VR space accurate positioning method based on head and limb tracking
By introducing a delay compensation feedback loop and a dynamic scene semantic library into the VR positioning system, combined with adaptive adjustment of user state, the problems of multimodal synchronization, posture error, scene feature changes and multi-user conflicts in VR positioning are solved, achieving accurate positioning and improved user comfort.
Patent Information
- Application Number
- CN202511575344.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-31
- Publication Date
- 2025-12-02
AI Technical Summary
Existing VR positioning technologies suffer from problems such as clock drift, transmission jitter, fixed weight fusion, single-band optical positioning, assumption of rigid skeletons, static pose calculation, single-modal prediction, lack of real-time updates for scene feature changes, equal bandwidth for multiple users, and high fatigue and overheating of devices, leading to positioning errors and user discomfort.
A multimodal data quality assessment is constructed by using a delay compensation feedback loop, dynamically adjusting the fusion weights, correcting joint angles with muscle force, establishing a dynamic scene semantic library and updating scene features in real time, allocating user priorities and bandwidth, automatically optimizing the sampling rate, constructing an error database, and achieving rapid parameter adaptation.
It achieves accurate positioning in dynamic scenarios, reduces positioning errors and user discomfort, improves system adaptability and stability, and reduces device overheating and multi-user conflicts.
Smart Images
Figure CN121048637A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of spatial positioning technology, and more specifically to a VR spatial precision positioning method based on head and limb tracking. Background Technology
[0002] In existing technologies, traditional VR positioning often uses the PTP protocol for multimodal data synchronization, but cross-modal transmission suffers from clock drift and transmission jitter, and latency is mostly compensated for with fixed values. Data fusion often uses preset fixed weights and does not consider the real-time noise level of sensors. Optical positioning relies on a single band, and is prone to losing marker points in strong light or low-texture scenes. Traditional VR pose estimation relies on IMU and optical marker points, assuming that the skeleton is a rigid structure and ignoring soft tissue deformation caused by muscle contraction. Joint angle calculation uses a static α angle and does not consider dynamic pose changes. Intent prediction often relies on a single modality and is susceptible to electromagnetic interference or latency. It relies on preset scene maps, which cannot be updated in real time when scene features change, leading to positioning misjudgments. In multi-user scenarios, equal bandwidth allocation is used, which can easily cause conflicts between high-priority users and low-priority users. Traditional VR positioning uses a fixed sampling rate, which can lead to increased device overheating and user discomfort under high fatigue conditions. When deploying in new scenes, positioning parameters need to be manually adjusted, which is time-consuming and relies on experience. Therefore, there is a need to provide a precise VR spatial positioning method based on head and limb tracking. Summary of the Invention
[0003] The purpose of this invention is to provide a VR spatial precision positioning method based on head and limb tracking. To solve the above-mentioned problems in the prior art, this invention achieves this through the following technical solution: In a first aspect, the VR spatial precise positioning method based on head and limb tracking provided in the embodiments of the present invention specifically includes the following steps: Step 1: By adding a delay compensation feedback loop, construct a multimodal data quality assessment index to quantify data credibility and assign pre-fusion weights; Step 2: Based on the adjusted sensor data, collect typical user postures and calculate the dynamic α angle to reduce posture errors. Convert EMG signals into muscle force and correct joint angles and marker offsets. Combine single-modal intent prediction with dual-modal fusion. Adaptively adjust the error compensation factor according to the motion state. Reduce the deviation between the intent angle and the measured angle by calculating the total error compensation. Step 3: Based on error compensation, establish a dynamic scene semantic library and update scene features in real time. Identify new features through semantic segmentation and update the semantic library to avoid misjudgment of localization. Assign user priority and allocate bandwidth based on scene type and action importance to resolve conflict issues when multiple users are densely concentrated. Step 4: Collect state parameters based on scene features and fuse them to obtain a comprehensive user state index. Automatically optimize the sampling rate, calculate the positioning error and analyze its source to correct the optimal parameters, build an error database, and call the optimal parameters for new scenes through matching.
[0004] Secondly, the VR spatial precision positioning system based on head and limb tracking provided in this embodiment of the invention specifically includes the following modules: Synchronization processing module: By adding a delay compensation feedback loop, a multimodal data quality assessment index is constructed to quantify data credibility and pre-fusion weights are assigned; Fusion computing module: Based on the adjusted sensor data, it collects typical user postures and calculates dynamic α angle to reduce posture error. It converts EMG signals into muscle force and corrects joint angles and marker offsets. It combines single-modal intent prediction to perform dual-modal fusion, adaptively adjusts error compensation factors according to motion state, and reduces the deviation between the intent angle and the measured angle by calculating the total error compensation. Management and scheduling module: Based on error compensation, it establishes a dynamic scene semantic library and updates scene features in real time. It identifies new features through semantic segmentation and updates the semantic library to avoid misjudgment of localization. It allocates user priority and bandwidth based on scene type and action importance to solve the conflict problem when multiple users are densely concentrated. Perception optimization module: Collects state parameters based on scene features and fuses them to obtain a comprehensive user state index, automatically adjusts the sampling rate, calculates positioning errors and analyzes their sources to correct the optimal parameters, builds an error database, and calls the optimal parameters for new scenes through matching.
[0005] The beneficial effects of this invention are: 1. It breaks through the limitations of fixed delay compensation and static weight fusion in traditional multimodal data synchronization. It achieves dynamic delay convergence through a feedback loop, adaptive weighting through quality quantization, and dynamically adjusts the optical strategy based on scene characteristics. It calculates the dynamic α angle based on typical user postures, replacing the static α angle and controlling the error within a preset range. It converts EMG signals into muscle force, corrects joint angles and marker offsets, and compensates for measurement errors caused by muscle exertion. It integrates EEG and EMG single-modal intent prediction, outputs the fused motion direction angle through confidence weighting, and adaptively adjusts the error compensation factor based on motion state. It deeply integrates muscle force modeling and intent prediction into posture correction, and solves the positioning deviation caused by skeletal and muscle deformation in dynamic movements through coupling analysis. 2. Establish a dynamic scene semantic library, update scene features in real time through semantic segmentation, and iterate the semantic library based on the difference between old and new features and update weights; allocate user priorities based on scene type and action importance, and dynamically allocate bandwidth according to priority to reduce the conflict rate of high-priority users; combine scene semantic understanding with multi-user resource scheduling, and avoid positioning misjudgment through dynamic scene updates; integrate EEG fatigue, heart rate variability, and joint angular velocity mean to obtain a comprehensive state index, and automatically adjust the sampling rate accordingly to balance accuracy and device load; establish an error database containing scene, environment, and user state, match new scenes through the KNN algorithm and call the optimal parameters to achieve rapid adaptation of positioning parameters; dynamically associate user physiological state with positioning system parameters to form a closed-loop optimization of state, sampling rate, and error correction. Attached Figure Description
[0006] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0007] Figure 1 This is a flowchart of the steps of the VR spatial precision positioning method based on head and limb tracking provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the VR spatial precision positioning system based on head and limb tracking provided in Embodiment 2 of the present invention. Detailed Implementation
[0008] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0009] Example 1: As Figure 1 As shown in the figure, the VR spatial precise positioning method based on head and limb tracking provided in this embodiment of the invention specifically includes the following steps: Step 1: By adding a delay compensation feedback loop, construct a multimodal data quality assessment index to quantify data credibility and assign pre-fusion weights; In a specific embodiment, a five-level topology of head-mounted display, limb sensor network, base station, synchronization controller and edge unit is adopted, and the data flow is clearly displayed through Mermaid diagram; the limb end IMU and EMG electrodes are integrated and packaged with flexible PCB, and the sampling point spacing is preset to be less than 5mm to ensure that the electromyographic signal and inertial data come from the same joint area and avoid errors caused by spatial misalignment; Real-time acquisition of sensor signal parameters, including: IMU noise variance, EEG signal-to-noise ratio, and full clock time difference; A delay compensation feedback loop is designed, and delay monitoring is added to the PTP-EMG synchronization process. The difference between the measured delay and the fixed delay is calculated to obtain the feedback delay difference. The feedback delay difference is multiplied by a preset feedback coefficient and summed with the fixed delay to obtain the dynamic compensation amount. The preset feedback coefficient delay is preset to 0.2. Through multiple dynamic tests and calibrations, it is ensured that the measured delay is stable within the range of [0.1ms, 0.2ms]. Based on the obtained signal parameters, a three-dimensional weighted fusion calculation is performed to obtain the data quality assessment index. Based on the obtained data quality assessment index, the fusion weights of the sensors are dynamically adjusted. The sum of the data quality assessment indices of all sensors is calculated, and the proportion of each sensor's data quality assessment index in the sum of the data quality assessment indices is analyzed as the fusion weights of the sensors. This solves the fusion error problem caused by high-noise data still having high weights, and at the same time provides a dynamic basis for fusion. Step 2: Based on the adjusted sensor data, collect typical user postures and calculate the dynamic α angle to reduce posture errors. Convert EMG signals into muscle force and correct joint angles and marker offsets. Combine single-modal intent prediction with dual-modal fusion. Adaptively adjust the error compensation factor according to the motion state. Reduce the deviation between the intent angle and the measured angle by calculating the total error compensation. In a specific embodiment, shoulder girdle point clouds of three typical user postures are collected. Typical postures include: natural drooping, abduction at 90° and elevation at 180°. The scapular plane is fitted for each typical posture and the α angle is calculated. The dynamic α angle is obtained by weighted averaging. The posture weights are calibrated based on the frequency of daily movements. To address the issue of excessive error in static α angle during dynamic motion, the error of dynamic α angle is controlled within a preset range; Muscle force calculation is based on EMG and a muscle model. The Hill muscle model is used to convert EMG electromyography signals into muscle force using the following formula: Calculate muscle strength ,in, The maximum isometric contractile force of the muscle. For EGM force mapping coefficients, EMG electromyography signal The actual joint angle at which the muscle attaches to the joint. The optimal joint angle for muscle contraction; Based on the fact that muscle force causes additional deviation in joint angle, for example, the deltoid muscle exertion increases the abduction angle of the joint; by calibrating the force-angle conversion coefficient through dynamic force and skeletal joint angle testing, the original joint angle measured by the IMU is obtained, the calculated muscle force is multiplied by the force-angle conversion coefficient to obtain the compensated joint angle, and the compensated joint angle is summed with the original joint angle measured by the IMU to obtain the corrected joint angle, thereby reducing the measurement error of the joint angle when the muscle exerts force; Nonlinear calculations are performed on muscle contraction, and the effect of force on soft tissue deformation is reflected by a muscle force correction term, using the formula: The offset of the marker point caused by muscle contraction at time t is calculated, where, This is the soft tissue coefficient, reflecting the mapping relationship between EMG amplitude and contraction amount. For time steps, For the sampling time window, EMG electromyography signal This is the square integral of the EMG electromyography signal within the sampling time window. For integration variables, Based on the offset, It is the hyperbolic tangent function. The peak value of the EMG signal represents the starting point of muscle activation. The preset time constant for muscle response. This is the force-angle conversion factor. For muscle strength; Based on the obtained marker point offset, the original marker point coordinates acquired by the optical camera are summed with the marker point offset to obtain the corrected marker point coordinates; The predicted motion direction angles were calculated by analyzing EEG single-modal intention prediction and EMG single-modal intention prediction respectively. Specifically, EEG single-mode prediction aims to predict the magnitude of potential changes via motion-related potentials (MRP). , Combine the motion direction angle EEG prediction value output by the SVM classifier The confidence level is calculated based on the obtained EEG prediction value of the motion direction angle, using the formula: Calculate the EEG confidence level ; EMG Single-Mode Intent Prediction: By utilizing the activation mode of the EMG, template matching is used to output the predicted motion direction angle EMG value. The obtained EMG electromyography signal is compared with the preset maximum EMG value to calculate the EMG confidence level. ; The motion direction angle is predicted by EEG based on EEG single-modal intention prediction analysis. and EEG confidence The motion direction angle predicted by EMG is obtained by combining EMG single-mode intention prediction analysis. and EMG confidence To perform dual-modal fusion prediction, the fusion formula is: The direction angle of fusion motion was calculated. ; It should be noted that the confidence level of a single EEG under electromagnetic interference is based on the fusion of motion direction angle. EMG prediction of motion direction angle when reduced to a preset confidence threshold The error is compensated by the delay of the motion direction angle EMG prediction value, thereby reducing the fusion prediction delay. Adaptive error compensation is performed based on the target's motion state to obtain the target's motion velocity. A preset base attenuation factor is then dynamically adjusted in conjunction with the motion velocity. The product of the preset adjustment factor and the motion velocity is then summed with the base attenuation factor to obtain the dynamic attenuation factor. This addresses the problem of insufficient drift suppression when the fixed base attenuation factor is applied to the target at high speed, thereby reducing drift error. Based on the obtained dynamic attenuation factor, the dual-modal error compensation formula is used: The total amount of error compensation was calculated. ,in, The time for the end of the integration process. The starting time of the integration. For integration variables, For angular velocity gain, The integral rate of change of the IMU angular velocity. For IMU angular velocity, For electromyography adaptation coefficient, This is an electromyographic signal. To integrate the angle difference correction term between the intended angle and the IMU measured angle, From the perspective of intent, This is the actual measured angle of the IMU, further reducing the deviation between the intended angle and the actual measured angle; If the target is stationary or in motion, i.e., EMG electromyography signal The formula then becomes: Focus on suppressing IMU cumulative drift; If the target moves at a constant velocity, i.e., the integral rate of change of the IMU angular velocity. The formula then becomes: It focuses on compensating for physiological deviations caused by muscle contraction; Step 3: Based on error compensation, establish a dynamic scene semantic library and update scene features in real time. Identify new features through semantic segmentation and update the semantic library to avoid misjudgment of localization. Assign user priority and allocate bandwidth based on scene type and action importance to resolve conflict issues when multiple users are densely concentrated. In a specific embodiment, a dynamic scene semantic library is established to update scene features in real time; The dynamic scene semantic library includes: scene type, static features, dynamic features, and environmental parameters; The edge unit collects scene data according to a preset collection cycle. The scene data includes, but is not limited to, LiDAR point clouds and illumination sensor values. New dynamic features are identified through a semantic segmentation model, and the semantic database is updated accordingly. The new scene features are subtracted from the old scene features, and the result of the subtraction is multiplied by the preset update weight. At the same time, it is combined with the historical semantic library to obtain the updated semantic library, which solves the problem of increased localization misjudgment rate caused by the addition of obstacles in the scene. User priority is assigned based on scenario type and the importance of user actions; For example, in a medical scenario: the surgeon's user priority is preset to 5, and the assistant's user priority is preset to 3; in an industrial scenario: the user priority of the machine operator is preset to 4, and the user priority of the monitoring user is preset to 2. Based on the obtained priorities, different frequency bandwidths are allocated to users with different priorities, using the following formula: Obtain user bandwidth ,in, for, Prioritize users To prioritize the highest-priority users, this addresses the issue of increased conflict rates among low-priority users during peak user activity, while also reducing conflict rates among high-priority users. Step 4: Collect state parameters based on scene features and fuse them to obtain a comprehensive user state index. Automatically optimize the sampling rate, calculate the positioning error and analyze its source to correct the optimal parameters, build an error database, and call the optimal parameters for new scenes by matching. In a specific embodiment, the user's state is perceived from multiple dimensions, and state parameters are collected and quantified. The state parameters include: EEG fatigue level, heart rate variability and mean joint angular velocity. In obtaining EEG signals The proportion of wave power to total brainwave power is used to obtain EEG fatigue level; The heart rate variability is obtained by collecting data through the PPG sensor of the head-mounted display, calculating the standard deviation of the heart rate interval and comparing it with a preset standard deviation threshold. The average joint angular velocity of the user's limb joints is obtained by IMU measurement; The obtained EEG fatigue level, heart rate variability and mean joint angular velocity are fused and normalized to obtain the user's comprehensive status index. The sampling rate is automatically optimized based on the obtained state comprehensive index. The complement of the state comprehensive index with respect to 1 is calculated and multiplied with the preset compensation sampling frequency to obtain the sampling compensation frequency. The sampling compensation frequency is summed with the preset sampling frequency to obtain the automatic sampling frequency, which solves the problem of equipment overheating and user discomfort caused by high sampling rate under high fatigue. In a scenario where the precise location is known, calculate the error between the final positioning location and the reference position; The difference between the final location and the reference location is calculated, and the L2 norm of the difference result is taken to obtain the positioning error. Analyze the sources of positioning error and correct the parameters. If the positioning error is greater than the preset error threshold and mainly comes from skeleton modeling, adjust the pose weights of skeleton modeling. If the positioning error is greater than the preset error threshold and mainly comes from error compensation, adjust the base attenuation factor of the dynamic attenuation factor. If the positioning error is less than or equal to the preset error threshold, no parameter correction is performed. Based on the obtained positioning error, an error database is constructed. The database entries include: scene type, environmental parameters, user status, optimal parameters, and positioning error. Environmental parameters include: light intensity and electromagnetic interference intensity; user status: comprehensive state index and EEG fatigue level; optimal parameters include: dynamic α angle, dynamic attenuation factor and attitude weight. When a new scene is started, the KNN algorithm is used to match the most similar scene in the database and the optimal parameters are directly called.
[0010] Example 2: Figure 2 As shown, the VR spatial precision positioning system based on head and limb tracking provided in this embodiment of the invention specifically includes the following modules: Synchronization processing module: By adding a delay compensation feedback loop, a multimodal data quality assessment index is constructed to quantify data credibility and pre-fusion weights are assigned; Fusion computing module: Based on the adjusted sensor data, it collects typical user postures and calculates dynamic α angle to reduce posture error. It converts EMG signals into muscle force and corrects joint angles and marker offsets. It combines single-modal intent prediction to perform dual-modal fusion, adaptively adjusts error compensation factors according to motion state, and reduces the deviation between the intent angle and the measured angle by calculating the total error compensation. Management and scheduling module: Based on error compensation, it establishes a dynamic scene semantic library and updates scene features in real time. It identifies new features through semantic segmentation and updates the semantic library to avoid misjudgment of localization. It allocates user priority and bandwidth based on scene type and action importance to solve the conflict problem when multiple users are densely concentrated. Perception optimization module: Collects state parameters based on scene features and fuses them to obtain a comprehensive user state index, automatically adjusts the sampling rate, calculates positioning errors and analyzes their sources to correct the optimal parameters, builds an error database, and calls the optimal parameters for new scenes through matching.
[0011] The above provides a detailed description of one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. The above formulas are all dimensionless numerical calculations, and the formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world situation. The preset parameters in the formulas are set by those skilled in the art based on actual conditions and historical experience, and can be adjusted according to actual conditions. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. All equivalent changes and improvements made in accordance with the scope of the present invention should still fall within the patent coverage of the present invention.
Claims
1. A VR spatial precision positioning method based on head and limb tracking, characterized in that, Includes the following steps: By incorporating a delay compensation feedback loop, a data quality assessment index is calculated to quantify data credibility, and pre-fusion weights are assigned. Based on the adjusted sensor to collect typical user postures and calculate dynamic α angle, EMG signals are converted into muscle force to correct joint angles and marker offsets. By combining single-modal intent prediction with dual-modal fusion, the error compensation factor is adaptively adjusted according to the motion state, and the deviation between the intent angle and the measured angle is reduced by calculating the total error compensation. Based on error compensation, a dynamic scene semantic library is established and scene features are updated in real time. New features are identified and the semantic library is updated through semantic segmentation. User priority and bandwidth allocation are based on scenario type and action importance; Based on scene features, state parameters are collected and fused to obtain a comprehensive user state index. The sampling rate is automatically optimized, the positioning error is calculated and its source is analyzed to correct the optimal parameters, and an error database is built. For new scenes, the optimal parameters are called by matching.
2. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for calculating the data quality assessment index is as follows: Design a delay compensation feedback loop, calculate the difference between the measured delay and the fixed delay to obtain the feedback delay difference, multiply the feedback delay difference by the preset feedback coefficient and sum it with the fixed delay to obtain the dynamic compensation amount; Based on the obtained signal parameters, a three-dimensional weighted fusion calculation is performed to obtain the data quality assessment index. Based on the obtained data quality assessment index, the fusion weights of the sensors are dynamically adjusted, the sum of the data quality assessment indices of all sensors is calculated, and the proportion of each sensor's data quality assessment index in the sum of the data quality assessment indices is analyzed as the fusion weight of the sensors.
3. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for converting EMG signals into muscle force is as follows: Muscle force calculation based on EMG and muscle models: EMG electromyographic signals are converted into muscle force using the formula: Calculate muscle strength ,in, The maximum isometric contractile force of the muscle. For EGM force mapping coefficients, EMG electromyography signal The actual joint angle at which the muscle attaches to the joint. The optimal joint angle for muscle contraction.
4. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for obtaining the marker point offset is as follows: Based on the fact that muscle force causes additional deviation in joint angle, the force-angle conversion coefficient is calibrated by dynamic force and skeletal joint angle test, the original joint angle measured by IMU is obtained, the calculated muscle force is multiplied by the force-angle conversion coefficient to obtain the compensated joint angle, and the compensated joint angle is summed with the original joint angle measured by IMU to obtain the corrected joint angle, thereby reducing the measurement error of joint angle when muscle force is exerted. Nonlinear calculations are performed on muscle contraction, and the effect of force on soft tissue deformation is reflected by a muscle force correction term, using the formula: The offset of the marker point caused by muscle contraction at time t is calculated, where, This is the soft tissue coefficient. For time steps, For the sampling time window, EMG electromyography signal This is the square integral of the EMG electromyography signal within the sampling time window. For integration variables, Based on the offset, It is the hyperbolic tangent function. The peak time of the EMG signal. The preset time constant for muscle response. This is the force-angle conversion factor. For muscle strength; Based on the obtained marker offset, the original marker coordinates acquired by the optical camera are summed with the marker offset to obtain the corrected marker coordinates.
5. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for performing dual-modal fusion is as follows: The predicted motion direction angles were calculated by analyzing EEG single-modal intention prediction and EMG single-modal intention prediction respectively. EEG single-mode intention prediction: potential changes via motion-related potentials (MRP) , Combine the motion direction angle EEG prediction value output by the SVM classifier The confidence level is calculated based on the obtained EEG prediction value of the motion direction angle, using the formula: Calculate the EEG confidence level ; EMG Single-Mode Intent Prediction: By utilizing the activation mode of the EMG, template matching is used to output the predicted motion direction angle EMG value. The obtained EMG electromyography signal is compared with the preset maximum EMG value to calculate the EMG confidence level. ; The motion direction angle is predicted by EEG based on EEG single-modal intention prediction analysis. and EEG confidence The motion direction angle predicted by EMG is obtained by combining EMG single-mode intention prediction analysis. and EMG confidence Perform dual-modal fusion prediction and calculate the fusion motion direction angle. .
6. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for calculating the total error compensation is as follows: Adaptive error compensation is performed based on the target's motion state to obtain the target's motion velocity. A preset base attenuation factor is then dynamically adjusted in conjunction with the motion velocity. The product of the preset adjustment factor and the motion velocity is then summed with the base attenuation factor to obtain the dynamic attenuation factor. This addresses the problem of insufficient drift suppression when the fixed base attenuation factor is applied to the target at high speed, thereby reducing drift error. Based on the obtained dynamic attenuation factor, the dual-modal error compensation formula is used: The total amount of error compensation was calculated. ,in, The time for the end of the integration process. The starting time of the integration. For integration variables, For angular velocity gain, The integral rate of change of the IMU angular velocity. For IMU angular velocity, For electromyography adaptation coefficient, This is an electromyographic signal. To integrate the angle difference correction term between the intended angle and the IMU measured angle, From the perspective of intent, This is the angle measured by the IMU.
7. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for updating the semantic library is as follows: Establish a dynamic scene semantic library and update scene features in real time; The dynamic scene semantic library includes: scene type, static features, dynamic features, and environmental parameters; The edge unit collects scene data according to a preset collection cycle. The scene data includes, but is not limited to, LiDAR point clouds and illumination sensor values. New dynamic features are identified through a semantic segmentation model, and the semantic database is updated accordingly. The new scene features are subtracted from the old scene features, and the result of the subtraction is multiplied by the preset update weights. This result is then combined with the historical semantic database to obtain the updated semantic database.
8. The VR spatial precision positioning method based on head and limb tracking according to claim 1, characterized in that, The method for allocating bandwidth is as follows: User priority is assigned based on scenario type and the importance of user actions; Based on the obtained priorities, different frequency bandwidths are allocated to users with different priorities, using the following formula: Obtain user bandwidth ,in, for, Prioritize users Prioritizes the largest user.
9. The VR spatial precision positioning method based on head and limb tracking according to claim 1, characterized in that, The method for automatically adjusting the sampling rate is as follows: By sensing user status from multiple dimensions, status parameters are collected and quantified. In obtaining EEG signals The proportion of wave power to total brainwave power is used to obtain EEG fatigue level; The heart rate variability is obtained by collecting data through the PPG sensor of the head-mounted display, calculating the standard deviation of the heart rate interval and comparing it with a preset standard deviation threshold. The average joint angular velocity of the user's limb joints is obtained by IMU measurement; The obtained EEG fatigue level, heart rate variability and mean joint angular velocity are fused and normalized to obtain the user's comprehensive status index. The sampling rate is automatically optimized based on the obtained state comprehensive index. The complement of the state comprehensive index with respect to 1 is calculated and multiplied with the preset compensation sampling frequency to obtain the sampling compensation frequency. The sampling compensation frequency is then summed with the preset sampling frequency to obtain the automatic sampling frequency.
10. The VR spatial precise positioning method based on head and limb tracking according to claim 1, characterized in that, The method for correcting the optimal parameters is as follows: In a scenario where the precise location is known, calculate the error between the final positioning location and the reference position; The difference between the final location and the reference location is calculated, and the L2 norm of the difference result is taken to obtain the positioning error. Analyze the sources of positioning error and correct the parameters. If the positioning error is greater than the preset error threshold and mainly comes from skeleton modeling, then adjust the pose weights of skeleton modeling. If the positioning error is greater than the preset error threshold and mainly comes from error compensation, the base attenuation factor of the dynamic attenuation factor will be adjusted. If the positioning error is less than or equal to the preset error threshold, no parameter correction will be performed.