A method and system for evaluating human balance function adapted to multitasking and complex scenarios thereof

Abstract

The application provides a human balance function evaluation method and system suitable for multi-task and complex scene, comprising: collecting plantar pressure distribution data of a subject in static standing and dynamic walking tasks and six-degree-of-freedom pose data of both feet calculated based on visual-inertial simultaneous localization and mapping technology; reconstructing the overall pressure center trajectory of the whole process of dynamic and static tasks through time synchronization, rigid body transformation and weighted fusion algorithm, realizing continuous and accurate pressure center calculation for the whole domain; extracting multi-dimensional evaluation indexes for dynamic and static tasks respectively, and adaptively calculating dynamic balance parameters in complex scenes such as turning, obstacle avoidance and stair climbing; mapping the multi-dimensional evaluation indexes to multiple physiological function sub-models, taking the health population norm database as the benchmark to calculate the comprehensive performance index of each sub-model, and outputting the balance control ability evaluation result of the subject. The application realizes joint evaluation of dynamic and static and self-adaption in complex scenes, and can accurately locate the mechanism of balance dysfunction.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation and health monitoring technology, and in particular to a method and system for assessing human balance function that is adaptable to multitasking and complex scenarios. Background Technology

[0002] Balance is a fundamental physiological function for maintaining postural stability, and its decline is a leading risk factor for accidental falls in the elderly. Objective and accurate assessment of balance ability is crucial for disease rehabilitation and injury prevention.

[0003] Currently, the main tool for clinical assessment of balance function is the Computerized Dynamic Posturography (CDP) system. This system performs quantitative analysis based on the center of pressure (COP) trajectory, but it relies on a fixed pressure testing platform or pressure walkway and can only perform static standing assessments in a laboratory environment. It cannot effectively measure dynamic activities such as walking, turning, and climbing stairs.

[0004] To overcome the limitations of fixed devices, researchers have proposed wearable evaluation schemes. One approach is based on pressure-monitoring insoles, which can continuously acquire plantar pressure distribution but cannot obtain the relative pose of the legs, only measuring plantar COP, making it difficult to integrate foot pressure into the overall human body COP. A second approach is based on inertial measurement units (IMUs), but IMU integration suffers from cumulative drift, and existing algorithms are mutually exclusive between dynamic and static tasks—zero-rate updates fail in static evaluation, and gradient descent is easily interfered with in dynamic walking. A third approach is a multi-sensor fusion scheme (IMU + pressure insole), which can achieve COP estimation in walking tasks, but still suffers from drift problems in static evaluation and is mostly limited to straight-line walking, failing in complex scenarios such as turning and obstacle avoidance.

[0005] Furthermore, existing virtual reality assessment schemes mostly treat virtual reality (VR) devices as visual feedback tools, lacking precise control over stimulus parameters such as visual flow and spatial audio; functional diagnostic schemes, on the other hand, have single assessment indicators and find it difficult to distinguish the contributions of different subsystems such as vision, vestibular sensation, and proprioception to balance control.

[0006] Therefore, existing balance function assessment technologies suffer from problems such as scenario fragmentation, mutual exclusion of dynamic and static elements, lack of analysis of complex scenarios, inability to quantify multi-sensory stimuli, and limited functional diagnostic dimensions. Summary of the Invention

[0007] In view of this, embodiments of the present invention provide a method and system for assessing human balance function that is adapted to multiple tasks and complex scenarios, in order to solve the problems of existing balance function assessment technologies, such as the separation of equipment and scenarios leading to the inability to cover real-life scenarios, the mutual exclusion of dynamic and static assessment tasks making it difficult for the same system to accurately measure static standing and dynamic walking, the lack of gait analysis capabilities in complex scenarios leading to the inability to quantify real movement characteristics such as turning and obstacle avoidance, the lack of precise control and quantitative analysis in multi-sensory stimulation assessment, and the difficulty in locating the root cause of functional impairment mechanisms due to the single functional diagnostic indicators.

[0008] On one hand, the present invention provides a method for assessing human balance function that is adapted to multitasking and complex scenarios, the method comprising:

[0009] Multi-source heterogeneous data were collected from subjects during a multi-task evaluation process that included static standing tasks and dynamic walking tasks. The multi-source heterogeneous data included: plantar pressure distribution data obtained by plantar pressure sensors installed on both feet, and bipedal six-degree-of-freedom pose data calculated by self-positioning trackers installed on both feet based on visual-inertial real-time localization and mapping technology.

[0010] The plantar pressure distribution data and the bipedal six-degree-of-freedom pose data are synchronized in time to obtain a time-aligned synchronized dataset;

[0011] Based on the synchronous dataset, the plantar pressure distribution data is mapped to a unified spatial coordinate system through rigid body transformation, and the overall pressure center trajectory of the subject during the entire process of static standing and dynamic walking is reconstructed by weighted fusion algorithm, so as to realize continuous and accurate pressure center calculation for the whole domain.

[0012] Based on the synchronous dataset and / or the overall center of pressure trajectory, multidimensional evaluation indicators are extracted for the static standing task and the dynamic walking task, respectively, and feature vector sets are constructed based on the multidimensional evaluation indicators. The multidimensional evaluation indicators include static balance parameters, dynamic balance parameters, spectral features, lower limb joint torque estimates, and multisensory stimulus response features. For the dynamic walking task, the dynamic balance parameters are adaptively calculated in various complex scenarios based on gait event detection and walking direction estimation. These complex scenarios include at least turning, obstacle avoidance, and climbing stairs. The spectral features are obtained by performing spectral analysis on the overall center of pressure trajectory, decomposing it into multiple feature frequency bands corresponding to different physiological regulatory functions, and calculating the power characteristics of each frequency band. The lower limb joint torque estimates are obtained by inverse calculation based on the overall center of pressure trajectory data during the support phase of the gait cycle. The multisensory stimulus response features are obtained by applying controllable sensory stimuli to the subject and quantifying the subject's response under different stimulus conditions, including response lag time and response gain.

[0013] The feature vector set is mapped to multiple preset physiological function sub-models. Based on the norm database of healthy people, the comprehensive efficacy index of each physiological function sub-model is calculated, and the assessment results of the subject's balance control ability in different physiological function dimensions are output.

[0014] In some embodiments of the present invention, multidimensional evaluation indicators are extracted based on the synchronized dataset and / or the overall pressure center trajectory, including:

[0015] The static equilibrium parameters are calculated based on the overall pressure center trajectory reconstructed in the static standing assessment task, including the center of gravity movement speed, the total path length of the pressure center trajectory, the maximum displacement of the pressure center, and the envelope area.

[0016] The dynamic balance parameters are calculated based on the gait events detected and the estimated direction of travel in the walking balance assessment task, including stride length, stride length, stride width, stride height, stride frequency, stride speed, gait cycle, support phase ratio, swing phase ratio, double support phase ratio, gait variability and gait symmetry index.

[0017] The spectral analysis uses wavelet decomposition to decompose the overall pressure center trajectory into multiple characteristic frequency bands, including: a low frequency band of 0-0.3Hz, a mid frequency band of 0.3-1Hz, and a high frequency band of 1-3Hz, which correspond to visual accommodation function, vestibular accommodation function, and proprioceptive accommodation function, respectively.

[0018] The response lag time is calculated by recording the time difference between the stimulus excitation time and the time difference between the acceleration at the center of pressure exceeding the baseline by 3 times the standard deviation threshold; the response gain is obtained by establishing a stimulus-response transfer function model and calculating the ratio of system output to input.

[0019] The estimated lower limb joint torque is obtained by inverse calculation based on the overall pressure center trajectory data of the support phase in the gait cycle. Specifically, it includes: identifying and extracting the overall pressure center trajectory data of the support phase in the gait cycle, constructing a high-dimensional spatiotemporal input vector containing the dynamic features of the current frame and the historical frame information within the preset stride length, introducing a multi-scale sliding window for segmentation and feature extraction, and using a cascaded random forest model to predict the torque of the ankle, knee and hip joints.

[0020] In some embodiments of the present invention, based on the synchronized dataset, the plantar pressure distribution data is mapped to a unified spatial coordinate system through rigid body transformation, and a weighted fusion algorithm is used to reconstruct the overall pressure center trajectory of the subject throughout the static standing and dynamic walking processes, including:

[0021] Based on the bipedal six-DOF pose data in the synchronized dataset, a rigid body transformation is used to map the preset local coordinates of the sensors to the world coordinate system, thereby obtaining the absolute coordinates of each pressure sensor in the world coordinate system. The expression for the rigid body transformation is:

[0022] ;

[0023] in, Indicates the first frame The side foot The absolute coordinates of each sensor in the world coordinate system; This represents the rotation matrix obtained by the rotation quaternion transformation; Represents the local coordinates of the sensor; This represents the fixed structural offset from the center of the tracker to the origin of the local coordinate system of the insole; This indicates the tracker's position coordinates in the world coordinate system;

[0024] By introducing adaptive weighting coefficients and combining the projected sensor absolute coordinates with real-time pressure values, the global pressure center in the world coordinate system is calculated using the following formula:

[0025] ;

[0026] ;

[0027] in, , They represent the first The overall pressure center of the frame is in the world coordinate system shaft and Position coordinates on the axis; This represents the filtered pressure value; Indicates the adaptive weighting coefficient; , These represent the sensor's position in the world coordinate system. shaft and Axis coordinates.

[0028] In some embodiments of the present invention, when a subject performs the dynamic walking task, gait event detection is performed based on the synchronized dataset and the overall center of pressure trajectory, including:

[0029] The total pressure on one foot is calculated using the following expression:

[0030] ;

[0031] in, Indicates the first frame Total pressure on the sole of the side foot; This represents the filtered pressure value;

[0032] A dual-threshold decision mechanism is used to detect gait events by setting a landing threshold and a takeoff threshold.

[0033] When the total pressure on the sole of one foot is greater than or equal to the landing threshold, and the total pressure on the sole of the foot in the previous frame is less than the landing threshold, it is determined to be a landing event. The current time is recorded as the landing time, and the current foot position is recorded as the landing position.

[0034] When the total pressure on the sole of one foot is less than or equal to the ground clearance threshold, and the total pressure on the sole of the foot in the previous frame is greater than the ground clearance threshold, it is determined to be a ground clearance event, and the current time is recorded as the ground clearance time.

[0035] The geometric center points of both feet are calculated using the following expression:

[0036] ;

[0037] in, Indicates the first Geometric center point of both feet in frame world coordinate system; , These represent the position coordinates of the left and right foot trackers in the world coordinate system, respectively.

[0038] A sliding window of preset duration is constructed, and a polynomial fitting is performed on the sequence of geometric center points of the two feet within the window to reconstruct the subject's walking path. The unit walking direction vector is obtained by calculating the normalized tangent vector of the fitted curve at the current moment.

[0039] In some embodiments of the present invention, the physiological function sub-model includes a visual modulation function sub-model, a vestibular modulation function sub-model, a proprioceptive modulation function sub-model, a motor control and executive function sub-model, and an anti-interference and adaptive control function sub-model.

[0040] In some embodiments of the present invention, the calculation of the comprehensive efficacy index of each physiological function sub-model includes:

[0041] Based on the norm database of healthy people, the feature components corresponding to each physiological function sub-model are standardized using the Z-Score standardization algorithm.

[0042] A multi-factor weighted fusion algorithm was used to fuse and calculate the standardized feature vectors to obtain the comprehensive efficacy index of each physiological function sub-model.

[0043] On the other hand, the present invention provides a human balance function assessment system adapted to multi-tasking and complex scenarios, the system comprising:

[0044] The plantar pressure acquisition module is installed on both feet of the subject to collect plantar pressure distribution data during the assessment process.

[0045] The bipedal pose tracking module includes two tracker modules, which are respectively installed on the subject's two feet to acquire bipedal six-DOF pose data;

[0046] A multi-sensory stimulation module for applying controllable sensory stimulation to a subject, the sensory stimulation including at least one of visual stimulation, auditory stimulation and proprioceptive stimulation;

[0047] The data processing module is communicatively connected to the plantar pressure acquisition module, the bipedal posture tracking module, and the multisensory stimulation module, and is configured to perform the steps of the aforementioned method.

[0048] The diagnostic decision module, which is connected in communication with the data processing module, is used to map the feature vector set to multiple preset physiological function sub-models, calculate the comprehensive efficacy index of each physiological function sub-model based on the healthy population norm database, and output the assessment results of the subject's balance control ability in different physiological function dimensions.

[0049] In some embodiments of the present invention, the plantar pressure acquisition module includes a plurality of pressure sensors disposed in the insole, the plurality of pressure sensors being disposed in the forefoot region and the heel region respectively;

[0050] The tracker module is rigidly mounted on the heel area of ​​the subject's shoes and maintains a fixed relative position with the plantar pressure acquisition module. Each tracker has a built-in visual sensor and inertial measurement unit, which calculates the six-degree-of-freedom pose data of the corresponding foot in real time based on visual-inertial real-time positioning and mapping technology.

[0051] In some embodiments of the present invention, the multi-sensory stimulation module includes:

[0052] The visual stimulation submodule includes a virtual reality headset for providing programmed, controlled visual stimuli to interfere with the subject's visual information;

[0053] The auditory stimulation submodule includes a three-dimensional spatial audio system for providing programmed and controlled auditory stimuli to interfere with the subject’s auditory spatial reference information;

[0054] The proprioceptive stimulation submodule is used to provide programmed and controllable proprioceptive stimuli to interfere with the subject's proprioceptive information, said proprioceptive stimuli including at least one of support surface perturbation stimuli and complex terrain stimuli.

[0055] In some embodiments of the present invention, the system further includes a safety protection module, which is communicatively connected to the data processing module and the multi-sensory stimulation module, and is used to receive the dynamic stability margin output by the data processing module, and control the multi-sensory stimulation module to terminate the stimulation output when the dynamic stability margin is lower than a preset safety threshold.

[0056] The present invention has the following beneficial effects:

[0057] This invention achieves unified hardware for both static and dynamic assessments, along with precise calculation of the entire pressure center. In existing technologies, static assessment relies on fixed force platforms, while dynamic assessment relies on pressure tracks or IMUs, resulting in hardware incompatibility and inconsistent data. This invention, through the deep integration of visual-inertial real-time positioning and mapping technology with plantar pressure monitoring, achieves for the first time continuous and precise reconstruction of the overall pressure center trajectory throughout both static standing and dynamic walking processes using a single wearable hardware setup. This completely solves the data fragmentation problem caused by the mutual exclusion of static and dynamic assessments in traditional solutions.

[0058] This invention enables adaptive gait analysis in complex scenarios. Existing wearable solutions are mostly limited to straight-line walking, and the models fail in complex scenarios such as turning, obstacle avoidance, and climbing stairs. This invention achieves adaptive and accurate calculation of dynamic balance parameters such as stride length, stride width, stride height, and stride frequency under complex movement trajectories through dual-threshold gait event detection and real-time walking direction estimation. It can be continuously performed in real-life scenarios, significantly improving the ecological validity of the evaluation results.

[0059] This invention represents a diagnostic upgrade from a single scoring system to one that traces the underlying mechanisms. Existing functional diagnostic schemes rely on single assessment indicators, making it difficult to differentiate the contributions of different subsystems such as vision, vestibular system, and proprioception to balance control. This invention maps multidimensional assessment indicators (static balance parameters, dynamic balance parameters, spectral characteristics, lower limb joint torque estimates, and multisensory stimulus response characteristics) to five physiological functional sub-models: visual accommodation, vestibular accommodation, proprioceptive accommodation, motor control and executive function, and interference resistance and adaptive control. This enables precise localization of the source and severity of balance dysfunction lesions.

[0060] This invention fills a gap in the assessment of multisensory integration abilities. Existing virtual reality assessment schemes mostly use VR as a visual feedback tool, lacking precise control over stimulus parameters such as visual flow and spatial audio. This invention, through programmed and controllable quantification of multisensory stimulation and response characteristics (response lag time, response gain), achieves an objective assessment of subjects' postural adjustment abilities under complex sensory conditions, providing an objective and quantitative basis for precise intervention of balance dysfunction and the development of personalized rehabilitation programs.

[0061] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0062] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0063] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings:

[0064] Figure 1 This is a schematic diagram illustrating the steps of a human balance function assessment method adapted to multitasking and complex scenarios in one embodiment of the present invention.

[0065] Figure 2 This is a flowchart illustrating a human balance function assessment method adapted to multi-tasking and complex scenarios in one embodiment of the present invention.

[0066] Figure 3 This is a schematic diagram of a data acquisition device in one embodiment of the present invention.

[0067] Figure 4 This is a distribution diagram of the foot pressure monitoring insole sensors in one embodiment of the present invention.

[0068] Figure 5 This is a schematic diagram illustrating the correspondence between gait events and step timing in one embodiment of the present invention.

[0069] Figure 6 This is a schematic diagram of the time distribution of each phase within the gait cycle in one embodiment of the present invention.

[0070] Figure 7 This is a schematic diagram of the structure of a human balance function assessment system adapted to multi-tasking and complex scenarios in one embodiment of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0072] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0073] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0074] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0075] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0076] It should be emphasized here that the step markers mentioned below are not a limitation on the order of the steps, but should be understood as meaning that the steps can be executed in the order mentioned in the embodiments, or in a different order than in the embodiments, or several steps can be executed simultaneously.

[0077] To address the problems of existing balance function assessment technologies, such as the disconnect between equipment and scenarios leading to inability to cover real-life situations, the mutual exclusion of static and dynamic assessment tasks making it difficult for the same system to accurately measure both static standing and dynamic walking, the lack of gait analysis capabilities in complex scenarios making it impossible to quantify real-world movement characteristics like turning and obstacle avoidance, the lack of precise control and quantitative analysis in multi-sensory stimulus assessment, and the difficulty in pinpointing the root cause of functional impairment due to the reliance on single functional diagnostic indicators, this invention provides a human balance function assessment method adapted to multi-task and complex scenarios, such as... Figure 1 As shown, the method includes the following steps S101~S105:

[0078] Step S101: Collect multi-source heterogeneous data from the subject during a multi-task evaluation process that includes static standing tasks and dynamic walking tasks. The multi-source heterogeneous data includes: plantar pressure distribution data obtained by plantar pressure sensors installed on both feet, and bipedal six-DOF pose data calculated by self-localization trackers installed on both feet based on visual-inertial real-time localization and mapping technology.

[0079] Step S102: Synchronize the plantar pressure distribution data with the bipedal six-DOF pose data in time to obtain a time-aligned synchronized dataset.

[0080] Step S103: Based on the synchronous dataset, the plantar pressure distribution data is mapped to a unified spatial coordinate system through rigid body transformation, and the overall pressure center trajectory of the subject during the entire process of static standing and dynamic walking is reconstructed by weighted fusion algorithm to achieve continuous and accurate pressure center calculation for the entire domain.

[0081] Step S104: Based on the synchronous dataset and / or the overall center of pressure trajectory, extract multidimensional evaluation indicators for both the static standing task and the dynamic walking task, and construct a feature vector set based on the multidimensional evaluation indicators. The multidimensional evaluation indicators include static balance parameters, dynamic balance parameters, spectral features, lower limb joint torque estimates, and multisensory stimulus response features. For the dynamic walking task, dynamic balance parameters are adaptively calculated in various complex scenarios based on gait event detection and direction estimation. Complex scenarios include at least turning, obstacle avoidance, and climbing stairs. Spectral features are obtained by performing spectral analysis on the overall center of pressure trajectory, decomposing it into multiple feature frequency bands corresponding to different physiological regulatory functions, and calculating the power characteristics of each frequency band. Lower limb joint torque estimates are obtained by inverse calculation based on the overall center of pressure trajectory data during the support phase of the gait cycle. Multisensory stimulus response features are obtained by applying controllable sensory stimuli to the subject and quantifying the subject's response under different stimulus conditions, including response lag time and response gain.

[0082] Step S105: Map the feature vector set to multiple preset physiological function sub-models, calculate the comprehensive efficacy index of each physiological function sub-model based on the healthy population norm database, and output the assessment results of the subject's balance control ability in different physiological function dimensions.

[0083] like Figure 2 The diagram shown is a flowchart of a human balance function assessment method adapted to multi-tasking and complex scenarios.

[0084] Step S101 aims to acquire raw data from the subjects during the balance function assessment process, providing a data foundation for subsequent spatiotemporal fusion and gait analysis. By simultaneously acquiring plantar biomechanical information and foot kinematic information, a comprehensive digital representation of the subjects' balance control process is achieved.

[0085] Before conducting a balance function assessment, subjects need to wear specialized data acquisition equipment, which mainly consists of two parts: a plantar pressure acquisition module and a bipedal posture tracking module.

[0086] The plantar pressure acquisition module is installed inside the subject's shoes to continuously collect plantar pressure distribution data throughout the assessment process. In some embodiments, such as Figure 3 As shown, this module is implemented in the form of a foot pressure monitoring insole. The insole has multiple built-in pressure sensors that collect pressure changes in various areas of the foot at a sampling frequency of approximately 100Hz.

[0087] In some embodiments, such as Figure 4 As shown, the plantar pressure monitoring insole incorporates eight pressure sensors, located in the forefoot and heel areas to cover key pressure-bearing areas of the foot. Specifically, four pressure sensors are positioned in the forefoot area, corresponding to the first, third, and fifth metatarsals and the area below the big toe; four pressure sensors are positioned in the heel area, corresponding to the medial and lateral calcaneus, the medial arch, and the lateral arch. This layout is designed based on the anatomical structure and biomechanical characteristics of the human foot—the first and fifth metatarsals and the heel area are the main pressure-bearing areas, the area below the big toe plays an important role during propulsion, and the arch area regulates the distribution of plantar pressure. By covering these key pressure-bearing areas, the insole can comprehensively capture the characteristics of plantar pressure distribution.

[0088] In some embodiments, the plantar pressure monitoring insole transmits the collected pressure data to the data processing module via wireless communication (such as Bluetooth).

[0089] The bipedal pose tracking module includes two trackers, each mounted on one of the subject's feet, to acquire six-degree-of-freedom pose data of the feet in space. The six-degree-of-freedom pose includes three translational components (i.e., spatial position coordinates) and three rotational components (i.e., spatial attitude angles), which can completely describe the motion state of the feet in three-dimensional space.

[0090] In some embodiments, the tracker is rigidly mounted on the heel area of ​​the subject's shoes and maintains a fixed relative position with the plantar pressure acquisition module. Preferably, the tracker is mounted 40-50 mm (e.g., 45 mm) directly above the origin of the heel coordinates of the corresponding side's plantar pressure monitoring insole, to avoid collisions with the tracker or affecting the subject's normal gait during walking. The rigid structure ensures no relative movement between the tracker and the insole, thereby guaranteeing the accuracy of subsequent spatial coordinate transformations.

[0091] Each tracker incorporates a built-in visual sensor and inertial measurement unit (IMU), using Visual-Inertial Simultaneous Localization and Mapping (VI-SLAM) technology to calculate the six-DOF pose data of the corresponding foot in real time. This technology utilizes a camera to capture static feature points in the environment, deeply fusing them with angular velocity and acceleration measured by the IMU. Inter-frame motion is estimated through front-end visual odometry, while back-end optimizations include bundle adjustment (BA) and global pose map optimization. Loop closure detection is used to eliminate accumulated errors, ultimately outputting high-precision, drift-free six-DOF pose data, including three-dimensional position coordinates and rotation quaternions.

[0092] Compared with traditional pure inertial navigation schemes, VI-SLAM technology uses environmental visual features as an external absolute reference, fundamentally overcoming the cumulative drift problem of IMU integration methods: in static standing evaluation, visual features provide stable pose constraints; in dynamic walking evaluation, visual-inertial fusion can both cope with rapid movement and suppress IMU integration errors.

[0093] The tracker also collects data at a sampling frequency of approximately 100Hz and transmits it wirelessly to the data processing module.

[0094] It should be noted that the plantar pressure monitoring insole and the bipedal self-positioning tracker use different wireless transmission protocols and independent sampling clocks, resulting in time synchronization issues in the data collected by the two. This step only completes the acquisition of raw data; data time synchronization will be specifically handled in subsequent step S102. However, to ensure the feasibility of subsequent time synchronization, this step adds a high-precision timestamp to each frame of data during acquisition, recording the precise moment of data acquisition, providing a basis for cross-correlation delay estimation and unified time reference mapping in step S102.

[0095] In some embodiments, the system further includes a virtual reality headset worn on the subject's head to provide visual stimulation and construct a virtual assessment scenario. The headset's built-in display and spatial audio system can present programmed, controllable visual flows, geometric illusion stimuli, and three-dimensional spatial audio interference, creating a highly immersive and isolated virtual environment for the subject. The headset and tracker belong to the same underlying ecosystem, enabling the audiovisual interference output by the headset during the assessment process to achieve native spatiotemporal alignment with the pose data collected by the tracker at the system level, thereby ensuring high temporal resolution and low synchronization error in subsequent multisensory stimulation assessments.

[0096] Step S102 aims to perform time synchronization processing on the plantar pressure distribution data and the bipedal six-DOF pose data acquired in step S101. This eliminates the time asynchrony problems introduced by different sensors using independent sampling clocks, different wireless transmission protocols, and inconsistent sampling frequencies, resulting in a time-aligned synchronized dataset. This solves the time misalignment problem of heterogeneous data in a multi-sensor system, providing a strictly time-matched synchronized dataset for subsequent spatial fusion (step S103) and gait analysis (step S104). It ensures that each frame of data corresponds to the same physical moment, thereby guaranteeing the accuracy of overall pressure center reconstruction and gait parameter calculation.

[0097] Before formal time synchronization, the raw data needs to be preprocessed to eliminate dimensional differences between different data types and provide standardized input for subsequent cross-correlation analysis. Specifically, before each experiment, subjects wore the device and walked naturally for about 5 seconds. The system collected plantar pressure data and tracker pose data during this process as calibration samples.

[0098] The foot pressure change rate amplitude sequence and vertical acceleration sequence were extracted from the collected data. The foot pressure change rate amplitude was obtained by summing the measurements from eight pressure sensors on one foot, then calculating the absolute value after time difference. The vertical acceleration was taken from the vertical component of the triaxial acceleration data collected by the tracker's built-in IMU.

[0099] In some embodiments, since there may be slight differences in the original sampling frequencies of the insole and the tracker, and the sampling times are not completely uniform, the module uses a linear interpolation algorithm to resample the two sequences to a uniform time grid of 100Hz. Subsequently, the resampled sequences are Z-score normalized to eliminate the influence of dimensions, resulting in a normalized pressure sequence and a normalized acceleration magnitude sequence, which prepares for subsequent cross-correlation analysis.

[0100] After obtaining the standardized sequences, the discrete cross-correlation function is used to calculate the similarity between the two sequences at different time delays, thereby accurately estimating the relative delay between the insole data and the tracker data. The principle of cross-correlation calculation is that during walking, the amplitude of the rate of change of foot pressure and the vertical acceleration both exhibit rhythmic fluctuations synchronized with the gait cycle. When the two signal waveforms are similar but have a time offset, the cross-correlation function will show a peak at the time delay corresponding to the offset.

[0101] In some embodiments, the formula for calculating the discrete cross-correlation function is shown in formula (1):

[0102] ; (1)

[0103] in, Indicates a hysteresis index as The cross-correlation function value at time; Indicates the length of the resampled data sequence; Indicates that the index is The standardized foot pressure change rate amplitude; Indicates that the index is Standardized vertical acceleration.

[0104] The module determines the optimal alignment index by searching for the global peak position of the cross-correlation function, as shown in formula (2):

[0105] ; (2)

[0106] in, This indicates the optimal lag index.

[0107] The relative delay between devices is calculated based on the optimal lag index, as shown in formula (3):

[0108] ; (3)

[0109] in, This indicates the resampling frequency; preferably, the resampling frequency is set to 100Hz.

[0110] The relative delay is calculated based on formula (3). A positive value indicates that the tracker data arrived before the insole data; a negative value indicates that the insole data arrived before the tracker data.

[0111] After obtaining the relative delay between devices, the module corrects the original timestamps of the left and right insoles according to the delay amount, as shown in formulas (4) and (5):

[0112] ; (4)

[0113] ; (5)

[0114] in, , These represent the timestamps of the corrected left and right insole data, respectively. , These represent the original timestamps for the left and right insoles, respectively. , These represent the relative delay between the left and right insoles and the corresponding side trackers, respectively.

[0115] After completing the timestamp correction, a standard uniform time grid with a frequency of 100Hz is constructed. Using this grid as the time reference, a linear interpolation algorithm is used to map the corrected pressure data and tracker pose data to a unified time axis. Finally, a synchronization dataset for subsequent analysis is generated. The information contained in each frame of synchronization data is shown in formula (6):

[0116] ; (6)

[0117] in, Indicates the first The frame synchronization dataset; Represents a timestamp on a unified time grid; , These are 1×8 vectors, representing the pressure values ​​of each sensor in the left and right feet after resampling; , These are 1×3 vectors, representing the three-dimensional position coordinates of the left and right foot trackers in the world coordinate system; , These are 1×4 vectors, representing the rotation quaternions of the left and right foot trackers in the world coordinate system.

[0118] It should be noted that the time synchronization process described above is completed through initialization calibration before the start of each evaluation experiment. The obtained relative delay parameters can be directly applied to the time alignment of the data stream in subsequent formal evaluations. If the experimental environment changes or the equipment is restarted, calibration needs to be performed again to update the delay parameters.

[0119] Step S103 aims to map the two-dimensional plantar pressure distribution data from the local sensor coordinate system to a unified three-dimensional world coordinate system using rigid body transformation based on the time-synchronized dataset obtained in step S102. A weighted fusion algorithm is then used to reconstruct the subject's overall center of pressure (COP) trajectory throughout the assessment process. This solves the problem in existing technologies where the separation of pressure information from both feet prevents the synthesis of the overall human COP, achieving natural fusion of pressure information from both feet in a unified spatial coordinate system. This provides core data support for subsequent balance function assessment and gait analysis. By accurately mapping the spatial positions of each pressure point on the sole to the world coordinate system, the physically meaningful overall COP trajectory can be calculated in real time and continuously, regardless of whether the subject is standing still, walking in a straight line, turning, or climbing stairs in complex scenarios.

[0120] To eliminate the influence of sensor noise and small fluctuations in plantar pressure on the calculation results, the pressure data in the synchronous dataset is first smoothed.

[0121] Specifically, a low-pass filter is used to filter the pressure sequences of the left and right foot sensors. In some embodiments of the invention, a fourth-order low-pass Butterworth filter is used. The filtered pressure value is denoted as... subscript Indicates the left foot or the right foot; Indicates the sensor number, Indicates the frame index.

[0122] After signal smoothing, the module uses bipedal six-DOF pose data from the synchronous dataset to map the preset sensor local coordinates to a unified world coordinate system through rigid body transformation, thereby obtaining the absolute coordinates of each pressure sensor in the world coordinate system.

[0123] First, a local sensor coordinate system needs to be established with the heel of the insole as the origin. In this local coordinate system, each pressure sensor has a fixed two-dimensional coordinate. Since the thickness of the insole is negligible, the vertical coordinate is set to zero.

[0124] In some embodiments, the first is defined The local coordinates of each sensor are shown in formula (7):

[0125] ; (7)

[0126] in, Indicates the first Local coordinates of each sensor; This indicates the distance of the sensor from the origin at the heel along the length of the insole; This indicates the distance the sensor deviates from the central axis along the width of the insole. These coordinate values ​​are fixed constants and do not change over time. This indicates transpose.

[0127] Since the plantar pressure monitoring insole and the foot tracker are rigidly connected and do not rotate relative to each other (fixed by a rigid support), the sensor's position in space is entirely determined by the tracker's pose. Therefore, using the tracker's pose information in the synchronized dataset, the sensor's local coordinates are projected onto the world coordinate system through rigid body transformation. The expression for rigid body transformation is shown in equation (8):

[0128] ; (8)

[0129] in, Indicates the first frame The side foot The absolute coordinates of each sensor in the world coordinate system; This represents the rotation matrix obtained by the rotation quaternion transformation; Represents the local coordinates of the sensor; This represents the fixed structural offset from the center of the tracker to the origin of the local coordinate system of the insole. This offset is obtained through calibration during device installation and is used as a constant in subsequent calculations. This indicates the position coordinates of the tracker in the world coordinate system, that is, the position of the tracker's center in the world coordinate system.

[0130] Through the rigid body transformation described above, the sensor position point, which was originally fixed in the local coordinate system, is accurately mapped to the world coordinate system, thus obtaining its spatial position coordinates that change over time. This process elevates the two-dimensional plantar pressure distribution to a three-dimensional spatial dimension, enabling the pressure information from both feet to be processed uniformly within the same world coordinate system.

[0131] After obtaining the absolute coordinates of each sensor in the world coordinate system, an adaptive weighting coefficient is introduced, and combined with the real-time pressure value, the overall pressure center in the world coordinate system is calculated by a weighted average algorithm.

[0132] It should be noted that, due to the nonlinear response error (i.e., the output signal is not strictly linearly related to the applied pressure) and the sensor edge effect (the sensitivity of the sensor edge region differs from that of the center region), directly using the original pressure value for weighted averaging may lead to biased calculation results. To compensate for these errors, this invention introduces adaptive weighting coefficients to correct the measurement errors of each sensor.

[0133] In some embodiments of the present invention, the adaptive weighting coefficients are obtained through a pre-trained machine learning model. Specifically, during system deployment, each sensor is calibrated using a standard test with a known pressure center (e.g., using a force table as the true value), and sensor response data under different pressure distributions are collected. With the goal of minimizing the error between the weighted calculation result and the true pressure center, weighting coefficients are learned that make the weighted calculation result closest to the true value. Once calibrated, the weighting coefficients are used as fixed parameters in subsequent practical applications.

[0134] Combining the projected sensor spatial coordinates with the real-time pressure value, the formulas for calculating the position coordinates of the overall pressure center on the x-axis and y-axis of the world coordinate system are shown in formulas (9) and (10):

[0135] ; (9)

[0136] ; (10)

[0137] in, , They represent the first The overall pressure center of the frame is in the world coordinate system shaft and Position coordinates on the axis; This represents the filtered pressure value; Indicates the adaptive weighting coefficient; , These represent the sensor's position in the world coordinate system. shaft and Axis coordinates.

[0138] The above calculation process is repeated for each frame of data to generate a continuous overall stress center trajectory for the subject throughout the assessment process. This trajectory is a two-dimensional time series that fully records the changes in the stress center over time during the assessment.

[0139] Step S104 aims to extract multidimensional assessment indicators based on the synchronous dataset obtained in step S102 and / or the overall stress center trajectory reconstructed in step S103. These indicators characterize the subject's balance control ability from different dimensions. A feature vector set is constructed based on the extracted multidimensional assessment indicators to provide input for subsequent balance function diagnosis (step S105).

[0140] Before extracting multidimensional assessment indicators, the subjects' assessment tasks need to be organized according to the assessment paradigms defined in the functional assessment module. The functional assessment module aims to achieve a comprehensive assessment of human balance function through a task-driven approach, and is divided into 36 assessment paradigms based on task type and sensory stimulus type.

[0141] The task types include standing balance assessment units and walking balance assessment units.

[0142] The standing balance assessment unit is used to evaluate the human body's balance ability in a relatively static state, involving both static and dynamic balance abilities. Assessment indicators include balance score, sensory score, COP envelope area, path length, average swing speed, maximum displacement, and standing dynamic stability margin.

[0143] The walking balance assessment unit is used to evaluate the human body's balance ability during continuous movement, involving self-dynamic balance and other-dynamic balance ability. Assessment indicators include stride length, stride width, stride speed, gait frequency, gait cycle distribution, gait symmetry, path length, and dynamic stability margin.

[0144] Sensory stimulation types include visual stimulation, auditory stimulation, and proprioceptive stimulation.

[0145] Visual stimuli include:

[0146] Normal visual feedback: providing a clear and stable interactive scene in the virtual environment, or displaying the real scene through the mixed reality perspective mode of the VR headset;

[0147] No visual feedback: Visual input is blocked by turning off the VR headset screen;

[0148] Visual interference: Using a VR engine to provide programmed and controllable visual flow or geometric illusion stimulation to create "visual-vestibular" conflict.

[0149] Auditory stimuli include:

[0150] Normal auditory feedback: Play natural background ambient sounds, or provide auditory cues based on the "COP-acoustic signal" model to guide the subject to adjust their center of gravity;

[0151] No auditory feedback: Apply white noise to block out ambient sounds, or mute the system.

[0152] Auditory interference: Applying sudden high-decibel noise or non-rhythmic sounds, or adjusting control parameters based on the "COP-acoustic signal" model to apply spatial orientation conflict, time delay interference, or intensity disturbance interference.

[0153] Proprioceptive stimuli include:

[0154] Normal support surface: Provides a solid, level, rigid support plane in static standing scenarios; provides a flat, unobstructed walking path in dynamic walking scenarios;

[0155] Support surface interference: In static standing scenarios, sponge pads are laid to weaken the subject's proprioception, or active and controllable physical disturbances are applied through dynamic balance boards; in dynamic walking scenarios, complex terrain elements such as ramps, irregular paths, and obstacles are introduced.

[0156] By combining the above task types and sensory stimulation types, the functional assessment module can generate 36 assessment paradigms.

[0157] Multidimensional assessment indicators include static balance parameters, dynamic balance parameters, spectral characteristics, lower limb joint torque estimates, and multisensory stimulus response characteristics.

[0158] Static balance parameters were calculated based on the reconstructed overall center of pressure trajectory from the static standing assessment task and were used to assess the subject’s balance control ability in a relatively static state.

[0159] In some embodiments, static equilibrium parameters include the velocity of the center of gravity movement, the total path length of the pressure center trajectory, the maximum displacement of the pressure center, and the envelope area.

[0160] Dynamic balance parameters are calculated based on gait events detected and estimated walking direction during the walking balance assessment task, and are used to assess the subject's balance control ability during continuous movement.

[0161] When subjects perform a walking balance assessment task, gait event detection and direction estimation are first performed based on a synchronous dataset and the overall center of pressure trajectory, providing a time reference and spatial reference system for the subsequent calculation of dynamic balance parameters.

[0162] Gait event detection aims to accurately identify key time points in the gait cycle, namely foot strike and take-off events. This invention employs a dual-threshold determination mechanism based on total plantar pressure to improve the robustness and accuracy of gait event detection.

[0163] First, the total plantar pressure on one side is calculated based on the filtered pressure data, as shown in formula (11):

[0164] ; (11)

[0165] in, Indicates the first frame Total pressure on the sole of the side foot; This represents the filtered pressure value.

[0166] To eliminate spurious phase switching caused by sensor noise or minute pressure fluctuations, this invention sets two different thresholds: a grounding threshold and a grounding threshold. and ground threshold And usually, the grounding threshold is greater than the grounding threshold, that is, a hysteresis comparison mechanism is used.

[0167] In some embodiments, the logic for determining gait events is as follows:

[0168] When the total pressure on one foot is greater than or equal to the set landing threshold, and the total pressure on the foot in the previous frame is less than the landing threshold, it is determined to be a landing event. The current time is recorded as the landing time, and the current foot position is recorded as the landing position.

[0169] When the total pressure on one foot is less than or equal to the set ground-lift threshold, and the total pressure on the foot in the previous frame is greater than the ground-lift threshold, it is determined to be a ground-lift event, and the current time is recorded as the ground-lift time.

[0170] Through the aforementioned dual-threshold determination mechanism, the module can accurately capture the landing and takeoff times of each step, effectively avoiding misjudgments caused by pressure fluctuations, even in complex scenarios such as uneven ground or going up and down stairs.

[0171] The purpose of movement direction estimation is to establish a real-time motion reference coordinate system for the subject, providing a directional benchmark for subsequent calculations of spatial parameters such as stride length and stride width. In straight-line walking, the movement direction is relatively fixed; however, in complex scenarios such as turning and obstacle avoidance, the movement direction changes continuously over time, thus requiring real-time estimation.

[0172] First, the geometric center points of both feet are calculated as a reference for the subject's overall position. The formula for calculating the geometric center point of the two feet in the frame world coordinate system is shown in formula (12):

[0173] ;(12)

[0174] in, Indicates the first Geometric center point of both feet in frame world coordinate system; , These represent the position coordinates of the left and right foot trackers in the world coordinate system, respectively.

[0175] Human walking paths exhibit significant continuity and predictability, making them suitable for fitting using kinematic laws. Based on this characteristic, the module employs a trajectory smoothing tangent method to estimate the direction of movement in real time. Specifically, a sliding window of preset duration (e.g., 2 seconds) is constructed, storing a sequence of historical geometric center points of both feet within the window. The least squares method is used to perform a polynomial fitting on this sequence, reconstructing the subject's smooth walking path. By calculating and normalizing the first derivative of the fitted curve at the current moment, a unit walking direction vector is obtained, pointing to the subject's current instantaneous direction of movement.

[0176] To further establish a complete motion reference coordinate system, a perpendicular unit vector is introduced. Calculate the unit lateral vector as shown in formula (13):

[0177] ; (13)

[0178] in, Represents the unit lateral vector; Represents the unit direction of travel vector; This represents a vertical unit vector.

[0179] The lateral vector is perpendicular to the direction of travel and lies in the horizontal plane, pointing to the subject's left side. Thus, a two-dimensional motion reference coordinate system consisting of the direction of travel D and the lateral direction I is established, providing a benchmark for the spatial decomposition of gait parameters.

[0180] like Figure 5 The diagram shown illustrates the correspondence between gait events and step timing.

[0181] Based on the detected gait events (landing time, takeoff time, and landing position) and the estimated gait direction vector, dynamic balance parameters are calculated. These dynamic balance parameters include stride length, stride width, stride height, stride frequency, stride speed, gait cycle, support phase ratio, swing phase ratio, double support phase ratio, gait variability, and gait symmetry index.

[0182] In some embodiments, for ease of description, the first is defined as Relevant variables for the next step: landing time The landing location is Time off the ground is The vector difference between the landing points of the front and rear feet is shown in formula (14):

[0183] ;(14)

[0184] Stride length refers to the projected distance in the direction of travel from the landing position of the rear foot to the landing position of the front foot during a single step, reflecting the magnitude of movement in one step. The formula for calculating stride length is shown in formula (15):

[0185] ; (15)

[0186] in, Indicates the first The stride length of each step; This represents the unit direction of travel vector.

[0187] Stride length, also known as stride length, refers to the distance traveled between two consecutive foot strikes on the same side, that is, the total distance traveled by taking one step with each foot. The formula for calculating stride length is shown in formula (16):

[0188] ; (16)

[0189] in, Indicates the first The stride length corresponding to each step; and They represent the first Second and third The length of each step.

[0190] Stride width refers to the lateral distance between the left and right feet when they land, perpendicular to the direction of travel. It is usually taken as an absolute value and reflects the base width during walking. The formula for calculating stride width is shown in formula (17):

[0191] ; (17)

[0192] in, Indicates the first The stride width of each step; Indicates the first The vector difference in the landing points of the front and back feet during the next step; This represents the unit lateral vector.

[0193] Step height refers to the maximum vertical height that the swinging foot lifts off the ground during the swinging motion. It reflects the foot's ground clearance and is particularly important in scenarios such as going up and down stairs and crossing obstacles. The formula for calculating step height is shown in formula (18):

[0194] ; (18)

[0195] in, Indicates the first The stride height of the next step; , They represent the first sequence The landing time and takeoff time of each step; Indicates the first Second landing position; Indicates the front foot is The position at that moment; This represents a vertical unit vector.

[0196] In some embodiments, for scenarios with height changes such as going up and down stairs, the step height can also be calculated, that is, the vertical height difference between the front and back foot landing positions, as shown in formula (19):

[0197] ; (19)

[0198] in, This indicates the vertical height difference between the front and back feet when they land.

[0199] Temporal parameters describe the temporal characteristics of gait, mainly including gait frequency, gait speed, gait period, and the proportion of each temporal phase.

[0200] Cadence refers to the number of steps taken per minute, reflecting the rhythm of gait. The formula for calculating cadence is shown in formula (20):

[0201] ; (20)

[0202] in, Indicates the first The instantaneous cadence corresponding to each step, measured in steps per minute.

[0203] Walking speed refers to the average speed of walking, reflecting the overall pace of movement. The formula for calculating walking speed is shown in formula (21):

[0204] ; (twenty one)

[0205] in, Indicates the first The instantaneous pace corresponding to each step; Indicates the first The stride length corresponding to each step; This indicates the time required to complete one full gait cycle.

[0206] Gait cycle refers to the time elapsed from the heel strike of one foot to the heel strike of the same foot again. The formula for calculating gait cycle is shown in formula (22):

[0207] ; (twenty two)

[0208] in, Indicates the first The duration of the current gait cycle ending with the next step.

[0209] like Figure 6 The diagram shown illustrates the time distribution of each phase within the gait cycle.

[0210] The gait cycle can be further divided into different phases, mainly including the support phase, swing phase, and double support phase. The support phase is the period when the foot contacts the ground and bears the body weight; the swing phase is the period when the foot leaves the ground and swings forward to prepare for the next contact; the double support phase is the period when both feet contact the ground simultaneously, usually occurring at the beginning (after heel strike) and end (before toe lift) of the support phase, and is the most stable phase during walking. The duration and proportion of each phase are important indicators for assessing gait stability and balance function. The formulas for calculating the proportion of each phase are as follows:

[0211] The formula for calculating the proportion of support is shown in formula (23):

[0212] ; (twenty three)

[0213] The formula for calculating the proportion of the swing phase is shown in formula (24):

[0214] ; (twenty four)

[0215] The formula for calculating the ratio of the two supports is shown in formula (25):

[0216] (25)

[0217] The physical quantities in formulas (23), (24), and (25) have been explained in the previous text, so they will not be repeated here.

[0218] Gait variability is used to reflect the degree of fluctuation of gait parameters and is an important indicator for evaluating gait stability. It is quantified by the coefficient of variation, and the calculation formula is shown in formula (26):

[0219] ; (26)

[0220] in, Indicates the selection of gait parameters coefficient of variation; Indicates parameters Standard deviation during the testing process; Indicates parameters The mean value during the testing process. The smaller the coefficient of variation, the more stable the gait.

[0221] The gait symmetry index is used to reflect the degree of symmetry of the motor function of the left and right limbs. It is quantified by the symmetry index and the calculation formula is shown in formula (27):

[0222] ;(27)

[0223] in, Indicates the selection of gait parameters The symmetry index; , Representing the left and right parameters respectively The average value during the testing process. The closer the symmetry index is to 0, the better the bilateral symmetry.

[0224] Multisensory stimulus response characteristics are obtained by applying controlled sensory stimuli to subjects and quantifying their responses under different stimulus conditions. This is used to assess subjects' postural regulation abilities under complex sensory conditions. Specifically, visual stimuli (such as visual flow and geometric illusions) and auditory stimuli (such as sound orientation shifts and frequency / volume perturbations) are provided through virtual reality headsets and spatial audio systems, and the subjects' response characteristics to these stimuli are quantified.

[0225] Response lag time is an important indicator for measuring how quickly a subject responds to a stimulus. It is calculated by recording the stimulus activation time and monitoring changes in the center of pressure acceleration. The time when the acceleration exceeds three times the baseline standard deviation threshold is recorded as the response onset time; the difference between the two is the response lag time. A shorter response lag time indicates a faster sensorimotor integration speed in the subject.

[0226] Response gain reflects the subject's sensitivity to stimuli. It is calculated as the ratio of system output (such as maximum shift of the center of pressure, weighted average oscillation velocity, etc.) to input (such as stimulus amplitude, frequency, etc.) by establishing a stimulus-response transfer function model. Excessively high or low gain may indicate sensory integration dysfunction.

[0227] The spectral characteristics were obtained by performing spectral analysis on the overall pressure center trajectory and were used to quantify the contribution of different sensory subsystems to balance control.

[0228] In some embodiments, the pressure center signal is decomposed into three characteristic frequency bands using the Symlet-8 wavelet basis function: low frequency (0-0.3Hz), mid frequency (0.3-1Hz), and high frequency (1-3Hz). According to research, these three frequency bands correspond to different sensory modulation mechanisms: the low frequency band mainly reflects the modulation function of the visual system, where visual information processing is relatively slow, primarily affecting low-frequency posture fluctuations; the mid frequency band corresponds to the modulation of the vestibular sensory system, which plays a major compensating role for mid-frequency posture disturbances; and the high frequency band characterizes the modulation of the proprioceptive system, which is most sensitive to rapid posture fluctuations. By calculating the normalized power of each frequency band, the contribution of different sensory subsystems to balance control can be quantified.

[0229] The estimated lower limb joint torque values ​​are obtained by inverse calculation based on the overall pressure center trajectory data during the support phase of the gait cycle, and are used to evaluate the mechanical output capability of the motion execution system in balance control.

[0230] In some embodiments, a dynamic inverse estimation algorithm based on a deep forest regression model is used to achieve real-time estimation of lower limb joint torques. This process mainly includes the following steps:

[0231] First, based on gait event detection results, the overall pressure center trajectory data during the support phase of the gait cycle is identified and extracted. A zero-phase fourth-order Butterworth low-pass filter is used to smooth the pressure center sequence and eliminate high-frequency noise interference.

[0232] Secondly, a high-dimensional spatiotemporal input vector is constructed. This vector not only contains the dynamic features of the current frame (such as the position of the pressure center, movement speed, acceleration, etc.), but also contains the temporal information of historical frames within a preset step size, so as to fully capture the dynamic characteristics of the motion.

[0233] Then, a multi-scale sliding window is introduced to segment and extract features from the input data, mining motion pattern information from different time scales.

[0234] Finally, the extracted features are input into a cascaded random forest regression model. The prediction results are optimized layer by layer through a multi-layer cascaded structure, and the torque estimates of the ankle, knee and hip joints are finally output.

[0235] Lower limb joint torques reflect the mechanical output capacity of the motor execution system in balance control. By combining them with sensory system assessment results, it is possible to effectively distinguish whether balance dysfunction stems from abnormal sensory input or motor execution deficits, providing a basis for precise intervention decisions.

[0236] After extracting the above five types of multidimensional evaluation indicators, the module constructs a feature vector set based on these indicators to provide input for subsequent balance function diagnosis.

[0237] Step S105 aims to map the feature vector set constructed in step S104 to multiple preset physiological function sub-models, calculate the comprehensive efficacy index of each physiological function sub-model based on the healthy population norm database, and finally output the assessment results of the subject's balance control ability in different physiological function dimensions.

[0238] The physiological function sub-model is an assessment framework built on the physiological mechanism of human balance control. Each sub-model corresponds to a specific physiological function dimension and is used to assess the balance control ability under that dimension.

[0239] In some embodiments of the present invention, the preset physiological function sub-model includes:

[0240] Visual accommodation function sub-model: Evaluating the contribution of the visual system to balance control. The visual system provides spatial orientation reference for posture control by perceiving visual reference information in the environment (such as horizontal lines, vertical lines, and the movement of surrounding objects). Balance control capability typically decreases when visual input is blocked or visual information is inaccurate.

[0241] Vestibular Modulation Function Sub-model: Assessing the contribution of the vestibular system to balance control. The vestibular system senses head movements and the head's direction relative to gravity, providing information about the body's movement and posture in space. Impaired vestibular function can lead to dizziness, vertigo, and balance disorders.

[0242] Proprioceptive Modulation Sub-model: Evaluating the contribution of the proprioceptive system to balance control. The proprioceptive system senses the position and motion of various body parts through receptors in muscles, tendons, and joints, providing information about the relative positions of body parts for postural control. Balance control ability significantly decreases when proprioceptive input is reduced.

[0243] Motor control and executive function sub-model: This sub-model evaluates the mechanical output capability of the motor executive system in balance control. It focuses on the torque output, joint coordination, and muscle activation patterns of lower limb joints during balance maintenance, reflecting the effectiveness of central nervous system commands.

[0244] The anti-interference and adaptive control function sub-model assesses subjects' postural regulation ability in the face of external disturbances or multisensory conflicts. This sub-model focuses on the subject's response speed, response amplitude, and ability to regain stability in response to sudden disturbances, reflecting the sensory integration and motor adaptation capabilities of the central nervous system.

[0245] To map the feature vector set constructed in step S104 to the five physiological function sub-models mentioned above, it is necessary to establish the correspondence between the feature components and the sub-models.

[0246] The assessment of visual accommodation function is achieved by comparing performance differences under different visual conditions.

[0247] For example, in visual assessment: compare the temporal feature vectors (such as center of gravity movement speed, COP envelope area) and normalized power changes in the low-frequency band (0-0.3Hz) under normal visual feedback conditions (a clear and stable virtual scene presented through a VR headset) and conditions without visual feedback conditions (a black screen on the VR headset blocking visual input). If the temporal indicators deteriorate significantly and the low-frequency power increases significantly under the closed-eye condition, it indicates that the subject has a strong dependence on visual information.

[0248] Another example is visual dependence: comparing the temporal feature vectors and low-frequency normalized power changes under visual interference conditions (providing visual flow or geometric illusion stimulation through a VR engine to create "visual-vestibular" conflict) and no visual feedback conditions. If the performance under visual interference conditions is significantly worse than that under no visual conditions, it indicates that the subjects are easily misled by inaccurate visual information and have excessive visual dependence.

[0249] The vestibular adjustment function is assessed by comparing the differences in performance with and without support surface interference:

[0250] The temporal characteristic vector and mid-frequency (0.3-1Hz) normalized power changes were compared under fixed support surface conditions and support surface disturbance conditions (active disturbance applied by a dynamic balancing plate or reduction of plantar sensation by laying a sponge pad). Individuals with normal vestibular function maintained relative stability under support surface disturbance conditions, with a moderate increase in mid-frequency power reflecting vestibular compensation activity; significant deterioration of temporal indicators or abnormal increases in mid-frequency power suggested vestibular dysfunction.

[0251] The assessment of proprioceptive accommodation function is achieved by comparing the differences in performance under conditions with and without support surface interference:

[0252] Compare the temporal feature vector and normalized power changes in the high-frequency band (1-3Hz) under fixed support surface conditions and support surface interference conditions (standing on a sponge mat, reducing the precise perception of plantar mechanoreceptors). Individuals with normal proprioception can maintain stability under support surface interference through visual and vestibular compensation; insufficient increase in high-frequency power suggests impaired proprioceptive function.

[0253] The assessment of motor control and executive function is based on the amplitude and contribution ratio of lower limb joint torques extracted under different paradigms. By analyzing the change patterns of joint torques under different assessment conditions, the adaptability of motor execution is evaluated.

[0254] The evaluation of anti-interference and adaptive control functions was achieved by comparing the performance differences under conditions with and without auditory interference and the response characteristics to multi-sensory stimuli.

[0255] After completing the mapping of feature vectors to physiological function sub-models, the module calculates the comprehensive efficacy index of each physiological function sub-model based on the healthy population norm database.

[0256] The healthy population norm database contains a large-scale baseline data of balanced function of healthy populations, which is classified in multiple dimensions according to age group, gender and physical indicators, and stores the standardized parameter range of healthy populations under different assessment tasks.

[0257] In some embodiments of the present invention, the norm database contains the following information:

[0258] Demographic information: age, sex, height, weight, clinical diagnostic background, etc.

[0259] Assessment task types: static standing, walking, multi-sensory stimulation, etc.;

[0260] Parameter distribution under each assessment task: mean, standard deviation, etc. of each assessment indicator.

[0261] For each feature component corresponding to a physiological function sub-model, a benchmark dataset matching the subject's basic information (age, gender, etc.) is first retrieved from the norm database. Then, the Z-Score normalization algorithm is used to map each feature component into a dimensionless standard deviation vector.

[0262] A multi-factor weighted fusion algorithm is used to fuse and calculate multiple feature components under the same physiological function sub-model to obtain a comprehensive efficiency index that represents the overall functional level of the sub-model.

[0263] In some embodiments, after obtaining the comprehensive efficacy index, the functional status of the subjects can be abnormally determined and graded according to preset thresholds.

[0264] After calculating the comprehensive efficacy index and identifying abnormalities in each physiological function sub-model, the results are output in a structured manner to form an assessment report of the subject's balance control ability in different physiological function dimensions, which can be used to analyze the subject's functional abnormalities and the degree of validation.

[0265] Corresponding to the above method, the present invention also provides a human balance function assessment system adapted to multi-tasking and complex scenarios, such as... Figure 7 As shown, the system includes:

[0266] The multisensory stimulation module 1 is used to apply controllable sensory stimulation to the subject, including at least one of visual stimulation, auditory stimulation and proprioceptive stimulation.

[0267] Data acquisition module 3 includes a plantar pressure acquisition module and a bipedal pose tracking module. The plantar pressure acquisition module is installed on both feet of the subject to collect plantar pressure distribution data during the assessment process. The bipedal pose tracking module includes two tracker modules, each installed on one of the subject's feet, to acquire six-DOF pose data of both feet.

[0268] The data processing module 4 is communicatively connected to the plantar pressure acquisition module, the bipedal posture tracking module, and the multi-sensory stimulation module, respectively. The data processing module is configured to execute the steps of the aforementioned human balance function assessment method adapted to multiple tasks and complex scenarios.

[0269] The diagnostic decision module 6 communicates with the data processing module and is used to map the feature vector set to multiple preset physiological function sub-models. Based on the healthy population norm database, it calculates the comprehensive efficacy index of each physiological function sub-model and outputs the subject's balance control ability assessment results in different physiological function dimensions.

[0270] In some embodiments, such as Figure 3 As shown, the plantar pressure acquisition module includes multiple pressure sensors installed in the insole, with the multiple pressure sensors respectively located in the forefoot area and the heel area.

[0271] The tracker module is rigidly mounted on the heel area of ​​the subject's shoes and maintains a fixed relative position with the plantar pressure acquisition module. Each tracker has a built-in visual sensor and inertial measurement unit, which calculates the six-degree-of-freedom pose data of the corresponding foot in real time based on visual-inertial real-time positioning and mapping technology.

[0272] In some embodiments, the multisensory stimulation module further includes:

[0273] The visual stimulation submodule 11 includes a virtual reality headset for providing programmed and controlled visual stimulation to assess the subject’s visual accommodation function.

[0274] Auditory stimulation submodule 12 includes a three-dimensional spatial audio system for providing programmed and controlled auditory stimulation to assess the subject’s anti-interference and adaptive control functions.

[0275] The proprioceptive stimulation submodule 13 is used to provide programmed and controllable proprioceptive stimulation to assess the subject's proprioceptive modulation function, including at least one of support surface perturbation stimulation and complex terrain stimulation.

[0276] In some embodiments, the system further includes a functional assessment module 2, which is used to organize the subject's assessment task according to the assessment paradigm defined by the functional assessment module before extracting multidimensional assessment indicators.

[0277] In some embodiments, the system further includes an information storage module 5. The information storage module receives structured datasets from the data processing module to construct a complete and traceable individual rehabilitation database, and consists of the following three units:

[0278] Subject Information Management Unit: Used to establish and maintain basic demographic profiles of subjects, including name, age, gender, height, weight, and clinical diagnostic background. Provides basic index tags for subsequent personalized diagnosis and norm matching.

[0279] Multidimensional data archiving unit: It categorizes and stores structured datasets from the data processing module, ensuring longitudinal consistency and traceability of test data from subjects at different rehabilitation stages. At the same time, it transforms scattered assessment data of the same subject into a continuous rehabilitation evolution curve, enabling intuitive historical comparison.

[0280] Healthy Population Norm Database: This database includes a large-scale baseline database of balanced functional abilities in healthy individuals. It is categorized in multiple dimensions by age group, gender, and physical indicators, storing standardized parameter ranges for healthy individuals under different assessment tasks.

[0281] In some embodiments, the system further includes a safety protection module 7. This safety protection module is communicatively connected to the data processing module and the multi-sensory stimulation module, and is used to receive the dynamic stability margin output by the data processing module. When the dynamic stability margin is lower than a preset safety threshold, the system controls the multi-sensory stimulation module to terminate the stimulation output.

[0282] Corresponding to the above method, the present invention also provides an electronic device including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor executing the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the electronic device performs the steps of the method as described above.

[0283] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned method. The computer-readable storage medium may be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0284] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0285] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0286] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0287] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for evaluating human balance function adapted to multitasking and its complex scene, characterized in that, The method includes: Multi-source heterogeneous data were collected from subjects during a multi-task evaluation process that included static standing tasks and dynamic walking tasks. The multi-source heterogeneous data included: plantar pressure distribution data obtained by plantar pressure sensors installed on both feet, and bipedal six-degree-of-freedom pose data calculated by self-positioning trackers installed on both feet based on visual-inertial real-time localization and mapping technology. The plantar pressure distribution data and the bipedal six-degree-of-freedom pose data are synchronized in time to obtain a time-aligned synchronized dataset; Based on the synchronous dataset, the plantar pressure distribution data is mapped to a unified spatial coordinate system through rigid body transformation, and the overall pressure center trajectory of the subject in the static standing task and the dynamic walking task is reconstructed by weighted fusion algorithm, so as to realize continuous and accurate pressure center calculation for the whole domain. Based on the synchronous dataset and / or the overall center of pressure trajectory, multidimensional evaluation indicators are extracted for the static standing task and the dynamic walking task, respectively, and feature vector sets are constructed based on the multidimensional evaluation indicators. The multidimensional evaluation indicators include static balance parameters, dynamic balance parameters, spectral features, lower limb joint torque estimates, and multisensory stimulus response features. For the dynamic walking task, the dynamic balance parameters are adaptively calculated in various complex scenarios based on gait event detection and walking direction estimation. These complex scenarios include at least turning, obstacle avoidance, and climbing stairs. The spectral features are obtained by performing spectral analysis on the overall center of pressure trajectory, decomposing it into multiple feature frequency bands corresponding to different physiological regulatory functions, and calculating the power characteristics of each frequency band. The lower limb joint torque estimates are obtained by inverse calculation based on the overall center of pressure trajectory data during the support phase of the gait cycle. The multisensory stimulus response features are obtained by applying controllable sensory stimuli to the subject and quantifying the subject's response under different stimulus conditions, including response lag time and response gain. The feature vector set is mapped to multiple preset physiological function sub-models. Based on the norm database of healthy people, the comprehensive efficacy index of each physiological function sub-model is calculated, and the assessment results of the subject's balance control ability in different physiological function dimensions are output.

2. The method for assessing human balance function adapted to multi-tasking and complex scenarios according to claim 1, characterized in that, Based on the synchronized dataset and / or the overall stress center trajectory, multidimensional evaluation metrics are extracted, including: The static equilibrium parameters are calculated based on the overall pressure center trajectory reconstructed in the static standing assessment task, including the center of gravity movement speed, the total path length of the pressure center trajectory, the maximum displacement of the pressure center, and the envelope area. The dynamic balance parameters are calculated based on the gait events detected and the estimated direction of travel in the walking balance assessment task, including stride length, stride length, stride width, stride height, stride frequency, stride speed, gait cycle, support phase ratio, swing phase ratio, double support phase ratio, gait variability and gait symmetry index. The spectral analysis uses wavelet decomposition to decompose the overall pressure center trajectory into multiple characteristic frequency bands, including: a low frequency band of 0-0.3Hz, a mid frequency band of 0.3-1Hz, and a high frequency band of 1-3Hz, which correspond to visual accommodation function, vestibular accommodation function, and proprioceptive accommodation function, respectively. The response lag time is calculated by recording the time difference between the stimulus excitation time and the time difference between the acceleration at the center of pressure exceeding the baseline by 3 times the standard deviation threshold; the response gain is obtained by establishing a stimulus-response transfer function model and calculating the ratio of system output to input. The estimated lower limb joint torque is obtained by inverse calculation based on the overall pressure center trajectory data of the support phase in the gait cycle. Specifically, it includes: identifying and extracting the overall pressure center trajectory data of the support phase in the gait cycle, constructing a high-dimensional spatiotemporal input vector containing the dynamic features of the current frame and the historical frame information within the preset stride length, introducing a multi-scale sliding window for segmentation and feature extraction, and using a cascaded random forest model to predict the torque of the ankle, knee and hip joints.

3. The method for assessing human balance function adapted to multi-tasking and complex scenarios according to claim 1, characterized in that, Based on the synchronized dataset, the plantar pressure distribution data is mapped to a unified spatial coordinate system through rigid body transformation, and a weighted fusion algorithm is used to reconstruct the overall pressure center trajectory of the subject throughout the static standing and dynamic walking processes, including: Based on the bipedal six-DOF pose data in the synchronized dataset, a rigid body transformation is used to map the preset local coordinates of the sensors to the world coordinate system, thereby obtaining the absolute coordinates of each pressure sensor in the world coordinate system. The expression for the rigid body transformation is: ； in, Indicates the first frame The side foot The absolute coordinates of each sensor in the world coordinate system; This represents the rotation matrix obtained by the rotation quaternion transformation; Represents the local coordinates of the sensor; This represents the fixed structural offset from the center of the tracker to the origin of the local coordinate system of the insole; This indicates the tracker's position coordinates in the world coordinate system; By introducing adaptive weighting coefficients and combining the projected sensor absolute coordinates with real-time pressure values, the global pressure center in the world coordinate system is calculated using the following formula: ； ； in, , They represent the first The overall pressure center of the frame is in the world coordinate system shaft and Position coordinates on the axis; This represents the filtered pressure value; Indicates the adaptive weighting coefficient; , These represent the sensor's position in the world coordinate system. shaft and Axis coordinates.

4. The method for assessing human balance function adapted to multi-tasking and complex scenarios according to claim 1, characterized in that, When the subject performs the dynamic walking task, gait event detection is performed based on the synchronized dataset and the overall center of pressure trajectory, including: The total pressure on one foot is calculated using the following expression: ； in, Indicates the first frame Total pressure on the sole of the side foot; This represents the filtered pressure value; A dual-threshold decision mechanism is used to detect gait events by setting a landing threshold and a takeoff threshold. When the total pressure on the sole of one foot is greater than or equal to the landing threshold, and the total pressure on the sole of the foot in the previous frame is less than the landing threshold, it is determined to be a landing event. The current time is recorded as the landing time, and the current foot position is recorded as the landing position. When the total pressure on the sole of one foot is less than or equal to the ground clearance threshold, and the total pressure on the sole of the foot in the previous frame is greater than the ground clearance threshold, it is determined to be a ground clearance event, and the current time is recorded as the ground clearance time. The geometric center points of both feet are calculated using the following expression: ； in, Indicates the first Geometric center point of both feet in frame world coordinate system; , These represent the position coordinates of the left and right foot trackers in the world coordinate system, respectively. A sliding window of preset duration is constructed, and a polynomial fitting is performed on the sequence of geometric center points of the two feet within the window to reconstruct the subject's walking path. The unit walking direction vector is obtained by calculating the normalized tangent vector of the fitted curve at the current moment.

5. The method for assessing human balance function adapted to multi-tasking and complex scenarios according to claim 1, characterized in that, The physiological function sub-models include visual regulation function sub-models, vestibular regulation function sub-models, proprioceptive regulation function sub-models, motor control and executive function sub-models, and interference resistance and adaptive control function sub-models.

6. The method for assessing human balance function adapted to multi-tasking and complex scenarios according to claim 1, characterized in that, Calculate the overall efficacy index of each physiological functional sub-model, including: Based on the norm database of healthy people, the feature components corresponding to each physiological function sub-model are standardized using the Z-Score standardization algorithm. A multi-factor weighted fusion algorithm was used to fuse and calculate the standardized feature vectors to obtain the comprehensive efficacy index of each physiological function sub-model.

7. A human balance function assessment system adapted to multi-tasking and complex scenarios, characterized in that, The system includes: The plantar pressure acquisition module is installed on both feet of the subject to collect plantar pressure distribution data during the assessment process. The bipedal pose tracking module includes two tracker modules, which are respectively installed on the subject's two feet to acquire bipedal six-DOF pose data; A multi-sensory stimulation module for applying controllable sensory stimulation to a subject, the sensory stimulation including at least one of visual stimulation, auditory stimulation and proprioceptive stimulation; The data processing module is communicatively connected to the plantar pressure acquisition module, the bipedal posture tracking module, and the multisensory stimulation module, respectively, and the data processing module is configured to perform the steps of the method according to any one of claims 1 to 6; The diagnostic decision module, which is connected in communication with the data processing module, is used to map the feature vector set to multiple preset physiological function sub-models, calculate the comprehensive efficacy index of each physiological function sub-model based on the healthy population norm database, and output the assessment results of the subject's balance control ability in different physiological function dimensions.

8. The human balance function assessment system adapted to multi-tasking and complex scenarios according to claim 7, characterized in that, The plantar pressure acquisition module includes multiple pressure sensors disposed in the insole, which are respectively disposed in the forefoot area and the heel area; The tracker module is rigidly mounted on the heel area of ​​the subject's shoes and maintains a fixed relative position with the plantar pressure acquisition module. Each tracker has a built-in visual sensor and inertial measurement unit, which calculates the six-degree-of-freedom pose data of the corresponding foot in real time based on visual-inertial real-time positioning and mapping technology.

9. The human balance function assessment system adapted to multi-tasking and complex scenarios according to claim 7, characterized in that, The multi-sensory stimulation module includes: The visual stimulation submodule includes a virtual reality headset for providing programmed, controlled visual stimuli to interfere with the subject's visual information; The auditory stimulation submodule includes a three-dimensional spatial audio system for providing programmed and controlled auditory stimuli to interfere with the subject’s auditory spatial reference information; The proprioceptive stimulation submodule is used to provide programmed and controllable proprioceptive stimuli to interfere with the subject's proprioceptive information, said proprioceptive stimuli including at least one of support surface perturbation stimuli and complex terrain stimuli.

10. The human balance function assessment system adapted to multi-tasking and complex scenarios according to claim 7, characterized in that, The system also includes a safety protection module, which is communicatively connected to the data processing module and the multi-sensory stimulation module. The safety protection module receives the dynamic stability margin output by the data processing module and controls the multi-sensory stimulation module to terminate the stimulation output when the dynamic stability margin is lower than a preset safety threshold.

Images