Early screening method and system for scoliosis based on continuous monitoring of plantar pressure and motion posture

Abstract

The application discloses a scoliosis early screening method and system based on continuous monitoring of plantar pressure and motion posture, relates to the field of human biomechanics multivariate measurement technology, and collects multiple-source physical measurement variables such as a plantar pressure distribution matrix sequence and a three-dimensional posture time sequence signal, is aligned with a time stamp, and is divided into independent gait cycle sequences through extreme value extraction; a pressure center dynamic swing trajectory and a multi-degree-of-freedom posture angle parameter in the sequence are extracted, a plantar support envelope area, a trajectory bias and a gait cycle symmetry degree feature are reconstructed and acquired, the area and bias features are reduced in dimension and spliced to construct a plantar pressure imbalance space vector, a nonlinear weighted modulation is performed on the space vector by using a penalty constraint condition constructed by using the symmetry degree feature, multi-dimensional conversion is performed to generate a multi-modal gait compensation fusion parameter, and the parameter is input into a geometric morphological mapping space to be inversely analyzed through a topological mechanism, and a spatial geometric physical quantity, namely a three-dimensional space measurement deflection angle, is output. The system is used for realizing the above method.

Description

Technical Field

[0001] This invention relates to the field of multivariate measurement technology in human biomechanics, specifically to a method and system for early screening of scoliosis based on continuous monitoring of plantar pressure and movement posture. Background Technology

[0002] Abnormal distortions in the three-dimensional geometric shape of the human spine are a significant factor jeopardizing the physical growth and development of adolescents and their daily physical health. Because early micro-adjustments in spinal curvature are highly concealed, and the trunk and skeletal structures typically lack readily apparent features in daily life, they are difficult for subjects and their families to detect in a timely manner without professional intervention. To meet the need for long-term, high-frequency daily monitoring that does not interfere with daily activities, there is an urgent need to utilize non-radioactive, portable wearable sensing devices to continuously and quantitatively capture and measure the compensatory biomechanical characteristics and posture features of the lower limbs during linear walking or movement.

[0003] However, compensation caused by skeletal deformities is a complex dynamic biomechanical evolution process. Existing measurement and assessment methods mostly rely on whole-spine medical imaging or manual geometric inspection in a static standing state. These static spatial parameter measurements within a single time period cannot reflect the dynamic posture evolution of the trunk during the actual continuous gait cycle. In addition, human walking in open environments is easily affected by transient external factors such as slight road surface undulations, uneven weight-bearing on one side, or transient muscle fatigue, leading to highly random physiological gait asymmetry. Existing data processing systems mostly use fixed thresholds or single variables for force comparison, making it difficult to deeply decouple and separate these transient environmental artifacts from the systematic mechanical compensation characteristics caused by rigid skeletal deformities in complex continuous monitoring signals. This results in data divergence, high false positive rates, and the accuracy of mapping multi-source heterogeneous signals to spatial geometric angles cannot meet practical engineering requirements. Therefore, it is still necessary to provide an early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture to improve the signal decoupling ability and geometric morphology analysis accuracy in the multivariate joint dynamic measurement process. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for early screening of scoliosis based on continuous monitoring of plantar pressure and movement posture to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture, comprising the following steps: S1, collecting multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time-series signal under continuous movement state; aligning the timestamps of both based on Bluetooth underlying protocol, extracting the vertical acceleration extreme value features of the three-dimensional posture time-series signal, and dividing the time-stamp-aligned multivariate joint data sequence into continuous independent gait cycle sequences; S2, extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom posture angle parameters within the independent gait cycle sequences, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing to obtain the plantar support envelope surface. Product and trajectory bias features; analyze the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait periodic symmetry features; S3, perform multi-dimensional dimensionality reduction and splicing of the foot support envelope area and trajectory bias features to construct a foot pressure imbalance space vector representing bilateral dynamic differences; extract gait periodic symmetry features to construct adaptive kinematic penalty constraints, perform nonlinear weighted modulation on the foot pressure imbalance space vector, and generate multimodal gait compensation fusion parameters through multi-dimensional heterogeneous signal conversion; S4, input the multimodal gait compensation fusion parameters into a preset geometric shape mapping space, and perform inverse spatial coordinate analysis of the parameters through a regression topology mechanism to output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measurement deflection angle.

[0006] In a preferred embodiment, the specific process of acquiring the multi-source physical measurement variables of the plantar pressure distribution matrix sequence and the three-dimensional attitude timing signal under continuous motion is as follows: the pressure resistance change is sensed at the intersection of the row and column electrodes in the flexible electrode array, and the multi-channel voltage discrete signal is obtained through the multiplexing array and analog-to-digital conversion channel. The plantar pressure distribution matrix sequence is generated after cascade amplification. The three-dimensional acceleration vector and three-dimensional angular velocity vector under motion are captured by the inertial measurement chip, and the three-dimensional attitude timing signal is generated in real time by the attitude calculation matrix.

[0007] In a preferred embodiment, the process of aligning the timestamps of both based on the Bluetooth underlying protocol, extracting the vertical acceleration extreme features of the three-dimensional attitude timing signal, and dividing the timestamp-aligned multivariate joint data sequence into continuous independent gait cycle sequences is as follows: The master control terminal synchronously sends the underlying broadcast wireless clock reference to the bilateral foot acquisition nodes. The bilateral foot acquisition nodes inject a global timestamp into the data packet header according to the wireless clock reference, thereby aligning the timestamps of the bilateral plantar pressure distribution matrix sequence and the bilateral three-dimensional attitude timing signal in the same time domain coordinate system. The local minimum and local maximum points of the vertical acceleration fluctuating with time in the timestamp-aligned three-dimensional attitude timing signal are extracted, and the data interval between two adjacent local minimum points is defined as an independent gait cycle sequence.

[0008] In a preferred embodiment, the specific process of extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom attitude angle parameters within an independent gait cycle sequence, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing to obtain the foot support envelope area and trajectory offset features is as follows: Calculate the force-weighted coordinates of each single-point array sensing unit within the independent gait cycle sequence to generate the dynamic swaying trajectory of the pressure center; analyze the three-dimensional attitude time-series signal within the independent gait cycle sequence to extract multi-degree-of-freedom attitude angle parameters; extract the maximum geometric outer contour boundary of the dynamic swaying trajectory of the pressure center in the two-dimensional horizontal plane, and calculate the area of ​​the closed region enclosed by the maximum geometric outer contour boundary as the foot support envelope area; perform projection decomposition on the dynamic swaying trajectory of the pressure center along the positive direction of motion and the horizontal vertical direction, calculate the maximum offset displacement vector of the dynamic swaying trajectory of the pressure center from the standard motion axis, and generate trajectory offset features.

[0009] In a preferred embodiment, the specific process of analyzing the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait period symmetry features is as follows: First multi-degree-of-freedom attitude angle temporal curves corresponding to the left-foot independent gait period sequence and second multi-degree-of-freedom attitude angle temporal curves corresponding to the right-foot independent gait period sequence are obtained respectively; temporal cross-correlation function operations are performed on the first and second multi-degree-of-freedom attitude angle temporal curves to extract the temporal phase difference at the peak position of the two sets of temporal curves; the temporal phase difference is input into a symmetry mapping function for overlap calculation to generate gait period symmetry features.

[0010] In a preferred embodiment, the specific process of constructing a foot pressure imbalance space vector characterizing the bilateral dynamic differences by multidimensional dimensionality reduction and splicing of the plantar support envelope area and trajectory bias features is as follows: The plantar support envelope area and trajectory bias features are subjected to dimensionless normalization processing using preset benchmark normal physical geometric quantities to obtain normalized area parameters and normalized bias parameters; the front-to-back distribution ratio features of the bilateral plantar pressure distribution matrix sequence are extracted to generate a longitudinally distributed asymmetric gradient; the longitudinally distributed asymmetric gradient, normalized area parameters, and normalized bias parameters are mapped to mutually orthogonal feature dimensions, and tensor splicing is performed to construct a multidimensional feature space, outputting the foot pressure imbalance space vector.

[0011] In a preferred embodiment, the specific process of extracting gait periodic symmetry features to construct adaptive kinematic penalty constraints, performing nonlinear weighted modulation on the foot pressure imbalance space vector, and generating multimodal gait compensation fusion parameters through multidimensional heterogeneous signal conversion is as follows: extracting a preset ideal steady-state extreme value benchmark, calculating the temporal difference deviation between the ideal steady-state extreme value benchmark and the gait periodic symmetry features, and constructing adaptive kinematic penalty constraints; and retrieving the discrete dynamic weight coefficients and kinematic compensation coefficients preset within the feature fusion module.

[0012] The dynamic feature components are obtained by performing spatial matrix scaling on the foot pressure imbalance space vector through discrete dynamic weight coefficients; the kinematic compensation components are obtained by performing positive scalar scaling on the adaptive kinematic penalty constraint through kinematic compensation coefficients; and the dynamic feature components and the kinematic compensation components are fused together to generate multimodal gait compensation fusion parameters.

[0013] In a preferred embodiment, the process of inputting the multimodal gait compensation fusion parameters into a preset geometric shape mapping space and performing inverse spatial coordinate analysis on the parameters through a regression topology mechanism to output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measured deflection angle, is as follows: A training feature set containing a calibrated deflection angle scalar and multi-source prior feature parameters is extracted; a minimum mean square error regression analysis model is constructed in the multi-dimensional feature space based on the training feature set to generate the geometric shape mapping space; the multimodal gait compensation fusion parameters are activated and projected onto a regression prediction analysis model in the geometric shape mapping space; the weight coefficient matrix and intercept bias parameter in the regression prediction analysis model are extracted, and inverse spatial coordinate analysis is performed on the multimodal gait compensation fusion parameters through the weight coefficient matrix and intercept bias parameter to output the three-dimensional spatial measured deflection angle.

[0014] The early scoliosis screening system based on continuous monitoring of plantar pressure and movement posture is used to execute the aforementioned early scoliosis screening method based on continuous monitoring of plantar pressure and movement posture. It includes: a data synchronization acquisition module for acquiring multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time-series signal during continuous movement; aligning the timestamps of both based on the Bluetooth underlying protocol, extracting the vertical acceleration extreme value features of the three-dimensional posture time-series signal, and dividing the time-stamp-aligned multivariate joint data sequence into continuous independent gait cycle sequences; and a multi-dimensional feature extraction module for extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom posture angle parameters within the independent gait cycle sequences, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing the plantar pressure... The system supports the features of the foot support envelope area and trajectory offset; analyzes the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait periodic symmetry features; a gait compensation fusion module is used to perform multi-dimensional dimensionality reduction and splicing of the foot support envelope area and trajectory offset features to construct a foot pressure imbalance spatial vector representing the bilateral dynamic differences; extracts gait periodic symmetry features to construct adaptive kinematic penalty constraints, performs nonlinear weighted modulation on the foot pressure imbalance spatial vector, and generates multi-modal gait compensation fusion parameters through multi-dimensional heterogeneous signal conversion; a deflection angle calculation module is used to input the multi-modal gait compensation fusion parameters into a preset geometric shape mapping space, and performs inverse spatial coordinate analysis of the parameters through a regression topology mechanism to output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial calculated deflection angle.

[0015] The technical effects and advantages of this invention are as follows:

[0016] (1) An early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture. In the signal acquisition and benchmark division stage, multi-source physical variables to be measured are acquired by collecting the plantar pressure distribution matrix sequence and three-dimensional posture time sequence signal under continuous movement. The underlying communication protocol is used to align the timestamp and vertical acceleration extreme value features to perform multi-variable joint gait cycle segmentation and establish a unified time domain transformation interval. Subsequently, the dynamic swing trajectory of the pressure center and multi-degree-of-freedom posture angle parameters are extracted in the independent gait cycle sequence. The plantar support envelope area, trajectory offset features and gait cycle symmetry features are reconstructed by using envelope extraction, vector decomposition and temporal differential phase analysis. Thus, the refined extraction of the dynamic contour boundary of the bilateral lower limbs and multi-degree-of-freedom kinematic morphological parameters is realized, providing a high-confidence underlying variable support with reference basis for subsequent elimination of spatiotemporal random artifact interference in open environments.

[0017] (2) The early screening system for scoliosis based on continuous monitoring of plantar pressure and movement posture, in the comprehensive analysis and morphological mapping stage, constructs a spatial vector of plantar pressure imbalance by multidimensionally reducing and splicing the plantar support envelope area and trajectory offset features, and implements nonlinear weighted modulation by constructing adaptive kinematic penalty constraints using gait period symmetry features. Multimodal gait compensation fusion parameters are generated through multidimensional heterogeneous signal conversion. Finally, the multimodal gait compensation fusion parameters are projected into a preset geometric morphological mapping space, and the spatial coordinates of the fusion parameters are reversed using a regression topology mechanism to calculate the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measurement deflection angle. Thus, by deeply nonlinearly reconstructing the feature vector representing dynamic differences and the constraint boundary representing kinematic deviations, the cross-rigid binding characteristics of dual-modal parameters are used to effectively eliminate flexible interference noise caused by load or terrain fluctuations, realizing high-precision topological mapping from multi-source open dynamic variables to the final structural geometric bending angle.

[0018] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0019] Figure 1 This is a flowchart of the early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture according to the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the extraction of the dynamic swaying trajectory of the pressure center and the support envelope area in an embodiment of the present invention;

[0021] Figure 3This is a schematic diagram of the regression topology mapping from multimodal fusion parameters to the measured deflection angle in an embodiment of the present invention;

[0022] Figure 4 This is a flowchart of the early scoliosis screening system based on continuous monitoring of plantar pressure and movement posture according to the present invention.

[0023] Figure 5 This is a hardware structure block diagram of the wearable gait information acquisition system in an embodiment of the present invention.

[0024] Figure 6 This is a schematic diagram of the three-dimensional attitude timing signal waveform and gait period division in an embodiment of the present invention. Detailed Implementation

[0025] This application provides an early screening method and system for scoliosis based on continuous monitoring of plantar pressure and movement posture. This method addresses the problems of divergence in spatial geometric morphology assessment data, high false positive rate, and low accuracy of indirect mapping of heterogeneous signals caused by instantaneous environmental interference and daily temporary flexible compensation noise in continuous dynamic monitoring.

[0026] The overall approach of the scheme in this application embodiment is as follows: First, the system performs time-stamp alignment on the synchronously acquired multi-source physical measurement variables through a low-level wireless broadcast mechanism, and captures the variation range of vertical acceleration kinematic feature points in the inertial attitude time-series signal. The aligned continuous data is jointly cropped and divided into independent gait cycle time-domain basic units to establish a common time-domain benchmark for feature analysis. Second, within the independent gait cycle sequence, the dynamic swaying trajectory of the pressure center at the force dynamics level and the multi-degree-of-freedom attitude angle parameters at the posture kinematics level are simultaneously separated. The dynamic swaying trajectory of the pressure center is subjected to geometric boundary envelope extraction and spatial orthogonal coordinate axis projection vector decomposition in the horizontal plane to calculate the dimensionless foot support envelope area and the trajectory offset feature representing the center point offset; at the same time, the temporal phase difference of the multi-degree-of-freedom attitude angle parameters in the bilateral alignment single-step cycle is analyzed to calculate and extract the gait cycle symmetry feature. Next, the extracted plantar support envelope area and trajectory offset features are dimensionality-reduced and stitched together using spatial feature axes to construct a foot pressure imbalance spatial vector representing the difference in bilateral dynamic force imbalance. Adaptive kinematic penalty constraints are then constructed using gait periodic symmetry features as feedback parameters. These penalty constraint boundaries are used to perform nonlinear weighted modulation of the foot pressure imbalance spatial vector on a spatial matrix scale. Through cross-conversion of multidimensional heterogeneous signals, a unified multimodal gait compensation fusion parameter is formed. Finally, minimum mean square error regression analysis is performed using a training feature set containing prior calibration deflection angle data and multi-source feature parameters to pre-construct a geometric shape mapping space that maps from the multidimensional feature space to the entity's geometric angle. The multimodal gait compensation fusion parameter is input into the regression prediction analysis model, and its internally deployed weight coefficient matrix and intercept bias parameters are extracted. Inverse spatial geometric coordinate analysis is performed on the multimodal gait compensation fusion parameter, ultimately calculating and outputting a non-specific physical quantity corresponding to the actual degree of skeletal deformity—the three-dimensional spatially calculated deflection angle of the target object.

[0027] Example 1; please refer to Figure 1This invention provides a technical solution: an early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture, comprising the following steps: S1, collecting multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time-series signal under continuous movement state; aligning the timestamps of both based on Bluetooth underlying protocol, extracting the vertical acceleration extreme value features of the three-dimensional posture time-series signal, and dividing the time-stamp-aligned multivariate joint data sequence into continuous independent gait cycle sequences; S2, extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom posture angle parameters within the independent gait cycle sequences, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing to obtain the plantar support envelope area and... Trajectory bias features; analyze the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait periodic symmetry features; S3, perform multi-dimensional dimensionality reduction and splicing of the foot support envelope area and trajectory bias features to construct a foot pressure imbalance space vector representing bilateral dynamic differences; extract gait periodic symmetry features to construct adaptive kinematic penalty constraints, perform nonlinear weighted modulation on the foot pressure imbalance space vector, and generate multimodal gait compensation fusion parameters through multi-dimensional heterogeneous signal conversion; S4, input the multimodal gait compensation fusion parameters into a preset geometric shape mapping space, perform inverse spatial coordinate analysis of the parameters through a regression topology mechanism, and output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measurement deflection angle.

[0028] In this implementation scheme, step S1 mainly performs synchronous acquisition of multi-source heterogeneous signals and temporal reference division. The system acquires the plantar pressure distribution matrix sequence and three-dimensional posture time-series signal of the target object during continuous walking. The plantar pressure distribution matrix sequence refers to the set of discrete spatial force values ​​of each region of the foot continuously acquired in time sequence by a flexible array sensor, and the three-dimensional posture time-series signal refers to the time-varying spatial vectors of acceleration and angular velocity acquired synchronously by an inertial sensor. After unifying the time axis of the two physical quantities using the underlying wireless communication protocol, the system further extracts the local maximum and minimum points of vertical acceleration in the three-dimensional posture signal. These vertical acceleration extreme values ​​correspond to the physical and mechanical moments of heel contact with the ground and toe lift-off during human walking, thereby dividing the continuous time series into independent gait cycles. This operation provides a basic time transformation coordinate system for subsequent joint measurement of dynamic and kinematic variables within the same physical cycle.

[0029] Step S2 involves deconstructing and extracting spatial and kinematic parameters from the segmented independent gait cycles. At the mechanical level, the system analyzes the dynamic swaying trajectory of the pressure center, i.e., the continuous movement path of the equivalent center point of the overall force on the sole of the foot on the support surface. Envelope extraction is performed on this path, which involves calculating the maximum outer contour boundary covered by the geometric trajectory, thereby obtaining the closed plantar support envelope area. Simultaneously, the trajectory is decomposed by projection to obtain the physical offset from the standard linear motion axis, constituting the trajectory offset feature. At the kinematic level, the system extracts multi-degree-of-freedom posture angle parameters of the lower limb joints and calculates the temporal phase difference of the corresponding posture angle fluctuation curves of the left and right limbs on the time axis. The temporal phase difference reflects the time delay of bilateral movement alternation, and the system quantifies and extracts the gait cycle symmetry feature accordingly. This step completes the quantification and extraction of the core geometric and temporal attributes within different physical parameters.

[0030] Step S3 is responsible for performing joint data fusion and signal conversion of heterogeneous physical variables. The system first maps the extracted plantar support envelope area and trajectory bias features to mutually orthogonal mathematical dimensions for matrix combination, forming a spatial vector of foot pressure imbalance that comprehensively reflects the bilateral force imbalance state. Since external physical disturbances such as road surface undulations can cause random asymmetric movements in the human body, the system uses gait periodic symmetry features to construct adaptive kinematic penalty constraints. This constraint, as a dynamic weight adjustment mechanism, performs nonlinear weighted modulation on the foot pressure imbalance spatial vector. When severe random asymmetry occurs in the movement posture, the system automatically reduces the weight of the pressure imbalance vector using the penalty constraint. Through this cross-modal data modulation and conversion, the separated dynamic and kinematic features are cross-fused into a single multimodal gait compensation fusion parameter.

[0031] Step S4 performs topological analytical calculations from multidimensional transformation parameters to macroscopic physical quantities. The multimodal gait compensation fusion parameters generated by the fusion transformation in step S3 reside in a high-dimensional abstract data space. The system inputs these parameters into a geometric shape mapping space pre-established based on sample data. This geometric shape mapping space contains a quantitative mapping relationship between the trunk's pose state and underlying foot pressure and movement characteristics. The system utilizes a regression topology mechanism to perform inverse spatial coordinate system calculations on the input high-dimensional feature parameters within this mapping space. The inverse analysis process directly skips the medical-level qualitative disease analysis, using a pure geometric projection transformation to ultimately calculate and output an objective, non-specific physical measurement result: the target object's trunk's measured deflection angle in three-dimensional space.

[0032] Specifically, the process of acquiring the plantar pressure distribution matrix sequence and three-dimensional attitude timing signal of multi-source physical measurement variables under continuous motion is as follows: pressure resistance changes are sensed at the intersection of row and column electrodes in the flexible electrode array; multi-channel discrete voltage signals are acquired through multiplexing array and analog-to-digital conversion channel; and plantar pressure distribution matrix sequence is generated after cascade amplification; the three-dimensional acceleration vector and three-dimensional angular velocity vector under motion are captured by inertial measurement chip; and three-dimensional attitude timing signal is generated in real time by attitude calculation matrix.

[0033] In this implementation scheme, the mechanical deformation is first converted into an electrical signal by leveraging the reduced contact resistance caused by the increased contact area of ​​the internal conductive particles in the distributed flexible array layer under mechanical pressure. A multiplexing circuit sequentially activates the cross-sensing points of each row and column electrode at a preset scanning frequency to obtain the analog voltage at that physical location. This analog voltage is then amplified through a multi-stage cascaded operational amplifier and input to an analog-to-digital converter channel for discrete quantization. The specific nonlinear mapping calculation process for converting this multi-channel discrete voltage signal into a plantar pressure distribution matrix is ​​expressed as follows: ;in, Indicates time The resolved pressure amplitude at the sensor array point located in row u and column v; This represents the factory-preset calibration conversion constant of the sensor material; Indicates the quantization step resolution of the analog-to-digital conversion channel; This indicates the total gain of the cascaded amplifier circuit. This indicates the voltage value of the constant voltage reference drive excitation source; This represents the fixed matching reference resistor connected in series in the voltage divider sampling circuit; This represents the dynamic equivalent resistance value of the corresponding site caused by external pressure deformation. This represents the base zero bias constant under passive static conditions; This represents the power function compensation exponent set for the hysteresis and creep effects of flexible polymer film materials. In the kinematic signal extraction stage, the microelectromechanical gyroscope and accelerometer inside the inertial measurement chip sense the spatial motion of the carrier in orthogonal directions. The hardware calculation unit uses a quaternion integration algorithm to perform discrete-time accumulation of the angular velocity and cross-decouples it from the acceleration-gravity vector. The real-time attitude calculation matrix construction process of this three-dimensional attitude time-series signal is expressed as: Based on this, extract the attitude direction cosine matrix. ;in, This represents the updated spatial attitude quaternion vector; This represents the spatial attitude quaternion vector of the previous system sampling period; This represents a fourth-order antisymmetric operator matrix constructed from the arrangement of elements of the three-dimensional angular velocity vector captured at the current moment; This represents the fixed sampling discrete period of the inertial sensor. Represents the mapping function of quaternions The generated direction cosine matrix serves as the core coordinate transformation reference for subsequent spatial projection and kinematic analysis of the three-dimensional attitude time-series signal.

[0034] Specifically, based on the Bluetooth underlying protocol to align the timestamps of both, the vertical acceleration extreme features of the three-dimensional attitude timing signal are extracted, and the multivariate joint data sequence after timestamp alignment is divided into continuous independent gait cycle sequences. The specific process is as follows: The master control terminal synchronously sends the underlying broadcast wireless clock reference to the bilateral foot acquisition nodes. The bilateral foot acquisition nodes inject a global timestamp into the data packet header according to the wireless clock reference, so as to realize the timestamp alignment of the bilateral plantar pressure distribution matrix sequence and the bilateral three-dimensional attitude timing signal in the same time domain coordinate system. The local minimum and local maximum points of vertical acceleration fluctuating with time in the three-dimensional attitude timing signal after timestamp alignment are extracted, and the data interval between two adjacent local minimum points is defined as an independent gait cycle sequence.

[0035] In this implementation, the multi-node time-domain alignment process utilizes the connectionless broadcast communication mechanism of the Bluetooth Low Energy (BLE) underlying layer. The master control terminal embeds the system absolute time counter value of the handheld device's master control chip into the broadcast physical layer message it sends. After capturing the broadcast preamble, the RF antennas of the dual-foot acquisition nodes trigger an external hardware interrupt and latch the local timer count value within the node. This process strictly calibrates the local independent clock by compensating for the fixed physical delay of the over-the-air wireless RF wave transmission. The timestamp alignment conversion rule under the same time-domain coordinate system is expressed as follows: ;in, This represents the globally standardized timestamp sequence value after alignment of the nth hardware sampling data point; This represents the relative clock tick count recorded by the underlying timer of the foot acquisition node; This represents the initial physical offset difference between the absolute reference time of the broadcast system and the local initialization start time of the foot. This represents a fixed air link transmission time compensation constant pre-determined by the crystal oscillator frequency offset and message length between the sender and receiver. After acquiring the multivariate joint data sequence with absolute timestamp alignment, the system processes the gravity-removed acceleration component perpendicular to the geographic horizontal plane in the three-dimensional posture sequence to extract independent gait cycles. Since the subject experiences a significant and steep vertical high-frequency reverse impact at the moment the heel touches the ground during walking, the algorithm sets a data sliding time window on this one-dimensional time-series data stream, searching for local minima below a preset dynamic judgment threshold as gait cycle segmentation anchor points. The local segmentation judgment condition for this extreme feature is expressed as: and ;in, The vertical acceleration discrete amplitude corresponding to the timestamp m is represented; w represents the width of the sliding decision window on one side and the number of points, which is determined by converting the extreme value of the reciprocal of the highest normal walking frequency of the human body to determine the limit of the number of points. This refers to the dynamic vertical impact threshold used to exclude high-frequency wandering noise. To address the issue that traditional fixed values ​​cannot accommodate variations in test surface hardness and subject weight, the publicly disclosed method for determining this dynamic vertical impact threshold is as follows: Within an initial calibration acquisition window containing three consecutive complete pre-walks, the statistical baseline state parameters of the vertical acceleration components during that time period are calculated, and the calculation is performed. ;in, This represents the expected baseline arithmetic mean of the vertical acceleration components within the calibration acquisition window. This parameter represents the discrete standard deviation of the data fluctuation within the calibration acquisition window. This represents the confidence interval attenuation penalty ratio factor set based on prior engineering experience. The system compares two adjacent timestamp segmentation anchor points on the time axis that simultaneously satisfy both the minimum and dynamic threshold conditions, and losslessly extracts all pressure matrix sequences and 3D attitude sequences between them to form a complete closed-loop independent gait cycle sequence for subsequent multi-dimensional fusion processing. Figure 6 As shown in the figure, this diagram visually illustrates the underlying kinematic data captured by the system during continuous walking. The upper part is the acceleration magnitude fluctuation curve of the inertial measurement unit (IMU), where the vertical axis represents the acceleration magnitude and the horizontal axis represents time. The lower part is the angular velocity magnitude fluctuation curve of the IMU, where the vertical axis represents the angular velocity magnitude and the horizontal axis represents time. The labels "devo" (representing device A) and "dev1" (representing device B) in the figure represent the bilateral foot-level data acquisition nodes worn on the left and right lower limbs of the subject, respectively. Figure 6 As shown, the upper graph shows the acceleration variation of the inertial measurement unit (IMU), and the lower graph shows the angular velocity variation of the IMU. The two sets of timing waveforms output by device A and device B are strictly aligned on the time axis (horizontal axis), which objectively verifies the effectiveness of the system's underlying global timestamp binding based on the Bluetooth wireless clock reference, eliminating the timing drift caused by independent sampling of multiple nodes. In the acceleration magnitude curve, due to the alternating movement of both feet, the signal exhibits a regular alternation of peaks and troughs. The local minimum troughs of the curve in the figure precisely correspond to the transient mechanical impact reference point generated when the subject's feet touch the ground. The system directly reads this acceleration magnitude waveform and locks these troughs as segmentation anchor points by applying a dynamic vertical impact threshold, thereby dividing the continuous data stream into independent gait cycle sequences. In addition, the lower angular velocity magnitude curve synchronously reflects the spatial attitude change waveform of the lower limb swing phase within the independent gait cycle, providing a reliable data source for the subsequent extraction of multi-degree-of-freedom attitude angle parameters and timing phase differences.

[0036] Specifically, the process of extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom attitude angle parameters within an independent gait cycle sequence, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing to obtain the foot support envelope area and trajectory offset features is as follows: Calculate the force-weighted coordinates of each single-point array sensing unit within the independent gait cycle sequence to generate the dynamic swaying trajectory of the pressure center; analyze the three-dimensional attitude time-series signal within the independent gait cycle sequence to extract multi-degree-of-freedom attitude angle parameters; extract the maximum geometric outer contour boundary of the dynamic swaying trajectory of the pressure center in the two-dimensional horizontal plane, and calculate the area of ​​the closed region enclosed by the maximum geometric outer contour boundary as the foot support envelope area; perform projection decomposition on the dynamic swaying trajectory of the pressure center along the positive direction of motion and the horizontal and vertical directions, calculate the maximum offset displacement vector of the dynamic swaying trajectory of the pressure center from the standard motion axis, and generate trajectory offset features.

[0037] In this implementation scheme, the system performs spatial weighted summation on the data from the single-point array sensing units at each discrete time to calculate the dynamic swaying trajectory of the pressure center. This calculation process is expressed as follows: ; ;in, and These represent the discrete sampling points. The horizontal and vertical coordinates of the pressure center at the location; and These represent the physical number of rows and columns of the sensor array, respectively. and Represents the physical position constant of the sensor node in row p and column q; This represents the real-time force value collected at that node at that moment. Subsequently, based on the pre-analyzed attitude direction cosine matrix, the Euler angle sequence of the lower limbs is extracted, completing the direct extraction of multi-degree-of-freedom attitude angle parameters. For the continuous set of pressure center coordinate points generated in the two-dimensional horizontal plane, the system calls the Graham scan algorithm to extract the outer limit coordinate points to form the maximum geometric outer contour boundary, and uses the polygon vertex determinant formula to calculate the area of ​​the closed region as the foot support envelope area. The calculation model is expressed as: ;in, Indicates the area of ​​the plantar support envelope; This represents the total number of vertices that constitute the boundary of the maximum outer contour. and Let represent the absolute coordinates of the h-th boundary vertex, and specify . and This forms a closed loop. Finally, the extreme differences of the pressure center in the orthogonal directions within a single cycle are extracted to calculate the maximum offset displacement vector. The calculation formula is: ;in, This indicates the lateral trajectory offset characteristic of the motion; This represents the reference coordinates of the standard motion axis formed by connecting the start and end points of this independent gait cycle. This step, through matrix analysis in a multivariable spatial coordinate system, completes the digital transformation from discrete dynamic variables to macroscopic geometric boundary characteristics. Combined with... Figure 2 As shown, in this two-dimensional orthogonal coordinate system, the horizontal axis represents the horizontal lateral displacement of the system, and the vertical axis represents the displacement in the positive direction of motion. The discretely distributed scatter plot in the figure intuitively reflects the pressure center within the independent gait cycle calculated by the formula. The dynamic swaying trajectory over time; the outer dashed closed region represents the maximum geometric outer contour boundary extracted by the Graham scan algorithm, and the area of ​​the overlay inside this boundary is the foot support envelope area obtained by integration. .also, Figure 2 The standard longitudinal motion centerline is marked, along with the maximum lateral distance of the extreme edges of the scatter group from this centerline. This physical distance directly maps the trajectory offset features extracted by the algorithm. This figure illustrates that the proposed scheme effectively achieves the objective measurement conversion from microscopic array electrical signals to macroscopic physical geometric contours.

[0038] Specifically, the process of analyzing the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait period symmetry features is as follows: First multi-degree-of-freedom attitude angle time-series curves corresponding to the left-foot independent gait period sequence and second multi-degree-of-freedom attitude angle time-series curves corresponding to the right-foot independent gait period sequence are obtained respectively; Temporal cross-correlation function operations are performed on the first and second multi-degree-of-freedom attitude angle time-series curves to extract the temporal phase difference at the peak position of the two sets of time-series curves; The temporal phase difference is input into a symmetry mapping function for overlap calculation to generate gait period symmetry features.

[0039] In this implementation scheme, the system extracts the sagittal plane pitch angle time-series data of the left foot independent gait cycle sequence as the first multi-degree-of-freedom attitude angle time-series curve, and extracts the corresponding right foot independent gait cycle data as the second multi-degree-of-freedom attitude angle time-series curve. The system performs a time-domain cross-correlation function operation on the two sets of time-series curves to extract the time-series phase difference. The specific operation rules are as follows: ;in, Indicates the discrete offset as The numerical relationships between time domain parameters; This represents the total length of sampling points for an independent gait cycle; This indicates the timing signal for the left foot attitude angle; This represents the timing signal of the right foot attitude angle after offset. The system iterates through the sequence to find... The offset limit when taking the global maximum value The timing phase difference is calculated by converting the time delay between the two peaks. The timing phase difference is then input into a symmetry mapping function to calculate the gait period symmetry feature. The mapping model is expressed as: ;in, The value representing the normalized gait periodicity symmetry characteristic; The theoretical half-cycle phase difference constant represents the human body's normal bipedal walking state. This represents the weighting coefficient for gait asymmetry attenuation. The method for determining the gait is as follows: A clinical biomechanical gait calibration database, collected under legal authorization and thoroughly desensitized to eliminate personal privacy features, is introduced. The actual phase distribution variance of the sample sets identified as normal physiological gait in the database is extracted, and the reciprocal of this variance is calculated for parameter fitting. This step achieves the numerical algebraic transformation of heterogeneous kinematic time-series signals into dimensionless symmetric characteristic variables.

[0040] Specifically, the process of constructing a foot pressure imbalance space vector representing bilateral dynamic differences by multidimensionally reducing and stitching together the plantar support envelope area and trajectory bias features is as follows: The plantar support envelope area and trajectory bias features are subjected to dimensionless normalization processing using preset baseline normal physical geometric quantities to obtain normalized area parameters and normalized bias parameters; the anterior-posterior distribution ratio features of the bilateral plantar pressure distribution matrix sequence are extracted to generate a longitudinally distributed asymmetric gradient; the longitudinally distributed asymmetric gradient, normalized area parameters, and normalized bias parameters are mapped to mutually orthogonal feature dimensions, and tensor stitching is performed to construct a multidimensional feature space, outputting the foot pressure imbalance space vector.

[0041] In this implementation plan, the system extracts standard physical scale baselines from a legally authorized and anonymized gait database of healthy adolescents, and establishes preset baseline physical quantities for the plantar envelope area and trajectory offset. The system calls upon the plantar support envelope area and trajectory offset features generated in previous steps, performs dimensionless normalization through ratio calculation, and obtains the conversion formulas for the normalized area parameter and the normalized offset parameter, expressed as follows: ; ;in, This represents the normalized area parameter after eliminating full-scale individual differences. This represents the preset baseline physical quantity of the plantar envelope area; This represents the normalized bias parameter that eliminates the dimensions of absolute displacement. This represents the preset physical quantity of the motion trajectory bias reference. Subsequently, the system divides the unilateral plantar pressure distribution matrix sequence into forefoot and heel physical regions according to anatomical proportions. Distribution proportion features are extracted through discrete integration, and a longitudinal asymmetric gradient is generated. This signal conversion process is represented as follows: ;in, represents the longitudinally distributed asymmetric gradient; e and g represent the discrete spatial indices of the forefoot and heel sensor nodes, respectively. and These represent the total number of independent sensor array units distributed in the front and rear regions, respectively. This represents the cumulative stress value collected by the e-th front node within the intact support phase; This represents the cumulative stress value of the g-th subsequent node within the corresponding period. Finally, the system sets the dynamic feature space as a mutually orthogonal basis system, performs tensor concatenation on the multidimensional parameters obtained from the above transformation to construct a multidimensional feature space, and the output is represented as... ;in, This represents the column matrix of the output foot pressure imbalance space vector; This represents the overall left-right asymmetry index component, which is pre-analyzed from the bilateral plantar pressure distribution matrix sequence.

[0042] Specifically, the process of extracting gait periodic symmetry features to construct adaptive kinematic penalty constraints, performing nonlinear weighted modulation on the foot pressure imbalance space vector, and generating multimodal gait compensation fusion parameters through multidimensional heterogeneous signal conversion is as follows: extracting the preset ideal steady-state extreme value benchmark, calculating the temporal difference deviation between the ideal steady-state extreme value benchmark and the gait periodic symmetry features, and constructing adaptive kinematic penalty constraints; and retrieving the discrete dynamic weight coefficients and kinematic compensation coefficients preset within the feature fusion module.

[0043] In this implementation scheme, the system first extracts the ideal steady-state extreme value benchmark representing the absolutely symmetrical walking of the human body on both sides, and calls the gait periodic symmetry feature calculated in the previous steps. It then constructs adaptive kinematic penalty constraints through time-domain differential deviation calculation. The differential constraint mapping relationship is expressed as follows: ;in, This represents the numerical value of the adaptive kinematic penalty constraint used to quantify the magnitude of kinematic compensation; Represents the ideal steady-state extreme value reference physical constant; This represents the normalized gait periodicity symmetry feature generated by the preceding analysis. Subsequently, the system retrieves the discrete dynamic weight coefficients and kinematic compensation coefficients preset within the feature fusion module, performs nonlinear weighted modulation and heterogeneous cascade fusion on the foot pressure imbalance spatial vector at the spatial matrix scale, and generates the underlying transformation operation of the multimodal gait compensation fusion parameters, represented as follows: ;in, This represents the multimodal gait compensation fusion parameter generated after multidimensional heterogeneous signal conversion; k represents the feature dimension traversal index of the multidimensional feature space. The total physical dimension value of the foot pressure imbalance space vector; Represents the scaling coefficients of the discrete dynamics weight coefficient vector in the k-th dimension; The specific scalar value of the feature extraction in the corresponding dimension represents the spatial vector of foot pressure imbalance. Indicates the kinematic compensation coefficient; This represents the nonlinear activation exponential constant used to perform positive scalar scaling. The method for determining the discrete dynamics weight coefficients and kinematic compensation coefficients is as follows: Gait data from a legally authorized and anonymized clinical multimodal biomechanical validation cohort, stripped of all sensitive privacy information, is aggregated. A feedforward deep neural network model containing a multidimensional feature input layer and a fully connected fusion layer is built on edge computing nodes. The gold standard physical distortion angle of the collected samples is used as the supervision label for model training. The gradient descent algorithm combined with a momentum optimizer is used to iteratively minimize the mean squared error loss function between the predicted output and the true label. After the model training converges and the prediction error of the physical validation set is less than the preset engineering standard, the kernel weight network tensor parameters, which are embedded in the fully connected fusion layer, are extracted and sent to the detection terminal as the physical conversion modulation constants directly invoked during the implementation example.

[0044] The dynamic feature components are obtained by performing spatial matrix scaling on the foot pressure imbalance space vector through discrete dynamic weight coefficients; the kinematic compensation components are obtained by performing positive scalar scaling on the adaptive kinematic penalty constraint through kinematic compensation coefficients; and the dynamic feature components and the kinematic compensation components are fused together to generate multimodal gait compensation fusion parameters.

[0045] In this implementation scheme, the system calls the pre-stored discrete dynamic weight coefficient matrix to perform dot product scaling on the characteristic dimension of the foot pressure imbalance space vector output by the previous processing node, and calculates the dynamic characteristic components. The specific operation is expressed as follows: ;in, Indicates the first Dynamic characteristic components of each gait processing cycle; d represents the total feature dimension of the foot pressure imbalance spatial vector; d represents the spatial index of the feature dimension. This represents the discrete dynamic weight coefficient corresponding to the d-th dimension; This represents the specific extracted value of the foot pressure imbalance space vector in this dimension within the current gait cycle. Subsequently, the system extracts the kinematic compensation coefficients and performs positive scalar scaling on the adaptive kinematic penalty constraints to calculate the kinematic compensation components, represented as... ;in, Indicates the kinematic compensation component; Indicates the kinematic compensation coefficient; This represents the adaptive kinematic penalty constraint condition calculated independently within the current cycle. To obtain the final fusion parameters, the system performs spatial concatenation and bias fusion on the two heterogeneous feature components mentioned above to generate multimodal gait compensation fusion parameters. The underlying concatenation calculation logic is represented as follows: ;in, Indicates the parameters for multimodal gait compensation fusion; This represents the inherent static instrument measurement bias constant of the system. To ensure the physical objectivity and compliance of the underlying parameter settings, the aforementioned discrete dynamic weighting coefficients... With kinematic compensation coefficient The determination method is as follows: by accessing a legally authorized open-source database of gait biomechanics of non-specific populations that has undergone strict desensitization and de-identification processing by the data provider, the extreme values ​​of bottom stress distribution, spatial posture deviation limits, and corresponding baseline geometric distortion data are extracted from a large number of samples. After removing environmental noise using principal component analysis, the principal component matrix is ​​fitted and solved using partial least squares method, and this set of static modulation coefficients is solidified to achieve unified dimensional transformation and nonlinear cascade modulation of heterogeneous physical parameters in the same mathematical coordinate system.

[0046] Specifically, the process of inputting the multimodal gait compensation fusion parameters into a preset geometric shape mapping space and performing inverse spatial coordinate analysis on these parameters through a regression topology mechanism to output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measurement deflection angle, is as follows: Extracting a training feature set containing a calibrated deflection angle scalar and multi-source prior feature parameters; constructing a minimum mean square error regression analysis model in the multi-dimensional feature space based on the training feature set to generate the geometric shape mapping space; activating and projecting the multimodal gait compensation fusion parameters onto a regression prediction analysis model in the geometric shape mapping space; extracting the weight coefficient matrix and intercept bias parameters from the regression prediction analysis model; and performing inverse spatial coordinate analysis on the multimodal gait compensation fusion parameters using the weight coefficient matrix and intercept bias parameters to output the three-dimensional spatial measurement deflection angle.

[0047] In this implementation, the inverse analysis of spatial geometric coordinates relies on a geometric shape mapping space pre-constructed within the system. This space is generated using a static X-ray skeleton projection geometry dataset that has undergone image masking and anonymization as a training feature set. The system builds a minimum mean square error regression analysis model on this feature set, incorporating elastic network hybrid regularization constraints. The loss function optimization logic of its underlying regression topology is expressed as follows: ;in, This represents the mixed objective loss function for global regression optimization; This represents the total number of valid samples in the desensitized training feature set; s represents the discrete traversal index of a single sample. This indicates the scalar geometric reference value for the calibrated deflection angle; This represents the multi-source prior feature parameter matrix corresponding to the training feature set samples; This represents the weight coefficient matrix in the regression prediction analysis model that is to be optimized and converged. Indicates the intercept offset parameter; and L2 and L1 regularization hyperparameters, respectively, are introduced to suppress overfitting, reduce network complexity, and improve model feature sparsity. During the model construction phase, the system utilizes an adaptive moment estimation gradient descent algorithm to achieve... The convergence to the global minimum allows for the identification and extraction of the optimal weight coefficient matrix. With intercept offset parameter The parameters are then burned into the measurement terminal. During the real-time calculation phase of dynamic continuous monitoring, the system activates the multimodal gait compensation fusion parameters and projects them into this mapping space. It then directly calls the pre-set topology parameters to perform inverse analytical calculation of spatial coordinates, expressed as: ;in, This represents the spatial geometric physical quantity of the target object in the final inverse analysis output, namely the three-dimensional spatial calculated deflection angle. This process breaks through the limitation of traditional multimodal evaluation algorithms that only provide black-box classification risk probabilities. By introducing strongly constrained regression topology, it directly maps complex heterogeneous gait mechanics fusion parameters into specific spatial geometric angle physical quantities, achieving high-precision physical inversion from discrete electrical signals from non-specific sensors to the deflection angle of a macroscopic continuous shape. Combined with... Figure 3 As shown in the figure, this diagram illustrates the three-dimensional topological structure obtained through inverse analysis of the target space coordinates. Figure 3 In the defined 3D space, the X-axis (dynamic feature components) and Y-axis (kinematic compensation components) constitute the underlying fusion feature basis, and the Z-axis represents the output spatial geometric physical quantity (measured deflection angle). The smooth 3D mesh surface in the figure represents the preset geometric shape mapping space solidified by the training feature set and the hybrid objective loss function; the black solid points scattered in the space represent the multimodal gait compensation fusion parameters calculated in real time under different gait cycles. The algorithm activates specific regression topology nodes, projects the heterogeneous compensation features of the X and Y axis dimensions along the surface normal, directly reverse analyzes and outputs the corresponding Z-axis height (i.e., deflection angle). Figure 3 This intuitively verifies that the topological mechanism can achieve high-precision geometric angle inversion without relying on any disease qualitative analysis, solely through nonlinear coordinate mapping of spatial surfaces.

[0048] Example 2; please refer to Figure 4An early scoliosis screening system based on continuous monitoring of plantar pressure and movement posture, used to execute the early scoliosis screening method based on continuous monitoring of plantar pressure and movement posture described in the embodiments, includes: a data synchronization acquisition module, used to acquire multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time-series signal under continuous movement state; aligning the timestamps of the two based on Bluetooth underlying protocol, extracting the vertical acceleration extreme value feature of the three-dimensional posture time-series signal, and dividing the time-stamp aligned multivariate joint data sequence into continuous independent gait cycle sequences; a multi-dimensional feature extraction module, used to extract the dynamic sway trajectory of the pressure center and multi-degree-of-freedom posture angle parameters within the independent gait cycle sequences, performing envelope extraction and vector decomposition on the dynamic sway trajectory of the pressure center, and reconstructing the obtained... The system employs a gait compensation fusion module to extract gait periodicity features by analyzing the temporal phase difference of multi-degree-of-freedom attitude angle parameters, performing multi-dimensional dimensionality reduction and splicing of the foot support envelope area and trajectory offset features to construct a foot pressure imbalance spatial vector characterizing bilateral dynamic differences. It also extracts gait periodicity features to construct adaptive kinematic penalty constraints, performs nonlinear weighted modulation on the foot pressure imbalance spatial vector, and generates multi-modal gait compensation fusion parameters through multi-dimensional heterogeneous signal conversion. Finally, a deflection angle calculation module inputs the multi-modal gait compensation fusion parameters into a preset geometric shape mapping space, performs inverse spatial coordinate analysis of the parameters through a regression topology mechanism, and outputs the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial calculated deflection angle.

[0049] In this implementation scheme, the data synchronization acquisition module, acting as the underlying hardware driver and primary data scheduling center, is responsible for acquiring multi-source physical measurement variables and dividing the time domain into basic units. This module controls the underlying scanning timing of the flexible electrode array and the sampling period of the inertial measurement chip to drive the front-end sensing hardware to acquire the plantar pressure distribution matrix and three-dimensional attitude sequence. Subsequently, the communication microcontroller within the module calls the clock synchronization instruction of the Bluetooth protocol stack to forcibly bind the data streams generated by the two independent hardware sets to the same global timestamp. In this time coordinate system, the extreme value detection logic unit within the module scans the vertical acceleration sequence in real time, locking specific trough positions as the start and end boundaries of physical actions, and precisely cutting and packaging the continuous heterogeneous data stream into independent gait cycle data blocks, pushing them to the next-level processing link. Combined with... Figure 5As shown, this data synchronization acquisition module is implemented based on the architecture of a wearable gait information acquisition system. The system hardware mainly consists of a power supply module, a data acquisition module, a microcontroller (such as an ESP32 chip), and a human-computer interaction module (a mobile phone host computer). The data acquisition circuit synchronously connects the plantar pressure sensor (i.e., a flexible electrode array) and the inertial sensor located at the measurement end. Under the stable power supply provided by the step-down circuit, the data acquisition module acquires the raw signals of foot pressure and three-dimensional posture in real time and transmits them to the microcontroller. The microcontroller uses its internal clock to uniformly timestamp the multi-source signals, and then uses the Bluetooth Low Energy (BLE) protocol to wirelessly transmit the aligned multi-variable joint data sequence to the human-computer interaction module. This hardware architecture diagram clearly shows the complete physical link of electrical signals from front-end physical sensing, convergence, to wireless transmission, providing underlying hardware support for the accurate spatiotemporal synchronization of multi-source physical measurement variables.

[0050] The multidimensional feature extraction module, serving as the system's geometric and kinematic analysis execution unit, receives the independent gait cycle sequence output from the higher-level module. Within this module, the dynamics processing thread performs a weighted traversal of the pressure matrix coordinate points within the sequence, calculates the dynamic swaying trajectory of the pressure center, and invokes a two-dimensional edge-finding algorithm and coordinate axis projection instructions to calculate the plantar support envelope area and trajectory offset vector values. Simultaneously, the kinematics processing thread, running synchronously within the module, reads Euler angle data from the posture sequence, extracts the posture change curves of the left and right lower limbs, and uses the digital signal processing unit to perform temporal cross-correlation operations to find peak delays, transforming the temporal phase difference of the two-sided curves into gait cycle symmetry features. This module converts the original array electrical signal into independent quantized parameters with specific physical representations.

[0051] The gait compensation fusion module is responsible for the data concatenation and dimensional transformation of multidimensional heterogeneous signals in the system. The module's memory allocation unit extracts the envelope area and bias features, combines them with the longitudinally distributed asymmetric gradient according to the system's preset orthogonal dimension address allocation rules, and maps them in memory to construct a high-dimensional foot pressure imbalance space vector. Next, the module's floating-point unit reads the gait periodicity symmetry features and calculates the time-domain difference between it and the pre-stored ideal steady-state benchmark to generate penalty constraints. Using the discrete dynamic weights and kinematic compensation coefficients stored in the module's registers, the floating-point unit performs matrix scaling and scalar scaling instructions on the foot pressure imbalance space vector and penalty constraints, reconstructing two types of physically different parameters into a unified multimodal gait compensation fusion parameter at the underlying level.

[0052] The deflection angle calculation module, serving as the system's terminal topology inversion calculation unit, pre-loads regression analysis model weight parameters generated from prior feature sets into its internal non-volatile memory. After receiving gait compensation fusion parameters, this module loads these input parameters into a preset geometric shape mapping space via its internal inference engine. The inference engine directly calls the stored weight coefficient matrix and intercept bias parameters to perform corresponding matrix multiplication and bias addition operations on the fusion parameters, completing the inverse analytical calculation of spatial coordinates. This module performs low-level geometric reconstruction on the compensation feature data fused from the front end, ultimately outputting the spatial geometric physical quantity representing the target object's entity structure—the three-dimensional spatially calculated deflection angle—directly at the system's output bus.

[0053] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0057] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0058] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A computer-executed method for early screening of scoliosis based on continuous monitoring of plantar pressure and movement posture, characterized in that, Includes the following steps: S1. Collect multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time sequence signal under continuous motion state; Based on the Bluetooth underlying protocol, the timestamps of the two are aligned, the vertical acceleration extreme value features of the three-dimensional attitude timing signal are extracted, and the multivariate joint data sequence after timestamp alignment is divided into continuous independent gait cycle sequences. S2. Extract the dynamic swaying trajectory of the pressure center and the multi-degree-of-freedom attitude angle parameters within the independent gait cycle sequence. Perform envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center to reconstruct and obtain the foot support envelope area and trajectory offset features. Analyze the temporal phase difference of the multi-degree-of-freedom attitude angle parameters to extract the gait cycle symmetry features. S3. Multidimensional reduction and splicing of the foot support envelope area and trajectory bias features are performed to construct a foot pressure imbalance space vector representing the bilateral dynamic differences; gait period symmetry features are extracted to construct adaptive kinematic penalty constraints, and nonlinear weighted modulation is performed on the foot pressure imbalance space vector. Multimodal gait compensation fusion parameters are generated through multidimensional heterogeneous signal conversion. S4. Input the multimodal gait compensation fusion parameters into the preset geometric shape mapping space, and perform inverse spatial coordinate analysis on the parameters through the regression topology mechanism to output the spatial geometric physical quantity of the target object, namely the three-dimensional spatial measurement deflection angle.

2. The early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture executed by a computer according to claim 1, characterized in that: The specific process for acquiring multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional attitude time-series signal under continuous motion is as follows: The pressure resistance change is sensed at the intersection of row and column electrodes in the flexible electrode array. The multi-channel discrete voltage signal is obtained through a multiplexed array and analog-to-digital conversion channel. The signal is then amplified and processed by cascade to generate a sequence of plantar pressure distribution matrix. The three-dimensional acceleration vector and three-dimensional angular velocity vector under motion are captured by an inertial measurement chip, and the three-dimensional attitude timing signal is generated in real time by the attitude calculation matrix.

3. The early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture executed by a computer according to claim 1, characterized in that: Based on the Bluetooth underlying protocol, the timestamps of both are aligned, the vertical acceleration extreme value features of the three-dimensional attitude timing signal are extracted, and the multivariate joint data sequence after timestamp alignment is divided into continuous independent gait periodic sequences. The specific process is as follows: The main control terminal synchronously sends the underlying broadcast wireless clock reference to the bilateral foot end acquisition nodes. The bilateral foot end acquisition nodes inject a global timestamp into the data packet header according to the wireless clock reference, so as to align the timestamps of the bilateral foot plantar pressure distribution matrix sequence and the bilateral three-dimensional attitude timing signal in the same time domain coordinate system. Local minima and local maxima of vertical acceleration fluctuating with time in the three-dimensional attitude timing signal after timestamp alignment are extracted, and the data interval between two adjacent local minima is defined as an independent gait cycle sequence.

4. The early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture executed by a computer according to claim 1, characterized in that: The specific process of extracting the dynamic swaying trajectory of the pressure center and multi-degree-of-freedom attitude angle parameters within an independent gait cycle sequence, performing envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center, and reconstructing to obtain the plantar support envelope area and trajectory offset features is as follows: Calculate the force-weighted coordinates of each single-point array sensor unit in an independent gait cycle sequence to generate a dynamic swaying trajectory of the pressure center; Analyze the three-dimensional attitude time-series signal within an independent gait period sequence and extract multi-degree-of-freedom attitude angle parameters; Extract the maximum geometric outer contour boundary of the dynamic swing trajectory of the pressure center in the two-dimensional horizontal plane, and calculate the area of ​​the closed region enclosed by the maximum geometric outer contour boundary as the foot support envelope area. The dynamic swing trajectory of the pressure center is decomposed by projection along the positive direction of motion and the horizontal and vertical directions. The maximum offset displacement vector of the dynamic swing trajectory of the pressure center deviating from the central axis of the standard motion is calculated, and the trajectory offset feature is generated.

5. A computer-executed early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture, as described in claim 1 or 4, characterized in that: The specific process of analyzing the temporal phase difference of multi-degree-of-freedom attitude angle parameters to extract gait periodic symmetry features is as follows: Obtain the time-series curves of the first multi-degree-of-freedom attitude angles corresponding to the independent gait cycle sequence of the left foot, and the time-series curves of the second multi-degree-of-freedom attitude angles corresponding to the independent gait cycle sequence of the right foot, respectively. The time-domain cross-correlation function is performed on the first multi-degree-of-freedom attitude angle timing curve and the second multi-degree-of-freedom attitude angle timing curve to extract the timing phase difference between the two sets of timing curves at the peak position; The timing phase difference is input into the symmetry mapping function to calculate the overlap degree, thereby generating the gait periodic symmetry feature.

6. The method for early screening of scoliosis based on continuous monitoring of plantar pressure and movement posture executed by a computer according to claim 1, characterized in that: The specific process of constructing a spatial vector representing the foot pressure imbalance characteristic of bilateral dynamic differences by multidimensional reduction and splicing together the plantar support envelope area and trajectory bias features is as follows: The foot support envelope area and trajectory offset features are subjected to dimensionless normalization processing by pre-set benchmark normal physical geometric quantities to obtain normalized area parameters and normalized offset parameters. Extract the anterior-posterior distribution ratio features of the bilateral plantar pressure distribution matrix sequence to generate a longitudinally distributed asymmetric gradient; The vertically distributed asymmetric gradient, normalized area parameter, and normalized bias parameter are mapped to mutually orthogonal feature dimensions. Tensor splicing is then performed to construct a multidimensional feature space, and the foot pressure imbalance space vector is output.

7. A computer-executed early screening method for scoliosis based on continuous monitoring of plantar pressure and movement posture, as described in claim 1 or 6, characterized in that: The specific process of extracting gait periodic symmetry features to construct adaptive kinematic penalty constraints, performing nonlinear weighted modulation on the foot pressure imbalance space vector, and generating multimodal gait compensation fusion parameters through multidimensional heterogeneous signal conversion is as follows: Extract the preset ideal steady-state extreme value benchmark, calculate the time-domain difference deviation between the ideal steady-state extreme value benchmark and the gait period symmetry feature, and construct adaptive kinematic penalty constraints; Retrieve the discrete dynamic weight coefficients and kinematic compensation coefficients preset within the feature fusion module; The dynamic characteristic components are obtained by performing spatial matrix scaling on the foot pressure imbalance space vector through discrete dynamic weighting coefficients. The kinematic compensation components are obtained by performing positive scalar scaling on the adaptive kinematic penalty constraints using kinematic compensation coefficients. Heterogeneous feature cascade fusion is performed on the dynamic feature components and kinematic compensation components to generate multimodal gait compensation fusion parameters.

8. The method for early screening of scoliosis based on continuous monitoring of plantar pressure and movement posture executed by a computer according to claim 1, characterized in that: The multimodal gait compensation fusion parameters are input into a preset geometric shape mapping space. The parameters are then analyzed in reverse spatial coordinates through a regression topology mechanism to output the spatial geometric physical quantity of the target object, i.e., the three-dimensional spatial measurement of the deflection angle. The specific process is as follows: Extract a training feature set containing a scalar of the calibration deflection angle and multi-source prior feature parameters; Based on the training feature set, a minimum mean square error regression analysis model is constructed in a multidimensional feature space to generate a geometric shape mapping space. The multimodal gait compensation fusion parameters are activated and projected into a regression prediction analysis model within the geometric shape mapping space; Extract the weight coefficient matrix and intercept bias parameters from the regression prediction analysis model, and perform inverse spatial coordinate analysis on the multimodal gait compensation fusion parameters through the weight coefficient matrix and intercept bias parameters to output the three-dimensional spatial calculation of the deflection angle.

9. A scoliosis early screening system based on continuous monitoring of plantar pressure and movement posture, used to execute the early scoliosis early screening method based on continuous monitoring of plantar pressure and movement posture executed by a computer as described in any one of claims 1-8, characterized in that, include: The data synchronization acquisition module is used to acquire multi-source physical measurement variables of plantar pressure distribution matrix sequence and three-dimensional posture time sequence signal under continuous motion state; Based on the Bluetooth underlying protocol, the timestamps of the two are aligned, the vertical acceleration extreme value features of the three-dimensional attitude timing signal are extracted, and the multivariate joint data sequence after timestamp alignment is divided into continuous independent gait periodic sequences. The multidimensional feature extraction module is used to extract the dynamic swaying trajectory of the pressure center and the multi-degree-of-freedom attitude angle parameters within an independent gait cycle sequence. It performs envelope extraction and vector decomposition on the dynamic swaying trajectory of the pressure center to reconstruct and obtain the plantar support envelope area and trajectory offset features. It also analyzes the temporal phase difference of the multi-degree-of-freedom attitude angle parameters to extract the gait cycle symmetry features. The gait compensation fusion module is used to perform multidimensional dimensionality reduction and splicing of the plantar support envelope area and trajectory bias features to construct a foot pressure imbalance space vector that characterizes the bilateral dynamic differences; extract gait periodic symmetry features to construct adaptive kinematic penalty constraints, perform nonlinear weighted modulation on the foot pressure imbalance space vector, and generate multimodal gait compensation fusion parameters through multidimensional heterogeneous signal conversion; The deflection angle calculation module is used to input the multimodal gait compensation fusion parameters into the preset geometric shape mapping space, and perform inverse spatial coordinate analysis on the parameters through the regression topology mechanism to output the spatial geometric physical quantity of the target object, namely the three-dimensional spatial calculation deflection angle.

Images