technologies mainly include a pedestrian dead reckoning method based on inertial sensors and a wireless positioning method based on Bluetooth, a wireless local area network or ultra-wideband. The inertial positioning relies on acceleration and angular velocity integration to calculate displacement and heading, and is easily affected by sensor noise and cumulative errors. The wireless positioning estimates the position through signal strength or ranging information, and is obviously affected by multipath effects, shielding and environmental changes. To improve positioning accuracy, some solutions attempt to fuse inertial positioning with wireless positioning, but mostly use simple weighting or single filtering structure, which is difficult to maintain stable accuracy in complex indoor environments.

[0003] Existing fusion positioning solutions generally have problems of insufficient modeling of historical errors, limited use of trajectory rationality and environmental constraints, resulting in drift accumulation, position jumping or discontinuity of positioning results in long-time operation. At the same time, some methods do not dynamically evaluate the credibility of wireless observations, and lack smoothing and time resampling processing of correction results, which is difficult to output continuous, stable and actual walking regularity of human indoor positioning trajectory, limiting the application effect in high-precision indoor positioning scenarios.

[0004] Therefore, how to provide a high-precision indoor positioning system based on a smart bracelet is a problem that those skilled in the art need to solve. SUMMARY

[0005] One object of the present application is to provide a high-precision indoor positioning system based on a smart bracelet. The present application is based on multi-source perception fusion and federal Kalman dynamic correction method, realizes high-precision indoor positioning of smart bracelet, and has the advantages of continuous trajectory, controllable error and strong environmental adaptability.

[0006] According to the high-precision indoor positioning system based on a smart bracelet of the present application, the system comprises: a data processing module for acquiring and preprocessing multi-source perception data collected by the smart bracelet, and generating a standardized inertial time series data set; a gait estimation module for performing gait event detection and pedestrian dead reckoning processing based on the standardized inertial time series data set, and outputting a relative trajectory sequence at the bracelet end; a wireless observation module for acquiring indoor wireless positioning observation data and preprocessing and credibility evaluation, and generating a wireless observation sequence and an observation credibility sequence; The fusion positioning module is used to weight the wireless observation sequence based on the observation confidence sequence in a unified coordinate system, and to fuse it with the relative trajectory sequence of the wristband in time series, and output the initial fused position sequence. The trajectory constraint module is used to apply trajectory rationality constraints to the initial fused position sequence and correct abnormal position points, and output the constrained position sequence; The error correction module is used to construct a dynamic positioning error correction model based on federated Kalman filtering based on standardized inertial time series datasets, wireless observation sequences, constrained position sequences and historical positioning results, and to perform online correction processing and output the corrected position sequence. The trajectory output module is used to smooth and resample the corrected position sequence over time to generate a continuous indoor positioning trajectory and extract the human position determination result set.

[0007] Optionally, modules can be integrated using the following methods: Acquire multi-source sensing data collected by the smart bracelet and preprocess it to obtain a standardized inertial time-series dataset; Based on a standardized inertial time-series dataset, gait event detection and pedestrian dead reckoning are performed to obtain a relative trajectory sequence from the wristband. Acquire indoor wireless positioning observation data output by indoor wireless positioning infrastructure, perform preprocessing and reliability assessment, and obtain observation reliability sequence and wireless observation sequence; The wireless observation sequence is weighted by confidence in a unified coordinate system, and the weighted wireless observation sequence is fused with the relative trajectory sequence of the wristband in time series processing to output the initial fused position sequence. Apply trajectory rationality constraints to the initial fused position sequence and correct abnormal position points to obtain the constrained position sequence; Based on standardized inertial time-series datasets, wireless observation sequences, constrained position sequences, and historical positioning results, a dynamic positioning error correction model based on federated Kalman filtering is constructed, and online correction processing is performed to output the corrected position sequence. The corrected position sequence is smoothed and time-resampled to generate a continuous indoor positioning trajectory and extract the human position determination result set.

[0008] Optionally, the multi-source sensing data includes triaxial acceleration data, triaxial angular velocity data, and geomagnetic data, and the preprocessing includes timestamp alignment, outlier removal, and noise filtering.

[0009] Optionally, obtaining the relative trajectory sequence at the wristband end includes the following specific steps: Gravity component separation and coordinate system-aware processing are performed on the triaxial acceleration and triaxial angular velocity data in the standardized inertial time series dataset to obtain the vertical acceleration and angular velocity sequences. Gait event detection is performed based on vertical acceleration sequences. Peak detection and inter-peak interval constraints are used to determine the timing of continuous gait events, thus obtaining a gait event sequence. The step frequency parameter is calculated based on the gait event sequence, and the step frequency parameter is the reciprocal of the time interval between two adjacent gait events; Gait intensity features are extracted and step length parameters are calculated based on the time intervals corresponding to gait event sequences. The heading angle sequence is calculated based on the angular velocity sequence and geomagnetic data. Each moment of the heading angle sequence is obtained by correcting the rotation angle obtained by the integral of the angular velocity and the geomagnetic pointing constraint. The heading increment parameter is calculated based on the heading angle sequence. The heading increment parameter is a sequence of turning changes formed by the difference in heading angles at the times of two adjacent gait events. Pedestrian dead reckoning is performed based on gait event sequences, stride parameters, and heading increment parameters, with a position update performed every time a gait event is detected. The relative position points obtained from multiple position updates are combined according to the time sequence corresponding to the gait events to obtain a continuous relative position sequence, and the continuous relative position sequence is output as the relative trajectory sequence of the wristband.

[0010] Optionally, obtaining the observation confidence sequence and the wireless observation sequence includes the following specific steps: The indoor wireless positioning observation data output by the indoor wireless positioning infrastructure during the target positioning period is obtained, and the data format is unified and the timestamp is standardized to obtain the original wireless observation record set with timestamp identifier. Based on the original wireless observation record set, the observation type and observation source are analyzed to obtain the observation type field and the observation source field; Based on the observation type field and the observation source field, the original wireless observation record set is subjected to observation value extraction and observation value standardization to generate a standardized wireless observation record set. The standardized wireless observation record set is subjected to observation time alignment processing. A time matching method with the positioning time as the alignment reference is adopted to map the standardized wireless observation record with a time difference of less than the positioning time to the corresponding positioning time, so as to obtain the wireless observation sequence organized according to the positioning time. The initial observation confidence level is calculated based on the wireless observation record at each positioning time in the wireless observation sequence, and the initial observation confidence level sequence is obtained. The initial observation confidence sequence is updated by using the ratio of the initial observation confidence value corresponding to each positioning time to the maximum initial observation confidence value in the initial observation confidence sequence as the normalized confidence value. This generates an observation confidence sequence, which is then output after being timestamped and associated with the wireless observation sequence.

[0011] Optionally, the output of the initial fusion position sequence includes the following specific steps: The wireless observation sequence, the observation confidence sequence, and the relative trajectory sequence of the wristband are obtained. The wireless observation sequence and the relative trajectory sequence of the wristband are then converted to coordinate expressions in a unified coordinate system to obtain the wireless observation position sequence and the relative trajectory position sequence in the unified coordinate system. The wireless observation position sequence under the unified coordinate system is weighted based on the observation confidence sequence, and the confidence weight of the wireless observation position at each positioning time is calculated according to the corresponding observation confidence value. In the case of multiple wireless observation location records at the same positioning time, a weighted wireless observation location is calculated based on the confidence weight of each wireless observation location record, and the weighted wireless observation locations obtained at each positioning time are combined in chronological order to obtain a weighted wireless observation location sequence. Time alignment processing is performed on the weighted wireless observation position sequence and the relative trajectory position sequence. A time matching method with the positioning time as the alignment reference is adopted to map the relative trajectory positions with a time difference of less than the positioning time to the corresponding positioning time, so as to obtain the aligned relative trajectory position sequence organized according to the positioning time. At each positioning time, time series fusion is performed based on the weighted wireless observation position and the aligned relative trajectory position, and the fused positions obtained at each positioning time are combined in chronological order to output the initial fused position sequence.

[0012] Optionally, obtaining the constrained position sequence includes the following specific steps: Acquire and parse indoor map data to obtain an indoor map constraint dataset; Indoor map constraint processing is performed on the initial fused location sequence based on the indoor map constraint dataset. Wall crossing and passageway judgment are performed on each location point in the initial fused location sequence to obtain the map constraint judgment sequence. For invalid location points in the map constraint determination sequence, map constraint correction processing is performed. The corrected mapping position is calculated based on the minimum spatial distance between the invalid location point and the boundary of the nearest feasible passage area. The corrected mapping position is used as the map correction location point to replace the invalid location point, and the map-constrained location sequence is obtained. Acquire reachable region data and parse it into a reachable region constraint dataset; Based on the reachable region constraint dataset, reachable region constraint processing is performed on the map-constrained location sequence. Region affiliation and region connectivity are determined for each location point in the map-constrained location sequence to obtain the reachable region determination sequence. For the unreachable location points in the reachable region determination sequence, reachable region correction processing is performed. Reachable correction location points are generated by mapping unreachable location points to the nearest point of the adjacent reachable boundary, and the unreachable location points are replaced with reachable correction location points to obtain the reachable region constrained location sequence. After performing velocity continuity constraint processing on the position sequence based on the reachable region constraint, the displacement distance and time interval between the position points corresponding to adjacent positioning times are calculated to obtain the instantaneous velocity sequence; A velocity continuity constraint determination is performed on the instantaneous velocity sequence. Based on the relationship between each instantaneous velocity value in the instantaneous velocity sequence and the corresponding allowable velocity upper limit, a continuity determination mark is assigned to each instantaneous velocity value, resulting in a velocity continuity determination sequence. For the position points in the speed continuity determination sequence that are determined to be speeding, speed continuity correction processing is performed. Based on the speeding position points and the adjacent valid position points before and after them, position interpolation relationship is constructed to calculate the speed correction position points. The speed correction position points are then used to replace the speeding position points to obtain the constrained position sequence.

[0013] Optionally, the output of the corrected position sequence includes the following specific steps: A dynamic positioning error correction model based on federated Kalman filtering is constructed based on standardized inertial time-series datasets, wireless observation sequences, constrained position sequences, and historical positioning results. The positioning error dynamic correction model is a multi-sub-filter cooperative positioning correction model based on federated Kalman filtering, which includes an inertial sub-filter, a wireless sub-filter, and a federated fusion unit. In the positioning error dynamic correction model, based on the standardized inertial time series dataset and the relative trajectory sequence of the wristband, state prediction processing is performed in the inertial sub-filter to obtain the inertial state prediction result and the inertial state uncertainty measure. In the wireless sub-filter, observation update processing is performed based on the state input of the wireless sub-filter to obtain the wireless state estimation result and the wireless state uncertainty measure. The constraint residual sequence is determined based on the position deviation between the constrained position sequence and the inertial state prediction result and the position deviation between the constrained position sequence and the wireless state estimation result. Then, state correction processing is performed on the inertial state prediction result and the wireless state estimation result based on the constraint residual sequence to obtain the constrained inertial state and the constrained wireless state. Calculate the drift bias vector based on historical positioning results; The drift bias vector is written into the inertial sub-filter as the error state input, and the constrained inertial state is updated by error compensation based on the error state input to obtain the inertial state estimation result. In the federated fusion unit, the inertial state estimation results and the constrained wireless state are updated by federated fusion to obtain the federated fusion position estimation results; The federated fusion location estimation results are combined in the order of positioning time to output the corrected location sequence.

[0014] Optionally, the extraction of the human body location determination result set includes the following specific steps: Based on the corrected position coordinates and timestamp records corresponding to adjacent positioning times in the corrected position sequence, the position change and time interval are calculated to obtain the trajectory smoothing input sequence; Smoothing processing is performed based on the trajectory smoothing input sequence. A smooth position sequence is generated by updating the coordinates of the position point corresponding to each positioning time to the weighted average coordinates of multiple position points in the time neighborhood. The position points in the smooth position sequence are rearranged according to the continuous time axis. A time interpolation relationship is constructed based on the timestamp intervals of adjacent position points. The resampled position points corresponding to each resampled time are calculated under the time interpolation relationship to generate a resampled position sequence. The resampled location sequences are combined according to the temporal order of the resampled times to generate a continuous indoor positioning trajectory. The human body position determination result set is extracted based on the continuous indoor positioning trajectory. The human body position determination result set includes the position coordinates and corresponding timestamp records corresponding to each resampling time in the continuous indoor positioning trajectory.

[0015] The beneficial effects of this invention are: This invention constructs a multi-source information collaborative indoor positioning system centered on a smart wristband, achieving deep fusion and dynamic correction of inertial and wireless observation information. By performing normalized preprocessing, gait event detection, and pedestrian dead reckoning on the inertial time-series data, continuous relative motion trajectories can be obtained without relying on external infrastructure. Simultaneously, a reliability assessment and weighting mechanism for indoor wireless positioning observation data is introduced, enabling wireless observations to adaptively participate in position estimation based on their quality and stability during the fusion process. This effectively reduces positioning errors caused by signal fluctuations, multipath effects, or occlusion, improving the stability and reliability of the fused positioning results in complex indoor environments.

[0016] Furthermore, this invention introduces a trajectory rationality constraint mechanism based on the fused positioning results. By combining indoor maps, reachable areas, and velocity continuity constraints, it corrects location points that do not conform to spatial structure or motion patterns, making the positioning trajectory more consistent with real human walking behavior and suppressing abnormal jumps and unreasonable displacements at the source. Based on this, a dynamic positioning error correction model based on federated Kalman filtering is constructed. The inertial sub-filter and wireless sub-filter are modeled collaboratively, and drift bias compensation is performed using historical positioning results. This achieves continuous online correction of inertial drift errors and wireless observation errors, effectively mitigating error accumulation under long-term operating conditions and significantly improving positioning accuracy and time consistency.

[0017] Furthermore, this invention further eliminates high-frequency jitter and unifies the time scale by performing smoothing and temporal resampling processing on the corrected position sequence, outputting a continuous, smooth, and temporally consistent indoor positioning trajectory. This results in a final human position determination set that exhibits excellent performance in both spatial continuity and temporal stability. The overall technical solution achieves high-precision, robust, and long-term stable indoor positioning without requiring high-density deployment of positioning infrastructure. It possesses strong environmental adaptability and engineering application value, making it suitable for various practical application scenarios such as personnel positioning, behavior analysis, and indoor navigation. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0019] Fig. 1 This is a flowchart of a high-precision indoor positioning system based on a smart bracelet proposed in this invention. Fig. 2 This is a flowchart illustrating the trajectory rationality constraint of a high-precision indoor positioning system based on a smart bracelet, as proposed in this invention. Fig. 3 This is a dynamic correction diagram of positioning error for a high-precision indoor positioning system based on a smart bracelet proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figs. 1-3 A high-precision indoor positioning system based on a smart bracelet includes: The data processing module is used to acquire and preprocess multi-source sensing data collected by the smart bracelet to generate a standardized inertial time-series dataset. The gait estimation module is used to perform gait event detection and pedestrian dead reckoning based on a standardized inertial time series dataset, and outputs the relative trajectory sequence of the wristband. The wireless observation module is used to acquire indoor wireless positioning observation data, perform preprocessing and reliability assessment, and generate wireless observation sequences and observation reliability sequences. The fusion positioning module is used to weight the wireless observation sequence based on the observation confidence sequence in a unified coordinate system, and to fuse it with the relative trajectory sequence of the wristband in time series, and output the initial fused position sequence. The trajectory constraint module is used to apply trajectory rationality constraints to the initial fused position sequence and correct abnormal position points, and output the constrained position sequence; The error correction module is used to construct a dynamic positioning error correction model based on federated Kalman filtering based on standardized inertial time series datasets, wireless observation sequences, constrained position sequences and historical positioning results, and to perform online correction processing and output the corrected position sequence. The trajectory output module is used to smooth and resample the corrected position sequence over time to generate a continuous indoor positioning trajectory and extract the human position determination result set.

[0022] In this embodiment, the modules are interconnected using the following method: Acquire multi-source sensing data collected by the smart bracelet and preprocess it to obtain a standardized inertial time-series dataset; Based on a standardized inertial time-series dataset, gait event detection and pedestrian dead reckoning are performed to obtain a relative trajectory sequence from the wristband. Acquire indoor wireless positioning observation data output by indoor wireless positioning infrastructure, perform preprocessing and reliability assessment, and obtain observation reliability sequence and wireless observation sequence; The wireless observation sequence is weighted by confidence in a unified coordinate system, and the weighted wireless observation sequence is fused with the relative trajectory sequence of the wristband in time series processing to output the initial fused position sequence. Apply trajectory rationality constraints to the initial fused position sequence and correct abnormal position points to obtain the constrained position sequence; Based on standardized inertial time-series datasets, wireless observation sequences, constrained position sequences, and historical positioning results, a dynamic positioning error correction model based on federated Kalman filtering is constructed, and online correction processing is performed to output the corrected position sequence. The corrected position sequence is smoothed and time-resampled to generate a continuous indoor positioning trajectory and extract the human position determination result set.

[0023] In this embodiment, the multi-source sensing data includes triaxial acceleration data, triaxial angular velocity data, and geomagnetic data, and the preprocessing includes timestamp alignment, outlier removal, and noise filtering.

[0024] In this embodiment, obtaining the relative trajectory sequence of the wristband includes the following specific steps: Gravity component separation and coordinate system-aware processing are performed on the triaxial acceleration and triaxial angular velocity data in the standardized inertial time series dataset to obtain the vertical acceleration sequence for gait detection and the angular velocity sequence for heading estimation. Gait event detection is performed based on vertical acceleration sequences. Peak detection and inter-peak interval constraints are used to determine the timing of continuous gait events, thus obtaining a gait event sequence. The step frequency parameter is calculated based on the gait event sequence, and the step frequency parameter is the reciprocal of the time interval between two adjacent gait events; Gait intensity features are extracted and step length parameters are calculated based on the time interval corresponding to the gait event sequence. The gait intensity features include the peak amplitude of vertical acceleration corresponding to the gait event, the change in peak amplitude, and the energy accumulation of vertical acceleration within the time interval. The step length parameter is the single-step displacement length obtained by mapping the gait intensity features and step frequency parameters together. The heading angle sequence is calculated based on the angular velocity sequence and geomagnetic data. The heading angle sequence is a sequence representing the change of the direction angle of the wristband's orientation over time in a unified coordinate system. The heading angle sequence at each moment is obtained by correcting the rotation angle obtained by the integral of the angular velocity and the geomagnetic pointing constraint. The heading increment parameter is calculated based on the heading angle sequence. The heading increment parameter is a sequence of turning changes formed by the difference in heading angles at the time of two adjacent gait events. The sequence of turning changes is used to characterize the orientation change corresponding to each step. Pedestrian dead reckoning is performed based on gait event sequences, step length parameters, and heading increment parameters. A position update is performed each time a gait event is detected. The position update includes: determining the basic displacement direction based on the heading angle sequence corresponding to the current gait event, correcting the basic displacement direction by combining the heading increment parameters between adjacent gait events, determining the displacement length for this position update based on the step length parameters, and accumulating the relative position coordinates along the displacement direction based on the relative position coordinates, displacement direction, and displacement length corresponding to the previous position update to obtain the updated relative position coordinates. The relative position points obtained from multiple position updates are combined according to the time sequence corresponding to the gait events to obtain a continuous relative position sequence. The continuous relative position sequence is then output as the relative trajectory sequence of the wristband. The relative trajectory sequence of the wristband includes the relative position coordinates at each moment and the timestamp record corresponding to the relative position coordinates.

[0025] In this embodiment, obtaining the observation confidence sequence and the wireless observation sequence includes the following specific steps: The indoor wireless positioning observation data output by the indoor wireless positioning infrastructure during the target positioning period is acquired, and the data format is unified and the timestamp is standardized to obtain the original wireless observation record set with timestamp. The indoor wireless positioning observation data includes one or more of Bluetooth beacon observation data, wireless local area network observation data and ultra-wideband observation data. Based on the original wireless observation record set, the observation type and observation source are analyzed to obtain the observation type field and the observation source field. The observation type field is used to identify the category of Bluetooth beacon observation data, wireless LAN observation data and ultra-wideband observation data. The observation source field is used to identify beacon identifier, access point identifier or base station identifier. Based on the observation type field and the observation source field, the original wireless observation record set is subjected to observation value extraction and observation value standardization to generate a standardized wireless observation record set. The observation values ​​include one or more of the following: Bluetooth beacon observation strength value, wireless LAN observation strength value, ultra-wideband ranging value, or ultra-wideband time difference of arrival value. The standardization process includes observation unit conversion, observation value range clipping, and abnormal observation removal. The standardized wireless observation record set is subjected to observation time alignment processing. A time matching method with the positioning time as the alignment reference is adopted to map the standardized wireless observation record with a time difference of less than the positioning time to the corresponding positioning time, so as to obtain the wireless observation sequence organized according to the positioning time. The initial observation confidence is calculated based on the wireless observation records at each positioning time in the wireless observation sequence to obtain the initial observation confidence sequence. The initial observation confidence is obtained by fusing the observation quality score and the observation stability score. The observation quality score is determined based on the degree of deviation between the observed value and the effective range of the observed value, and the observation stability score is determined based on the change range of the observed value at adjacent positioning times. The initial observation confidence sequence is updated by using the ratio of the initial observation confidence value corresponding to each positioning time to the maximum initial observation confidence value in the initial observation confidence sequence as the normalized confidence value. This generates an observation confidence sequence, which is then output after being timestamped and associated with the wireless observation sequence.

[0026] In this embodiment, the output of the initial fusion position sequence includes the following specific steps: The wireless observation sequence, the observation confidence sequence, and the relative trajectory sequence of the wristband are obtained. The wireless observation sequence and the relative trajectory sequence of the wristband are then converted to coordinate expressions in a unified coordinate system to obtain the wireless observation position sequence and the relative trajectory position sequence in the unified coordinate system. The wireless observation position sequence under the unified coordinate system is weighted based on the observation confidence sequence. For each positioning time, the confidence weight is calculated according to the corresponding observation confidence value. The confidence weight is determined by the ratio of the observation confidence value corresponding to the positioning time to the sum of the observation confidence values ​​in the observation confidence sequence. In the case of multiple wireless observation location records at the same positioning time, a weighted wireless observation location is calculated based on the confidence weight corresponding to each wireless observation location record. The weighted wireless observation locations obtained at each positioning time are combined in chronological order to obtain a weighted wireless observation location sequence. The weighted wireless observation location is a location vector obtained by summing the product of each wireless observation location and its corresponding confidence weight. Time alignment processing is performed on the weighted wireless observation position sequence and the relative trajectory position sequence. A time matching method with the positioning time as the alignment reference is adopted to map the relative trajectory positions with a time difference of less than the positioning time to the corresponding positioning time, so as to obtain the aligned relative trajectory position sequence organized according to the positioning time. At each positioning time, time series fusion is performed based on the weighted wireless observation position and the aligned relative trajectory position. The fused positions obtained at each positioning time are combined in chronological order to output an initial fused position sequence. The time series fusion is determined by weighting the wireless observation contribution and the relative trajectory contribution. The wireless observation contribution is the product of the weighted wireless observation position and the wireless fusion weight. The relative trajectory contribution is the product of the aligned relative trajectory position and the relative fusion weight. The wireless fusion weight is determined by the ratio of the observation confidence value at the positioning time to the sum of the observation confidence value at the positioning time and the maximum observation confidence value in the observation confidence sequence. The relative fusion weight is obtained by subtracting the wireless fusion weight from one.

[0027] In this embodiment, obtaining the constrained position sequence includes the following specific steps: Indoor map data is acquired and parsed to obtain an indoor map constraint dataset, which includes a set of wall boundaries, a set of passage areas, a set of room boundaries, and a set of floor connectivity structures. Indoor map constraint processing is performed on the initial fused location sequence based on the indoor map constraint dataset. Wall crossing and passage within the initial fused location sequence are determined for each location point to obtain a map constraint determination sequence. The wall crossing determination is a location point validity determination based on the intersection relationship between the line connecting adjacent location points and the wall boundary set. The passage within the channel determination is a location point validity determination based on the inclusion relationship between the location point coordinates and the passage area set. For invalid location points in the map constraint determination sequence, map constraint correction processing is performed. The corrected mapping position is calculated based on the minimum spatial distance between the invalid location point and the boundary of the nearest feasible passage area. The corrected mapping position is used as the map correction location point to replace the invalid location point, and the map-constrained location sequence is obtained. Acquire reachable region data and parse the reachable region data into a reachable region constraint dataset, wherein the reachable region constraint dataset includes a set of traversable regions, a set of intraversable regions, and a set of region connectivity relationships; Based on the reachable region constraint dataset, reachable region constraint processing is performed on the map-constrained location sequence. Region affiliation and connectivity are determined for each location point in the map-constrained location sequence to obtain a reachable region determination sequence. The region affiliation determination is a determination of the validity of the location point based on the inclusion relationship between the location point coordinates and the set of walkable regions. The region connectivity determination is a determination of path reachability based on the connectivity relationship determined by the region to which the adjacent location point belongs and the set of region connectivity relationships. For the unreachable location points in the reachable region determination sequence, reachable region correction processing is performed. Reachable correction location points are generated by mapping unreachable location points to the nearest point of the adjacent reachable boundary, and the unreachable location points are replaced with reachable correction location points to obtain the reachable region constrained location sequence. After performing velocity continuity constraint processing on the position sequence based on the reachable region constraint, the displacement distance and time interval between the position points corresponding to adjacent positioning times are calculated to obtain the instantaneous velocity sequence. Each instantaneous velocity value in the instantaneous velocity sequence is the ratio of displacement distance to time interval. A speed continuity constraint determination is performed on the instantaneous speed sequence. Based on the relationship between each instantaneous speed value in the instantaneous speed sequence and the corresponding allowable speed limit, a continuity determination mark is assigned to each instantaneous speed value to obtain a speed continuity determination sequence. The continuity determination mark includes an overspeed mark and a non-overspeed mark. For the position points in the speed continuity determination sequence that are determined to be speeding, speed continuity correction processing is performed. Based on the speeding position points and the adjacent valid position points before and after them, position interpolation relationship is constructed to calculate the speed correction position points. The speed correction position points are then used to replace the speeding position points to obtain the constrained position sequence.

[0028] In this embodiment, the output of the corrected position sequence includes the following specific steps: A dynamic positioning error correction model based on federated Kalman filtering is constructed based on standardized inertial time-series datasets, wireless observation sequences, constrained position sequences, and historical positioning results. The historical positioning result is a corrected position sequence record before the target positioning time stored in the positioning history cache; The positioning error dynamic correction model is a multi-sub-filter cooperative positioning correction model based on federated Kalman filtering, which includes an inertial sub-filter, a wireless sub-filter, and a federated fusion unit. The construction of the positioning error dynamic correction model includes: setting up an inertial sub-filter, a wireless sub-filter, and a federated fusion unit; configuring the standardized inertial time-series dataset and the relative trajectory sequence of the wristband as the state prediction input of the inertial sub-filter; configuring the wireless observation sequence as the observation update input of the wireless sub-filter; configuring the constrained position sequence as the constrained observation input for consistency constraint update, and configuring the drift bias vector calculation and error state access structure; configuring a fusion weight calculation structure based on state uncertainty metric in the federated fusion unit, which is used to perform federated fusion update on the inertial state estimation results output by the inertial sub-filter and the wireless state estimation results output by the wireless sub-filter, forming the positioning error dynamic correction model; In the positioning error dynamic correction model, based on the standardized inertial time series dataset and the relative trajectory sequence of the wristband, state prediction processing is performed in the inertial sub-filter to obtain the inertial state prediction result and the inertial state uncertainty measure. The state prediction processing includes: in the inertial sub-filter, based on the triaxial acceleration and triaxial angular velocity data of adjacent sampling times in the standardized inertial time-series dataset, calculating the velocity change and heading change corresponding to the positioning time; based on the inertial state estimation result corresponding to the previous positioning time, accumulating the velocity change and heading change to the velocity state and heading state corresponding to the previous positioning time to obtain the velocity prediction and heading prediction corresponding to the positioning time; based on the velocity prediction, heading prediction and the time interval between adjacent positioning times, recursively updating the position state corresponding to the previous positioning time to obtain the position prediction corresponding to the positioning time; based on the state transition relationship in the inertial sub-filter, recursively propagating the inertial state uncertainty measure corresponding to the previous positioning time to calculate the inertial state uncertainty measure corresponding to the positioning time, wherein the inertial state uncertainty measure includes the position uncertainty measure, velocity uncertainty measure and heading uncertainty measure; The state transition relationship is obtained by: determining the recursive form of the position state between adjacent positioning times based on the position change relationship between adjacent position points in the relative trajectory sequence of the wristband; determining the recursive form of the velocity state between adjacent positioning times based on the integral relationship of the three-axis acceleration data between adjacent sampling times; determining the recursive form of the heading state between adjacent positioning times based on the integral relationship of the three-axis angular velocity data between adjacent sampling times; and unifying the position state transition relationship, velocity state transition relationship, and heading state transition relationship into the state transition relationship of the inertial sub-filter. In the wireless sub-filter, observation update processing is performed based on the state input of the wireless sub-filter to obtain the wireless state estimation result and the wireless state uncertainty measure. The observation update process includes: in the wireless sub-filter, constructing a wireless observation vector corresponding to the positioning time based on the wireless observation record corresponding to the positioning time in the wireless observation sequence; calculating the wireless observation residual based on the positional deviation between the wireless observation vector and the wireless state estimation result corresponding to the previous positioning time; performing observation update calculation on the uncertainty measure of the wireless state estimation result corresponding to the previous positioning time based on the observation mapping relationship in the wireless sub-filter to obtain the wireless state uncertainty measure corresponding to the positioning time; and performing state correction on the wireless state estimation result corresponding to the previous positioning time based on the wireless observation residual and the updated wireless state uncertainty measure to obtain the wireless state estimation result corresponding to the positioning time. The observation mapping relationship includes: wireless position state mapping relationship and observation uncertainty mapping relationship; wherein, the wireless position state mapping relationship is used to map the wireless state estimation result in the wireless sub-filter to the position observation mapping quantity corresponding to the wireless observation vector; the observation uncertainty mapping relationship is used to map the uncertainty measure corresponding to the wireless state estimation result to the observation uncertainty representation consistent with the wireless observation vector; The constraint residual sequence is determined based on the position deviation between the constrained position sequence and the inertial state prediction result and the position deviation between the constrained position sequence and the wireless state estimation result. Then, state correction processing is performed on the inertial state prediction result and the wireless state estimation result based on the constraint residual sequence to obtain the constrained inertial state and the constrained wireless state. The drift bias vector is calculated based on historical positioning results. The drift bias vector is the statistical result of the position deviation between the corrected position sequence record before the target positioning time and the corresponding constrained inertial state in the time dimension. The statistical result includes the mean deviation vector and the deviation change magnitude vector. The drift bias vector is written into the inertial sub-filter as the error state input, and the constrained inertial state is updated by error compensation based on the error state input to obtain the inertial state estimation result. The error compensation update includes: in the inertial sub-filter, mapping the drift bias vector to the error state space corresponding to the constrained inertial state to obtain the error compensation amount corresponding to the positioning time; performing deviation correction processing on the position, velocity, and flight vector in the constrained inertial state based on the error compensation amount to obtain the processed inertial state correction amount; performing synchronous correction and update on the inertial state uncertainty measure based on the uncertainty propagation relationship corresponding to the error state input in the inertial sub-filter to obtain an inertial state uncertainty measure consistent with the inertial state correction amount; and outputting the inertial state correction amount and the corrected inertial state uncertainty measure together as the inertial state estimation result. In the federated fusion unit, the inertial state estimation result and the constrained wireless state are updated by federated fusion to obtain the federated fusion position estimation result. The federated fusion update includes: determining the inertial fusion weight based on the inertial state uncertainty measure, determining the wireless fusion weight based on the wireless state uncertainty measure, and performing weighted fusion of the inertial state estimation result and the constrained wireless state based on the inertial fusion weight and the wireless fusion weight. The federated fusion location estimation results are combined in the order of positioning time to output the corrected location sequence.

[0029] In this embodiment, the extraction of the human body location determination result set includes the following specific steps: Based on the corrected position coordinates and timestamp records corresponding to adjacent positioning times in the corrected position sequence, the position change and time interval are calculated to obtain the trajectory smoothing input sequence; Smoothing processing is performed based on the trajectory smoothing input sequence. A smooth position sequence is generated by updating the coordinates of the position point corresponding to each positioning time to the weighted average coordinates of multiple position points in the time neighborhood. The weighted average coordinates are obtained by summing the product of the coordinates of each position point in the time neighborhood and the corresponding smoothing weight. The smoothing weight is the weight value obtained by inverting the time difference between the timestamp of each position point in the time neighborhood and the timestamp of the current positioning time, and then normalizing it in the time neighborhood. The position points in the smooth position sequence are rearranged according to the continuous time axis. A time interpolation relationship is constructed based on the timestamp intervals of adjacent position points. The resampled position points corresponding to each resampled time are calculated under the time interpolation relationship to generate a resampled position sequence. The resampled location sequences are combined according to the temporal order of the resampled times to generate a continuous indoor positioning trajectory. The human body position determination result set is extracted based on the continuous indoor positioning trajectory. The human body position determination result set includes the position coordinates and corresponding timestamp records corresponding to each resampling time in the continuous indoor positioning trajectory.

[0030] Example 1: To verify the feasibility of this invention in practice, it was applied to an indoor positioning scenario in a large commercial complex. This scenario has a complex internal spatial structure, including multi-level connected areas, enclosed rooms, long corridors, and densely populated areas. Indoor wireless signals are easily affected by obstruction and multipath effects. A single wireless positioning method suffers from significant position jumps and discontinuous trajectories, while relying solely on inertial calculations easily leads to cumulative drift, making it difficult to meet the actual requirements for positioning continuity and stability.

[0031] In this scenario, a user wearing a smart bracelet walks freely indoors. The bracelet continuously collects triaxial acceleration, triaxial angular velocity, and geomagnetic data, which are preprocessed by a data processing module to form a standardized inertial time-series dataset. Based on this dataset, the system automatically identifies gait events and performs pedestrian dead reckoning, generating a relative trajectory sequence from the bracelet reflecting the relative displacement relationship of walking. Simultaneously, wireless positioning infrastructure deployed within the venue continuously outputs Bluetooth, Wi-Fi, and UWB observation data. The system performs unified processing and reliability assessment on these wireless observation data, obtaining a wireless observation sequence and an observation reliability sequence. Subsequently, in a unified coordinate system, the system weights the wireless observation information according to the observation reliability and fuses it with the bracelet's relative trajectory sequence over time to form an initial fused position sequence. For potential wall penetration, boundary crossing, or speed anomalies in this sequence, the system applies trajectory rationality constraints based on the indoor map and reachable area information, correcting abnormal positions to obtain a constrained position sequence. Building upon this foundation, the system introduces a dynamic positioning error correction model based on federated Kalman filtering. This model integrates inertial information, wireless observation information, constrained position information, and historical positioning results to correct positioning errors online, effectively suppressing inertial drift and smoothing wireless observation fluctuations, outputting a stable corrected position sequence. Finally, the system smooths and resamples the corrected position sequence over time to generate continuous indoor positioning trajectories, forming a human location determination result set that can be directly used for personnel location display and behavior analysis.

[0032] In practical applications, this system maintains trajectory continuity even when users walk continuously, turn, and enter different functional areas, avoiding sudden location changes and unreasonable jumps. By comparing the position trajectory performance before and after system operation, it can be observed that the positioning results are significantly improved in terms of spatial continuity, path rationality, and long-term stability. Especially in areas with unstable or short-term wireless signals, the system can still maintain reliable positioning results by relying on inertial information and dynamic error correction mechanisms, thus fully demonstrating the beneficial effects of this invention in improving positioning accuracy and robustness in complex indoor environments.

[0033] Table 1. Performance Comparison of the Invention and Traditional Indoor Positioning Methods

[0034] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0035] The average positioning error index shows that the traditional inertial estimation method has an average positioning error of 2.85 meters, while the traditional wireless positioning method has an average error of 2.12 meters. The method of this invention reduces this index to 1.48 meters. This improvement does not rely on the accuracy improvement of a single sensor, but rather stems from the introduction of an observation reliability weighting mechanism under a unified coordinate system. This actively weakens the impact of poor signal quality from wireless observation on the positioning results. At the same time, it utilizes the relative trajectory constraint of the wristband to constrain the short-term displacement direction, thereby significantly reducing the overall error level in the fusion stage.

[0036] Regarding the standard deviation of errors, the traditional inertial estimation method and the traditional wireless positioning method have deviations of 1.34 meters and 1.02 meters, respectively, while the method of this invention has a deviation of 0.68 meters, indicating a significant reduction in fluctuation amplitude. This result directly reflects the synergistic effect of trajectory rationality constraints and the dynamic error correction mechanism of federated Kalman filtering in this invention. By introducing indoor map constraints, reachability area constraints, and velocity continuity constraints into the fusion process, the amplification effect of abnormal location points on the error distribution is effectively eliminated, making the positioning results more stable in the time dimension.

[0037] In terms of trajectory continuity, the method of this invention achieves 93.4%, which is a significant improvement compared to 86.2% for traditional inertial estimation methods and 88.9% for traditional wireless positioning methods. This improvement mainly stems from the introduction of historical positioning results into the drift bias estimation during the dynamic correction stage of positioning errors. This allows the system to maintain the temporal continuity of the trajectory even when wireless observations are interrupted or inertial errors accumulate, thereby reducing trajectory interruptions and breaks.

[0038] In terms of the number of position jumps, the method of this invention achieves 3.7 jumps per 100 meters, significantly lower than the 7.6 jumps per 100 meters of the traditional inertial estimation method and the 5.9 jumps per 100 meters of the traditional wireless positioning method. This improvement directly corresponds to the processing logic in this invention that imposes trajectory rationality constraints on the initial fused position sequence. Through wall crossing determination, region connectivity determination, and velocity continuity correction, the system can proactively correct unreasonable jump positions during the fusion phase, rather than smoothing them out afterward, thus reducing the frequency of jumps from the source.

[0039] Regarding the long-term drift rate, the method of this invention achieves a value of 0.21 m / min, which is further reduced compared to 0.42 m / min for the traditional inertial estimation method and 0.27 m / min for the traditional wireless positioning method. This effect stems from the multi-sub-filter cooperative correction mechanism under the federated Kalman filter framework. Through information complementarity between the inertial sub-filter and the wireless sub-filter, and continuous updating of the drift bias vector, the accumulated time error is effectively suppressed.

[0040] In terms of effective positioning coverage, the method of this invention achieves 96.4%, which is higher than the 91.5% of the traditional inertial estimation method and the 93.8% of the traditional wireless positioning method. This indicates that in complex indoor environments, even if the wireless signal is blocked or partially failed, the system can still maintain stable output by relying on inertial and historical correction information, significantly improving overall availability.

[0041] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A high-precision indoor positioning system based on a smart bracelet, characterized in that, The method comprises the following steps: a data processing module is used to acquire multi-source perception data collected by the smart bracelet and perform preprocessing to generate a standardized inertial time series dataset; a gait estimation module is used to perform gait event detection and pedestrian dead reckoning processing based on the standardized inertial time series dataset, and output a bracelet-end relative trajectory sequence; a wireless observation module is used to acquire indoor wireless positioning observation data and perform preprocessing and reliability evaluation to generate a wireless observation sequence and an observation reliability sequence; a fusion positioning module is used to weight the wireless observation sequence based on the observation reliability sequence in a unified coordinate system, and perform time series fusion with the bracelet-end relative trajectory sequence to output an initial fusion position sequence; a trajectory constraint module is used to impose trajectory rationality constraints on the initial fusion position sequence and correct abnormal position points to output a constrained position sequence; an error correction module is used to construct a positioning error dynamic correction model based on federated Kalman filtering based on the standardized inertial time series dataset, the wireless observation sequence, the constrained position sequence and historical positioning results, and perform online correction processing to output a corrected position sequence; a trajectory output module is used to perform smoothing and time resampling processing on the corrected position sequence to generate a continuous indoor positioning trajectory and extract a human position determination result set. 2.The high-precision indoor positioning system based on smart bracelet of claim 1, wherein, The modules are realized through the following methods: multi-source perception data collected by the smart bracelet is acquired and preprocessed to obtain a standardized inertial time series dataset; gait event detection and pedestrian dead reckoning processing are performed based on the standardized inertial time series dataset to obtain a bracelet-end relative trajectory sequence; indoor wireless positioning observation data output by indoor wireless positioning infrastructure is acquired, preprocessed and reliability evaluated to obtain an observation reliability sequence and a wireless observation sequence; the wireless observation sequence is weighted in a unified coordinate system, and the weighted wireless observation sequence is fused with the bracelet-end relative trajectory sequence in time series to output an initial fusion position sequence; trajectory rationality constraints are imposed on the initial fusion position sequence and abnormal position points are corrected to obtain a constrained position sequence; a positioning error dynamic correction model based on federated Kalman filtering is constructed based on the standardized inertial time series dataset, the wireless observation sequence, the constrained position sequence and historical positioning results, and online correction processing is performed to output a corrected position sequence; the corrected position sequence is smoothed and time resampled to generate a continuous indoor positioning trajectory and extract a human position determination result set. 3.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The multi-source perception data includes three-axis acceleration data, three-axis angular velocity data and geomagnetic data, and the preprocessing includes timestamp alignment, outlier rejection and noise filtering. 4.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The bracelet-end relative trajectory sequence is obtained through the following specific steps: gravity component separation and coordinate system unification are performed on the three-axis acceleration data and three-axis angular velocity data in the standardized inertial time series dataset to obtain vertical acceleration and angular velocity sequences; gait event detection is performed based on the vertical acceleration sequence, and peak detection and peak interval constraint are used to determine continuous gait event times to obtain a gait event sequence; calculating a step frequency parameter based on the gait event sequence, the step frequency parameter being an inverse of a time interval between two adjacent gait events; extracting a gait intensity feature based on a time interval corresponding to the gait event sequence and calculating a step length parameter; calculating a heading angle sequence based on the angular velocity sequence and the geomagnetic data, each time point of the heading angle sequence being obtained by jointly correcting a rotation angle amount obtained by integrating the angular velocity and a geomagnetic direction constraint; calculating a heading increment parameter based on the heading angle sequence, the heading increment parameter being a sequence of steering change amounts formed by differences between heading angles corresponding to two adjacent gait event time points; performing pedestrian dead reckoning processing based on the gait event sequence, the step length parameter and the heading increment parameter, and performing position updating once at each time when a gait event is detected; combining relative position points obtained through multiple times of position updating according to a time sequence corresponding to the gait events to obtain a continuous relative position sequence, and outputting the continuous relative position sequence as a bracelet-end relative trajectory sequence. 5.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The observation credibility sequence and the wireless observation sequence are obtained by including the following specific steps: obtaining indoor wireless positioning observation data output by indoor wireless positioning infrastructure within a target positioning period, and performing data format unification and timestamp standardization processing to obtain a set of original wireless observation records with timestamp identification; analyzing observation types and observation sources based on the set of original wireless observation records to obtain an observation type field and an observation source field; performing observation value extraction and observation value standardization processing on the set of original wireless observation records based on the observation type field and the observation source field to generate a set of standardized wireless observation records; performing observation time alignment processing on the set of standardized wireless observation records, and using a time matching method taking a positioning time point as an alignment reference to map standardized wireless observation records with a time difference less than a time threshold from the positioning time point to the corresponding positioning time point to obtain a wireless observation sequence organized according to positioning time points; calculating an initial observation credibility based on wireless observation records at each positioning time point in the wireless observation sequence to obtain an initial observation credibility sequence; updating the initial observation credibility sequence by using a ratio of a value of the initial observation credibility corresponding to each positioning time point to a maximum initial observation credibility value in the initial observation credibility sequence as a normalized credibility to generate an observation credibility sequence, and outputting the observation credibility sequence and the wireless observation sequence after timestamp association. 6.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The output of the initial fusion position sequence includes the following specific steps: obtaining the wireless observation sequence, the observation credibility sequence and the bracelet-end relative trajectory sequence, and converting the wireless observation sequence and the bracelet-end relative trajectory sequence to coordinate expressions in a unified coordinate system to obtain a wireless observation position sequence and a relative trajectory position sequence in the unified coordinate system; performing weighted processing on the wireless observation position sequence in the unified coordinate system based on the observation credibility sequence, and calculating a credibility weight for a wireless observation position at each positioning time point according to a corresponding observation credibility value; For the case that multiple wireless observation position records exist at the same positioning moment, a weighted wireless observation position is calculated based on the reliability weight corresponding to each wireless observation position record, and the weighted wireless observation positions obtained at each positioning moment are combined in time sequence to obtain a weighted wireless observation position sequence; Time alignment processing is performed on the weighted wireless observation position sequence and the relative trajectory position sequence, and a time matching method taking the positioning moment as the alignment reference is adopted to map the relative trajectory positions with a time difference less than a time threshold from the positioning moment to the corresponding positioning moment, thereby obtaining an aligned relative trajectory position sequence organized by positioning moment; Time sequence fusion is performed on the weighted wireless observation position and the aligned relative trajectory position at each positioning moment, and the fusion positions obtained at each positioning moment are combined in time sequence to output an initial fusion position sequence. 7.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The obtaining of the constraint position sequence includes the following specific steps: Obtain indoor map data and perform analysis to obtain an indoor map constraint data set; Perform indoor map constraint processing on the initial fusion position sequence based on the indoor map constraint data set, and perform wall crossing judgment and passage judgment on each position point in the initial fusion position sequence to obtain a map constraint judgment sequence; Perform map constraint correction processing on the position points in the map constraint judgment sequence that are judged to be invalid, calculate a correction mapping position based on the minimum spatial distance between the invalid position point and the boundary of the nearest feasible passage region, and replace the invalid position point with the map correction position point as the correction mapping position to obtain a map constraint position sequence; Obtain reachable region data and parse the reachable region data into a reachable region constraint data set; Perform reachable region constraint processing on the map constraint position sequence based on the reachable region constraint data set, and perform region attribution judgment and region connectivity judgment on each position point in the map constraint position sequence to obtain a reachable region judgment sequence; Perform reachable region correction processing on the position points in the reachable region judgment sequence that are judged to be unreachable, generate reachable correction position points by mapping the unreachable position points to the nearest points on the adjacent reachable boundaries, and replace the unreachable position points with the reachable correction position points to obtain a reachable region constraint position sequence; Perform speed continuity constraint processing based on the reachable region constraint position sequence, calculate the displacement distance and time interval between the position points corresponding to adjacent positioning moments, and obtain an instantaneous speed sequence; Perform speed continuity constraint judgment on the instantaneous speed sequence, and assign a continuity judgment mark to each instantaneous speed value according to the size relationship between the instantaneous speed value and the corresponding allowable speed upper limit in the instantaneous speed sequence to obtain a speed continuity judgment sequence; Perform speed continuity correction processing on the position points in the speed continuity judgment sequence that are judged to be overspeed, construct a position interpolation relationship based on the overspeed position point and the adjacent valid position points before and after it, calculate a speed correction position point, and replace the overspeed position point with the speed correction position point to obtain a constraint position sequence. 8.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The output of the corrected position sequence includes the following specific steps: The positioning error dynamic correction model based on federated Kalman filtering is constructed based on a standardized inertial time series data set, a wireless observation sequence, a constraint position sequence, and historical positioning results; The positioning error dynamic correction model is a multi-sub-filter collaborative positioning correction model based on federated Kalman filtering, including an inertial sub-filter, a wireless sub-filter, and a federated fusion unit; In the positioning error dynamic correction model, based on the standardized inertial time series data set and the relative trajectory sequence of the bracelet end, state prediction processing is performed in the inertial sub-filter to obtain inertial state prediction results and inertial state uncertainty metrics; In the wireless sub-filter, observation update processing is performed based on wireless sub-filter state inputs to obtain wireless state estimation results and wireless state uncertainty metrics; The constraint residual sequence is determined according to the position deviation between the constraint position sequence and the inertial state prediction results, and the position deviation between the constraint position sequence and the wireless state estimation results, and the state correction processing is performed on the inertial state prediction results and the wireless state estimation results based on the constraint residual sequence, respectively, to obtain the constraint inertial state and the constraint wireless state; The drift bias vector is calculated based on the historical positioning results; The drift bias vector is written as an error state input into the inertial sub-filter, and error compensation update is performed on the constraint inertial state based on the error state input to obtain the inertial state estimation results; In the federated fusion unit, federated fusion update is performed on the inertial state estimation results and the constraint wireless state to obtain the federated fusion position estimation results; The federated fusion position estimation results are combined in sequence according to the positioning time to output the corrected position sequence. 9.The high-precision indoor positioning system based on smart bracelet of claim 2, wherein, The extraction of the human position determination result set includes the following specific steps: Based on the corrected position coordinates and timestamp records corresponding to adjacent positioning time in the corrected position sequence, the position change and time interval are calculated to obtain the trajectory smoothing input sequence; Based on the trajectory smoothing input sequence, smoothing processing is performed, and the smoothed position sequence is generated by updating the position point coordinates corresponding to each positioning time to the weighted average coordinates of multiple position point coordinates in the time neighborhood; According to the continuous time axis, the position points in the smoothed position sequence are time rearranged, the time interpolation relationship is constructed based on the time stamp interval corresponding to adjacent position points, and the resampling position points corresponding to each resampling time are calculated based on the time interpolation relationship to generate the resampling position sequence; The resampling position sequence is combined in sequence according to the time of the resampling time to generate the continuous indoor positioning trajectory; Based on the continuous indoor positioning trajectory, the human position determination result set is extracted, which includes the position coordinates and corresponding timestamp records corresponding to each resampling time in the continuous indoor positioning trajectory.

Images