A method for early warning of dynamic balance posture of lower limbs and center of gravity based on a body measurement integrated machine

Abstract

The present application provides a lower limb and center of gravity dynamic balance posture warning method based on a physical fitness all-in-one machine, including: obtaining the pelvic tilt angle, arch height and center of gravity position in a standard standing posture as individualized baseline data, and generating a dynamic balance reference value reflecting the individual's normal balance ability by statistically analyzing the baseline data; on the basis of the dynamic balance reference value, combining the center of gravity forward shift trend and the lower limb muscle compensation pattern, calculating the balance offset index in the standing posture change through a time series analysis algorithm to obtain a dynamic balance risk assessment result; if the dynamic balance risk assessment result exceeds the danger threshold, clustering analysis is performed on the correlation parameters of the pelvic tilt angle and arch height, grouping the correlation parameters, and identifying the potential causes of balance ability degradation.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a physical testing all-in-one machine. Background Art

[0002] The study of dynamic balance is of vital importance in the fields of human movement science and health management. It directly impacts the safety of individuals in daily activities, the optimization of athletic performance, and the prevention and control of fall risks in the elderly. With the aging population and the increase in chronic diseases, the use of technology to accurately assess and predict balance deterioration has become a crucial issue for improving quality of life and reducing the burden of healthcare. Existing methods often rely on traditional body measurement equipment or subjective observation. While these methods can provide a certain degree of balance, they suffer from insufficient accuracy, limited data, and a lack of personalized analysis. These limitations make it difficult to capture subtle dynamic features of postural changes and provide targeted early intervention recommendations. Against this backdrop, key challenges in the research field have become increasingly prominent, particularly the interaction between pelvic tilt angle, arch height, and static center of gravity position. Current technologies for analyzing these factors often struggle to accurately quantify the dynamic effects of pelvic tilt on lumbar curvature and anterior center of gravity, let alone reveal the complex adjustments in compensatory muscle activation patterns in the lower limbs. In addition, how to accurately associate the parameters of pelvic tilt and arch height through three-dimensional bone reconstruction algorithms, and set risk thresholds based on individualized baseline data, remains an unsolved technical problem. The existence of these challenges has led to a lack of real-time monitoring and scientific early warning basis for the posture change process in dynamic balance assessment, limiting the effectiveness of the technology in practical applications. Therefore, how to use the all-in-one body measurement device to accurately identify the associated parameters of pelvic tilt angle and arch height when the subject switches from the standard standing posture to the daily habitual standing posture, and analyze its impact on the center of gravity position and lower limb muscle activation pattern through three-dimensional bone reconstruction algorithms, and then establish a dynamic balance risk warning threshold based on individualized data, has become a key issue that needs to be overcome in this study. Summary of the Invention

[0003] The present invention provides a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one machine, which mainly includes:

[0004] The three-dimensional skeletal data of the subjects in standard and habitual standing postures were obtained. The built-in sensors of the all-in-one body measurement machine were used to collect the pelvic tilt angle and arch height parameters. The initial skeletal model was generated using a three-dimensional skeletal reconstruction algorithm to obtain the correlation parameters between the pelvis and the arch of the foot.

[0005] Extract the pelvic tilt angle change data from the initial skeletal model and combine it with the arch height parameter to determine the center of gravity position. Calculate the center of gravity offset caused by the pelvic position change and determine the center of gravity forward shift trend based on the time-varying trend of the offset.

[0006] Predicting changes in stance based on the forward center of gravity shift. When stance changes, acquiring and processing the electromyographic signals of key lower limb muscle groups to derive the lower limb muscle activation pattern. Combining the relative magnitude and coordination of muscle activation within the activation pattern, the compensatory pattern of lower limb muscles is determined, and the amplitude and distribution characteristics of compensatory muscle activation are determined.

[0007] Obtain the pelvic tilt angle, arch height, and center of gravity position in a standard standing posture as individualized baseline data. Statistical analysis of the baseline data is performed to generate a dynamic balance benchmark value that reflects the individual's normal balance ability.

[0008] Based on the dynamic balance baseline value, combined with the center of gravity forward shift trend and lower limb muscle compensation pattern, the balance deviation index in the standing posture change is calculated through the time series analysis algorithm to obtain the dynamic balance risk assessment result.

[0009] If the dynamic balance risk assessment result exceeds the danger threshold, cluster analysis is performed on the correlation parameters between the pelvic tilt angle and the arch height, and the correlation parameters are grouped to identify the potential causes of balance deterioration.

[0010] Based on the results of dynamic balance risk assessment, corresponding warning signals are generated, and the warning signals are correlated and mapped with the real-time change trends of the pelvic tilt angle and arch height. Combined with the potential causes of degeneration, balance degradation warning information is generated and fed back in real time through the physical measurement integrated system.

[0011] Furthermore, the three-dimensional skeletal data of the subject in the standard standing posture and the habitual standing posture are obtained, the pelvic tilt angle and the arch height parameters are collected by the built-in sensor of the body measurement all-in-one machine, and the initial skeletal model is generated by using the three-dimensional skeletal reconstruction algorithm to obtain the associated parameters of the pelvis and the arch of the foot, including:

[0012] The three-dimensional coordinate data of the skeletal feature points are collected, and the three-dimensional coordinate data are subjected to noise reduction processing by a median filter to obtain filtered feature point position data; the pressure center trajectory curve and the pressure value of the arch area are calculated based on the filtered feature point position data and the plantar pressure distribution data collected by the pressure sensor array; the angular velocity data are collected by a three-axis gyroscope sensor array arranged in the pelvic area, and the pelvic spatial posture angle is obtained based on the integration operation of the angular velocity data; the plantar surface equation is established based on the pressure value of the arch area, the longitudinal arch height and the transverse arch height parameters are calculated from the plantar surface equation, and a group of geometric constraint equations is established based on the pelvic spatial posture angle and the longitudinal arch height and the transverse arch height parameters, and the constraint equations are solved by the least squares method to obtain the positional relationship parameters between the pelvis and the arch.

[0013] Furthermore, the method extracts the pelvic tilt angle change data from the initial skeletal model, combines it with the arch height parameter to determine the center of gravity position, calculates the center of gravity offset caused by the pelvic position change, and determines the center of gravity forward shift trend based on the change trend of the offset over time, including:

[0014] Receive pelvic angle data sent by the skeletal surface equation, and process the pelvic angle data with a Kalman filter to obtain a pelvic tilt angle sequence; obtain pressure distribution data collected by a plantar pressure sensor array based on the pelvic tilt angle sequence, and calculate a pressure center point coordinate sequence and a foot arch height value through the pressure distribution data; use the pelvic tilt angle sequence and the foot arch height value to establish a human body center of mass distribution equation, and obtain the initial coordinates of the whole body center of gravity through linear superposition calculation; for the initial coordinates of the whole body center of gravity, use Euler angles to describe the three-dimensional posture changes of the pelvis, and calculate a center of gravity displacement vector sequence through a rigid body kinematic equation, and obtain a horizontal plane displacement field distribution through least squares fitting of the center of gravity displacement vector sequence.

[0015] Furthermore, the posture change is predicted based on the center of gravity forward shift trend. When the posture changes, the electromyographic signals of key lower limb muscle groups are acquired and processed to obtain the lower limb muscle activation pattern. The relative size and coordination of the activation degree of each muscle in the activation pattern are combined to determine the lower limb muscle compensation pattern, and the amplitude and distribution characteristics of the compensatory muscle activation are obtained, including:

[0016] A predicted spatial trajectory diagram is constructed based on the center of gravity displacement sequence data, and the deviation value between the center of gravity projection position and the support base area edge threshold is obtained from the predicted spatial trajectory diagram; the potential signals of the muscle groups around the lower limb joints are collected by surface electromyography sensors, and the potential signals are filtered using a center frequency bandpass filter to obtain a filtered electromyography signal waveform diagram; the filtered electromyography signal waveform diagram is subjected to wavelet transformation to obtain a wavelet coefficient matrix, and the signal mutual correlation coefficient matrix between the muscle groups is calculated from the wavelet coefficient matrix, and the muscle group synergy eigenvector is obtained by eigenvalue decomposition of the mutual correlation coefficient matrix; the Euclidean distance value is calculated based on the muscle group synergy eigenvector and the standard eigenvector constructed from standard anthropometric data, and a compensation measurement function is established from the Euclidean distance value to obtain a thermal distribution map of the muscle compensation degree.

[0017] Furthermore, the pelvic tilt angle, arch height, and center of gravity position in the standard standing posture are obtained as individualized baseline data, and a dynamic balance benchmark value reflecting the individual's normal balance ability is generated by statistically analyzing the baseline data, including:

[0018] Acquire pelvic marker point data collected by an optical tracker, and obtain a pelvic tilt angle sequence through three-dimensional spatial coordinate transformation based on the marker point data; receive a pressure distribution map collected by a plantar pressure sensor array, and calculate an arch support index based on the pressure distribution map, wherein the arch support index is obtained by calculating the longitudinal arch height value and the transverse arch height value of the arch area pressure curve; establish a center of mass position calculation equation based on the pelvic tilt angle sequence and the arch support index, and obtain the three-dimensional coordinates of the center of gravity from the calculation equation, wherein the calculation equation includes the segment mass ratio and the spatial coordinate transformation matrix; perform kernel density estimation on the three-dimensional coordinates of the center of gravity to obtain a center of gravity displacement envelope, extract a displacement direction and amplitude feature sequence from the center of gravity displacement envelope, and obtain a balance ability benchmark value based on the feature sequence using a Bayesian estimation method.

[0019] Furthermore, based on the dynamic balance baseline value, combined with the center of gravity forward shift trend and lower limb muscle compensation pattern, the balance deviation index in the standing posture change is calculated through a time series analysis algorithm to obtain the dynamic balance risk assessment results, including:

[0020] A time-frequency spectrum is obtained by performing wavelet transform on the center of gravity displacement sequence, and the dominant frequency and amplitude parameters of the center of gravity movement are extracted from the time-frequency spectrum to construct a balance state description vector. A recursive neural network is used to process the balance state description vector, and the center of gravity displacement rate curve is calculated through a sliding time window to obtain the balance offset. Envelope extraction is performed on the lower limb electromyographic signal, and the compensatory activation intensity is calculated based on the envelope amplitude. The temporal correlation between muscle groups is solved through the cross-correlation function to obtain the compensatory synergy feature. A feature fusion function is established based on the balance state description vector and the compensatory synergy feature, and the fused feature is normalized. The trend item of the risk feature sequence is extracted through time series analysis. If the trend slope exceeds the preset threshold, the balance risk level is determined.

[0021] Furthermore, if the dynamic balance risk assessment result exceeds the danger threshold, cluster analysis is performed on the parameters associated with the pelvic tilt angle and the arch height, and the associated parameters are grouped to identify potential causes of balance deterioration, including:

[0022] A risk status curve is constructed based on the dynamic balance risk assessment data, and the piecewise linear mapping value is calculated through the risk status curve to obtain the danger threshold interval parameter; the density clustering algorithm is used to process the pelvic tilt angle data, the fluctuation amplitude and change period are calculated from the angle data sequence, and the pelvic tilt abnormality parameter is obtained through the Euclidean distance metric; the pressure distribution surface is calculated based on the plantar pressure sensor array data, and the arch boundary curve is extracted from the pressure distribution surface using the region growing algorithm to obtain the arch morphology abnormality parameter; a parameter correlation matrix is ​​constructed for the pelvic tilt abnormality parameter and the arch morphology abnormality parameter, and the hierarchical clustering method is used to group and cluster the correlation matrix, and the dominant factor of balance ability degradation is obtained through intra-group variance calculation.

[0023] Furthermore, based on the dynamic balance risk assessment results, a corresponding warning signal is generated, and the warning signal is correlated with the real-time change trend of the pelvic tilt angle and the arch height. In combination with the potential causes of degradation, balance degradation warning information is generated and fed back in real time through the integrated physical test system, including:

[0024] A dynamic threshold range is obtained based on the distribution of historical data, and a fluctuation feature sequence is extracted from the dynamic threshold range to obtain a first warning parameter; the first warning parameter is predictively analyzed through a recursive neural network, and if the rate of decrease of the predictive analysis score exceeds a preset threshold, a first warning indicator of a corresponding level is generated; a support area change rate curve is obtained based on the plantar pressure sensor array, and a trend feature sequence is extracted from the change rate curve to obtain a second warning parameter; a mapping relationship table is established using the first warning indicator and the second warning parameter, and the warning prompt content is matched through the mapping relationship table to generate a warning data packet to the display terminal.

[0025] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0026] The present invention discloses a method for early warning of the dynamic balance posture of the lower limbs and the center of gravity based on a physical measurement all-in-one machine. The method generates an initial skeletal model and extracts key parameters by collecting three-dimensional skeletal data, pelvic tilt angle and arch height parameters of the subject in different standing postures. Combining the changes in the center of gravity position and the lower limb muscle activation pattern, the present invention determines the center of gravity forward shift trend and the lower limb muscle compensation pattern. By establishing individualized baseline data and dynamic balance reference values, the present invention calculates the balance offset index to achieve dynamic balance risk assessment. When the risk exceeds the threshold, the present invention identifies the potential causes of the deterioration of balance ability and generates corresponding early warning information. This method can monitor and evaluate individual balance ability in real time, providing a scientific basis for preventing the risk of falling and improving standing balance. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention is a flowchart of a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a physical fitness tester.

[0028] Figure 2 It is a schematic diagram of a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one machine according to the present invention. DETAILED DESCRIPTION

[0029] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] like Figure 1 -2. In this embodiment, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one device may specifically include:

[0031] S101. Collect three-dimensional skeletal data of the subject in the standard standing posture and the habitual standing posture through built-in sensors, use a three-dimensional skeletal reconstruction algorithm to generate an initial skeletal model and extract the correlation parameters of the pelvis and the arch of the foot, and process the collected data to ensure accuracy.

[0032] In this embodiment of the present invention, the integrated body measurement machine is equipped with multiple sensors to obtain skeletal and pressure data from subjects in different standing postures. Specifically, an optical tracker is used to collect 3D coordinate data of skeletal feature points on the subject's body surface. Noise reduction techniques are then incorporated into the data processing to improve measurement accuracy.

[0033] S1011. Obtain the three-dimensional coordinate data of the subject's skeletal feature points in the standard standing posture and the habitual standing posture through an optical tracker, synchronously record the motion trajectory of the marker points using a high-speed camera, and use a median filter to perform noise reduction on the coordinate data to obtain smooth feature point position data. Then, align the data through the rigid body transformation method to generate an initial skeletal model.

[0034] In practice, reflective markers with a diameter of 10 mm were placed on the subject's body surface. Anatomical landmarks such as the anterior superior pelvic spine and the greater trochanter of the femur were selected as marker locations. The optical tracker consisted of four high-speed cameras with a sampling frequency of 100 Hz. A median filter with a window size of 5 frames was used to eliminate jitter caused by skin and soft tissue deformation. The registration process used a rigid body transformation method to ensure that the feature point data was aligned with the anatomical structure template. A thin plate spline deformation function was then used to generate a personalized bone surface equation, providing a basis for subsequent analysis.

[0035] S1012. Use the pressure sensor array built into the all-in-one body measurement machine to collect plantar pressure distribution data, calculate the pressure center trajectory curve and the pressure value of the arch area, and calculate the pelvic spatial posture angle and forward tilt angle value based on the angular velocity data collected by the three-axis gyroscope sensor array. Combined with the arch height parameters, a set of geometric constraint equations is established, and the spatial position relationship parameters of the pelvis and the arch are obtained by solving them using the least squares method.

[0036] Specifically, the pressure sensor array consists of 16×16 sensing units, each measuring 10×10 mm, with a measurement range of 0 to 600 kPa and a sampling frequency of 50 Hz. When the displacement of the center of pressure trajectory is less than 5 mm within 20 consecutive frames, the stance is considered stable and the current pressure data is recorded. The sole of the foot is divided into three regions: the forefoot, midfoot, and heel. The arch boundaries are identified and pressure values ​​are extracted by the difference between adjacent sensing units. A three-axis gyroscope sensor is placed in the pelvis, with a measurement range of ±1000 degrees per second and a sampling frequency of 200 Hz. Euler angles are obtained through integration and converted to a gravity coordinate system to calculate the pelvic tilt angle. Plantar pressure data are processed using cubic spline interpolation with a node spacing of 5 mm to generate the plantar surface equation, from which the longitudinal and transverse arch height parameters are extracted. A set of geometric constraint equations was established with the pelvic center of gravity as the coordinate origin. The solution results showed that for every 1 degree change in the pelvic tilt angle, the longitudinal arch height changed by 0.8 mm and the transverse arch height changed by 0.5 mm, reflecting the biomechanical relationship between the two.

[0037] In an embodiment of the present invention, the pelvic tilt angle and arch height parameters obtained through the above steps can provide accurate input data for subsequent center of gravity offset and muscle activation pattern analysis. The collaborative work of the optical tracker and the pressure sensor ensures the stability and reliability of data acquisition, and the application of the three-dimensional bone reconstruction algorithm enables the correlation parameters of the pelvis and arch to be presented in a quantitative form, laying the foundation for dynamic balance assessment. It is understandable that the specific configuration of the sensor and the algorithm parameters can be adjusted by technical personnel according to actual needs without excessive restrictions.

[0038] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a physical fitness tester is provided. The following specific steps are used to achieve accurate analysis and evaluation of dynamic balance ability.

[0039] S102. Extract the pelvic tilt angle change characteristics from the initial skeletal model and determine the center of gravity position in combination with the arch height data. By analyzing the impact of pelvic posture changes on the center of gravity offset, calculate the time series trend of the offset and determine the direction of the center of gravity forward movement.

[0040] In this embodiment, based on a pre-built initial skeletal model, multidimensional data collected by sensors is used to analyze the dynamic relationship between the pelvis and the arch of the foot, thereby deducing the changing patterns of the center of gravity. During implementation, key parameters are first extracted from the skeletal model and then smoothed to ensure data accuracy and stability.

[0041] S1021. Obtain the pelvic tilt angle sequence from the initial skeletal model and smooth it through the Kalman filter. Calculate the pressure center coordinates and the arch height value based on the pressure distribution data collected by the plantar pressure sensor array. Use these parameters to establish a human body center of mass distribution model, and determine the initial coordinates of the whole body center of gravity through the linear superposition method.

[0042] Specifically, the pelvic anterior tilt angle series was extracted from a skeletal surface equation, which constructs a three-dimensional spatial model based on anatomical landmarks such as the anterior superior iliac spine and the posterosuperior pelvic spine. A Kalman filter was used to optimize the angle data, with a measurement noise covariance of 0.5 degrees and a process noise covariance of 0.1 degrees. This reduced the fluctuation range of the filtered data from the original ±3 degrees to ±0.8 degrees, ensuring smoothness of the angle series. The plantar pressure sensor array employed a 16×16 grid layout, with each element measuring 1 square centimeter. The measurement range covered 0 to 600 kPa. The coordinates of the pressure center point were calculated using a weighted average method, with the weights determined by the pressure values ​​of each sensor element. Arch height values ​​were obtained by cubic spline interpolation of the longitudinal arch pressure data with a 5-mm node spacing. The resulting smooth curve accurately reflects the arch morphology. The human center of mass distribution model comprehensively considers the influence of pelvic tilt on trunk position and the role of arch height in lower limb support. The mass distribution of each segment is based on anthropometric standards, for example, the trunk accounts for 43% and the lower limbs account for 32%. The center of mass coordinates of each segment are calculated through linear superposition, and a coordinate system is established with the center of the pelvis as the origin. Finally, a weighted average method is used to determine the initial coordinates of the entire body center of mass, laying the foundation for subsequent offset analysis.

[0043] S1022. For the initial coordinates of the center of gravity of the whole body, the Euler angle is used to represent the posture change of the pelvis in three-dimensional space. The center of gravity displacement vector sequence caused by the change of the pelvic position is calculated through the rigid body kinematic equation. The least squares fitting is performed on the sequence to obtain the displacement field distribution on the horizontal plane. The dominant direction of the displacement field is extracted through Fourier transform to determine the forward movement trend of the center of gravity.

[0044] In this step, Euler angles are obtained by integrating triaxial gyroscope data to describe the pelvic rotation angles in the sagittal, coronal, and transverse planes. The rigid body kinematic equations use the pelvis as the rigid core and calculate the dynamic effects of changes in its posture on the center of gravity position. For example, a 5-degree increase in pelvic anterior tilt causes the trunk center of gravity to shift forward by approximately 20 mm. The center of gravity displacement vector sequence records the spatial trajectory of the center of gravity during the sampling time, with a sampling frequency set to 100 Hz. The vector sequence is fitted using the least squares method to generate a displacement field curve on the horizontal plane, with a fitting error controlled within 2 mm to ensure accuracy. A Fourier transform further analyzes the frequency characteristics of the displacement field, extracting the direction of movement corresponding to the component with the largest amplitude. If the angle between this direction and the sagittal plane is less than 30 degrees, the center of gravity is considered to be shifting forward. Furthermore, the stability boundary is calculated by considering the effect of changes in arch height on the base of support area. For example, for every 100 square centimeters decrease in base area, the allowable forward center of gravity shift decreases by 15 mm, providing a quantitative basis for dynamic balance risk assessment.

[0045] In an embodiment of the present invention, by smoothing the pelvic tilt angle sequence and accurately calculating the plantar pressure data, the center of gravity shift characteristics in posture changes can be effectively captured. Compared with traditional methods that rely only on a single parameter, this method significantly improves the accuracy of center of gravity position derivation through the fusion analysis of multi-dimensional data. For example, the Kalman filter not only reduces the impact of measurement noise, but also provides stable input data for subsequent trend judgment. At the same time, the application of rigid body kinematic equations makes the calculation of center of gravity displacement more physically based, reflecting the biomechanical relationship between pelvic posture adjustment and lower limb support.

[0046] In order to further improve the comprehensiveness of the analysis, this step also takes into account the indirect effect of arch height on support stability. The support base area calculated by the human body center of mass distribution model can dynamically adjust the allowable range of center of gravity offset. For example, when the increase in arch height causes the base area to decrease, the system will update the stability boundary in real time to ensure that the judgment of the center of gravity forward trend is more in line with individual characteristics. It is understandable that the specific filtering parameters and fitting algorithms can be adjusted according to the actual application scenario. For example, in a high dynamic environment, the frequency resolution of the Fourier transform can be increased to capture more subtle displacement changes.

[0047] In an embodiment of the present invention, the determination of the center of gravity's forward shift trend provides a key basis for predicting subsequent changes in standing posture. By analyzing the distribution of the displacement field, users can intuitively understand the dynamic evolution of the body's balance state. For example, when standing statically, the natural swing range of the center of gravity position is usually plus or minus 10 mm. When the offset exceeds this range, the system can identify it as a significant posture adjustment. This analysis method based on multi-parameter fusion not only improves the scientific nature of the evaluation, but also provides reliable data support for real-time warnings. The specific fitting algorithm and threshold setting can be further optimized by technical personnel based on the characteristics of the subject group without excessive restrictions.

[0048] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a physical fitness tester is provided, which achieves accurate assessment of lower limb muscle activation and center of gravity changes through the following steps.

[0049] S103. Predict the changes in the subject's standing posture based on the tendency of the center of gravity to shift forward, and extract the muscle activation pattern by collecting and processing the electromyographic signals of the key muscle groups of the lower limbs when the standing posture is adjusted. Further analyze the characteristics of the activation degree and synergistic relationship of each muscle, so as to determine the compensatory pattern of the lower limb muscles and obtain their activation amplitude and distribution characteristics.

[0050] In this embodiment of the present invention, previously calculated center of gravity displacement data, combined with signals collected by surface electromyography sensors, dynamically analyzes stance changes and muscle compensation. This analysis first predicts the direction of stance change using the center of gravity trajectory, and then performs multi-level processing on the electromyography signals to extract key features.

[0051] S1031. Construct a predicted spatial trajectory diagram based on the center of gravity displacement sequence to determine the trend of standing posture changes. Use surface electromyography sensors to collect potential signals of the main muscle groups of the lower limbs and use bandpass filters for preprocessing to obtain filtered electromyography waveforms. Then, use wavelet transform to extract time-frequency features to analyze muscle activation intensity.

[0052] Specifically, a center of gravity displacement sequence is generated using plantar pressure data and pelvic posture parameters. The predicted spatial trajectory is bounded by the support base area, which is approximately 300 to 400 square centimeters during normal standing. When the center of gravity projection is less than 20 mm from the edge of the base, the system identifies a potential unstable state and predicts posture changes based on the tangent direction of the trajectory curve. For example, a displacement of the projection exceeding 10 mm per second indicates significant posture adjustment. Surface electromyography sensors are positioned for key lower limb muscle groups, including the rectus femoris, biceps femoris, tibialis anterior, and gastrocnemius, with a sampling frequency set to 1000 Hz. Raw electromyographic signals are often affected by 50 Hz power frequency interference and baseline drift. Therefore, a bandpass filter with a center frequency of 50 Hz and a passband range of 20 to 450 Hz is used for processing. This filtered signal retains the main frequency components of muscle contraction while effectively suppressing noise. Next, a wavelet transform is applied to the filtered electromyographic waveform, using a 6-layer decomposition using a fourth-order Dobesy wavelet basis function. Wavelet coefficients are extracted to reflect the signal's time-frequency distribution. The root mean square (RMS) values ​​of the wavelet coefficients were calculated using a 200-millisecond window and a 50% overlap to generate time-domain activation curves for each muscle group. The amplitudes of these curves are measured in microvolts and range from 50 to 200 microvolts during normal standing. This process clearly demonstrates the dynamic intensity of muscle contraction and provides reliable data for subsequent synergistic analysis.

[0053] S1032. Extract features from the filtered electromyographic signal, calculate the correlation coefficient matrix between muscle groups and perform eigenvalue decomposition to obtain collaborative eigenvectors, calculate the Euclidean distance value in combination with standard anthropometric data to quantify the degree of compensation, and generate a muscle compensation degree heat map to reflect the activation distribution characteristics.

[0054] In this step, the synergy between muscle groups is quantified using a correlation coefficient matrix. For example, the correlation coefficient between the rectus femoris and tibialis anterior muscles in normal stance is approximately 0.7, indicating strong temporal correlation between them. Eigenvalue decomposition is performed on the coefficient matrix to extract the first three eigenvectors, which explain over 80% of the synergy variance and form the core features of the muscle synergy pattern. Standard anthropometric data, derived from a normal stance electromyographic dataset of 100 healthy adults, contains synergy eigenvectors for each muscle group as a benchmark. The synergy eigenvectors for the current subject are compared with the standard vectors, and the difference is calculated using the Euclidean distance. In normal stance, the distance is less than 0.3, while it can exceed 0.5 when compensation is significant. A compensation metric function is established based on the Euclidean distance to calculate the compensatory activation strength of each muscle group. For example, a compensatory strength value exceeding 2.0 for the tibialis anterior indicates that it is taking on additional balance adjustments. The compensation heatmap uses color coding, with red indicating areas of high compensation and blue indicating near-normal conditions. This intuitively displays spatial distribution characteristics. For example, while the tibialis anterior muscle appears dark red, the adjacent gastrocnemius muscle may be only light blue, indicating a compensation strength below 0.5. This visualization facilitates understanding of regional differences in muscle compensation.

[0055] In this embodiment of the present invention, predicting posture changes through spatial trajectory mapping is crucial for triggering muscle activation analysis. For example, when the center of gravity projection exceeds the edge of the base, the system automatically initiates EMG signal acquisition, ensuring that the data corresponds to real-time posture adjustments. The combination of bandpass filtering and wavelet transform not only improves signal quality but also extracts deeper features of muscle activation. Compared to methods that rely solely on raw signals, this method can more accurately reflect the dynamic process of muscle contraction.

[0056] To further optimize the analysis, this step also considers the impact of the base of support area on standing stability. The outer contour of the plantar surface is measured using a pressure sensor array, and changes in the base area directly affect the threshold for determining center of gravity shift. For example, when the arch collapses, the base area increases, enhancing stability but potentially triggering compensatory adjustments in the lower limb muscles. Time and frequency domain analysis of electromyographic signals can capture subtle changes in these adjustments. For example, the short, high-intensity activation of the biceps femoris at the beginning of a posture change can cause its RMS value to surge from 100 to 300 microvolts.

[0057] In an embodiment of the present invention, the realization of compensation pattern analysis benefits from the integration of multi-level signal processing and feature extraction technology. The calculation of the cross-correlation coefficient matrix reveals the collaborative working mechanism between muscle groups, and the generation of the heat map provides intuitive feedback to the user. For example, when the tibialis anterior muscle is overactivated due to the forward shift of the center of gravity, the red area of ​​its compensatory strength on the heat map prompts the user that there may be insufficient arch support. This data-based visual analysis not only improves the scientific nature of the assessment, but also provides a targeted basis for subsequent intervention measures. It is understandable that the filter parameters and feature extraction algorithms can be adjusted according to actual needs. For example, in a high-noise environment, the number of wavelet decomposition layers can be increased to improve signal resolution.

[0058] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one machine is provided, and the evaluation of individualized balance ability is achieved through the following steps.

[0059] S104. Collect the data of the subject's pelvic tilt angle, arch height, and center of gravity position in a standard standing posture as an individualized baseline, and generate a dynamic balance benchmark value reflecting normal balance ability through multidimensional statistical analysis.

[0060] In an embodiment of the present invention, a personalized balance benchmark is constructed based on multi-source data collected by an optical tracker and a plantar pressure sensor array, combined with spatial transformation and probability estimation methods. Specifically, key parameters are first accurately measured and processed.

[0061] S1041. Use an optical tracker to collect pelvic marker point data and calculate the forward tilt angle sequence. At the same time, obtain the pressure distribution map and extract the arch support index through the plantar pressure sensor array. Based on these parameters, establish a center of mass position calculation model to determine the three-dimensional coordinates of the center of gravity.

[0062] Specifically, the optical tracker uses four high-speed cameras to capture the motion trajectory of 10 mm diameter reflective markers on the anterior and posterior superior iliac spines of the subjects, with a sampling frequency of 100 Hz. The pelvic tilt angle is defined as the angle between the sagittal plane of the pelvis and the direction of gravity, and the normal value is usually in the range of 10 to 15 degrees. The system monitors the rate of change of the angle. When the change is less than 0.5 degrees per second in 20 consecutive frames, it is judged as a steady-state stance. It is then smoothed using a weighted average filter, and the filter window is set to 10 frames to ensure data stability. The plantar pressure sensor array consists of 16×16 units, each unit has an area of ​​1 square centimeter, and the measurement range is 0 to 600 kPa. By setting a pressure threshold of 50 kPa to segment the plantar contour, the support area is calculated to be approximately 280 square centimeters. Cubic spline interpolation is performed on the pressure data of the arch area with a node spacing of 5 mm to generate a smooth curve, from which the highest points of the longitudinal arch and the transverse arch are extracted. They are located at 40% of the foot length and the middle of the metatarsal bone, respectively. The normal longitudinal arch height is 15 to 25 mm, and the transverse arch height is 8 to 12 mm. The arch support index is calculated by the ratio of height value to foot length, reflecting the support capacity of the foot. The center of mass position calculation model is based on the segment parameter method, which divides the human body into the trunk and lower limbs, with mass distribution of 43% and 32% respectively. The center of mass position of each segment is adjusted by the pelvic tilt angle and the arch height. For example, an increase in the tilt angle by 1 degree causes the center of mass of the trunk to move forward by 2 mm, and a decrease in the arch height by 1 mm causes the center of gravity to drop by 0.5 mm. Finally, the three-dimensional coordinates of the center of gravity are generated through spatial coordinate transformation.

[0063] S1042. Perform kernel density estimation on the three-dimensional coordinates of the center of gravity to generate a displacement envelope and extract a displacement feature sequence. Use the Bayesian estimation method to fuse multi-dimensional feature data and calculate a dynamic balance reference value that reflects the individual's balance ability.

[0064] In this step, kernel density estimation uses a Gaussian kernel function to analyze the distribution of center of gravity positions. The kernel bandwidth is set to 5 mm, and the area with a probability density greater than 0.1 is taken as the fluctuation range, forming an elliptical envelope, the major axis of which is usually consistent with the sagittal plane of the human body. Principal component analysis extracts the displacement characteristics within the envelope. The first principal component reflects the displacement in the anterior-posterior direction, with an amplitude usually less than 20 mm. The second principal component reflects the left-right direction, with an amplitude less than 10 mm, and the displacement frequency is between 0.3 and 1 Hz. These features constitute a descriptive function of the dynamic balance state. The displacement amplitude and frequency parameters are fused through Bayesian estimation to generate a comprehensive score ranging from 0 to 100, with higher scores indicating stronger balance ability. This method can effectively quantify the center of gravity stability of an individual in a standard standing posture, providing a reference for subsequent risk assessment.

[0065] In this embodiment of the present invention, the accuracy of baseline data is ensured through the collaborative work of an optical tracker and a pressure sensor. Compared to traditional static measurements, this method more comprehensively reflects the normal performance of an individual's balance ability through dynamic feature extraction and probabilistic analysis. For example, the introduction of the arch support index not only quantifies the foot's support of the center of gravity but also reveals the potential impact of changes in arch morphology on overall stability.

[0066] To improve the practicality of the baseline values, this step also combines steady-state data with dynamic fluctuation characteristics. The envelope generated by kernel density estimation can intuitively demonstrate the natural swing range of the center of gravity in space. For example, when standing normally, the fluctuation range in the front-to-back direction reflects the subject's ability to adapt to gravity, while the smaller amplitude in the left-to-right direction indicates strong lateral stability. Bayesian estimation improves the robustness of the score through probabilistic fusion. For example, when the subject's arch height is low, the system will adjust the score weight based on historical data to emphasize the contribution of lower limb support to balance.

[0067] In the embodiment of the present invention, the process of generating dynamic balance baseline values ​​fully considers the scientific nature of data acquisition and analysis. The high-frequency sampling of the optical tracker ensures the temporal resolution of the pelvic angle, while the spatial resolution of the pressure sensor provides reliable support for the arch parameters. The establishment of the baseline value not only provides a basis for individualized early warning, but also provides a comparison standard for subsequent posture adjustment analysis. The filter window or kernel bandwidth can be adjusted according to the characteristics of the subject group. For example, the steady-state judgment threshold can be appropriately relaxed for the elderly to adapt to their posture fluctuation characteristics.

[0068] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one machine is provided, and a quantitative assessment of dynamic balance risk is achieved through the following steps.

[0069] S105. Based on the dynamic balance baseline value, combined with the center of gravity forward shift trend and lower limb muscle compensation pattern, the balance deviation index in the standing posture change is calculated through a time series analysis algorithm, thereby generating a dynamic balance risk assessment result.

[0070] In this embodiment of the present invention, the center of gravity displacement and electromyographic signal data extracted in the early stage are used to comprehensively evaluate the dynamic changes in the balance state through time-frequency analysis and feature fusion technology. When implementing this, the center of gravity displacement sequence is first decomposed at multiple scales to extract key features.

[0071] S1051. Perform time-frequency analysis on the center of gravity displacement sequence through wavelet transform and construct a balance state description vector. Use recursive neural network to process the displacement data and calculate the balance offset. At the same time, perform envelope extraction on the lower limb electromyographic signals to obtain compensatory activation intensity and synergistic characteristics.

[0072] Specifically, the wavelet transform uses the Dobechy 4th-order wavelet basis function to perform a 6-layer decomposition of the center of gravity displacement sequence, extracting a time-frequency spectrum in the 0.5 to 2 Hz frequency band, which corresponds to the primary frequency range for human posture adjustment. The time-frequency spectrum shows that during normal standing, the dominant frequency of center of gravity movement is approximately 0.8 Hz, with an amplitude typically less than 10 mm. However, during posture adjustment, the frequency may rise to 1.5 Hz, and the amplitude may increase to over 15 mm. The equilibrium state description vector is composed of the dominant frequency, displacement amplitude, and velocity parameters. This vector is processed through a recursive neural network using a long-short-term memory structure. The input layer contains the displacement coordinates and velocity components, and the hidden layer has 64 neurons, which can capture long-term dependencies in the time series. The sliding time window is 1 second long, with a 50% overlap rate. The displacement rate curve is calculated. When the rate exceeds 20 mm per second, it indicates a significant shift in the equilibrium state. Lower limb electromyographic signals were collected from the rectus femoris, biceps femoris, tibialis anterior, and gastrocnemius muscles at a sampling frequency of 1000 Hz. A waveform envelope was generated through rectification and smoothing. The amplitude of the envelope reflects the strength of muscle contraction, ranging from 50 to 200 microvolts during normal standing and potentially rising to 400 microvolts during compensation. A cross-correlation function analyzes the temporal relationship between muscle groups. For example, the correlation coefficient between the rectus femoris and tibialis anterior muscles during compensation can increase from 0.6 to 0.9, indicating enhanced synergy. These features collectively constitute a description of compensatory synergy, providing a basis for subsequent fusion analysis.

[0073] In an embodiment of the present invention, the balance risk is further quantified through feature fusion and time series analysis. Based on the balance state description vector and the compensatory synergy feature, a weighted summation method is used to construct a feature fusion function, and the weight is determined by principal component analysis. For example, the displacement amplitude weight is 0.4, and the myoelectric synergy weight is 0.3. The fusion feature is normalized and mapped to the range of 0 to 1 to generate a balance risk feature sequence. Time series analysis extracts trend terms and periodic terms. The trend term calculates the slope by polynomial fitting, and the periodic term identifies the fluctuation frequency by Fourier transform. When the trend slope exceeds 0.1 per second and lasts for more than 3 seconds, the system determines it to be a high-risk state. The risk assessment results are presented in the form of scores. The normal population score is between 80 and 100, the mild risk is 60 to 80, the moderate risk is 40 to 60, and the severe risk is less than 40.

[0074] In order to enhance the scientific nature of the assessment, this step also conducts a multi-scale analysis of the compensatory coordination features. Wavelet decomposition divides the electromyographic signal into four frequency bands, corresponding to rapid reflex regulation, voluntary control regulation, posture maintenance, and slow drift. In the compensatory state, the energy proportion of the rapid reflex frequency band increases from 15% in normal times to 40%, reflecting the degree of muscle activity in emergency adjustments. Multi-scale features are reconstructed to generate a compensation pattern description vector, which, after being fused with the center of gravity data, can more comprehensively characterize the dynamic process of balance shift. For example, when the tibialis anterior muscle is overactivated due to the forward shift of the center of gravity, the increase in the energy of its rapid adjustment component indicates a potential risk of lower limb fatigue.

[0075] In an embodiment of the present invention, the application of the time series analysis algorithm significantly improves the real-time and accuracy of risk assessment. The recursive neural network captures small changes in displacement rate through a sliding window, ensuring sensitivity to the initial stage of posture adjustment. The wavelet transform provides a multi-dimensional perspective on the movement of the center of gravity. For example, the correlation between the moment of increased frequency and increased amplitude can be intuitively observed through the time-frequency spectrum. This multi-level analysis method not only reveals the causal relationship between center of gravity shift and muscle compensation, but also provides users with quantitative risk feedback. For example, when the score drops to 60, it prompts that attention should be paid to posture adjustment. The window length or network parameters can be adjusted according to the actual scenario. For example, in a dynamic environment, the window can be shortened to 0.5 seconds to improve the response speed.

[0076] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a body measurement all-in-one machine is provided, which implements in-depth analysis and early warning of balance risks through the following steps.

[0077] S106. If the dynamic balance risk assessment result exceeds the preset danger threshold, the correlation parameters of the pelvic tilt angle and the arch height are clustered and analyzed to identify the potential causes of balance ability degradation and generate corresponding warning signals.

[0078] In this embodiment of the present invention, when the risk assessment results indicate a significant decrease in balance ability, the system automatically initiates a parameter analysis process, using multidimensional data mining technology to locate abnormal factors. During implementation, the risk status is first quantified and a threshold range is determined.

[0079] S1061. Construct a risk status curve based on the dynamic balance risk assessment data and calculate the danger threshold interval. At the same time, use a density clustering algorithm to analyze the fluctuation characteristics of the pelvic tilt angle sequence, extract abnormal parameters through Euclidean distance measurement, and use plantar pressure distribution data to generate abnormal arch morphology characteristics.

[0080] Specifically, the dynamic balance risk assessment results are expressed as a score from 0 to 100, with scores above 80 considered normal, 60 to 80 considered mild risk, 40 to 60 considered moderate risk, and below 40 considered severe risk. A risk status curve is plotted using time series data, reflecting the changing trend of the score over time. When the slope of the curve decreases by more than 1 point per minute, it indicates a worsening deterioration in balance ability. A piecewise linear mapping divides the score into risk intervals. For example, the interval below 40 is defined as the danger threshold, triggering subsequent analysis. The pelvic tilt angle is measured using an optical tracker, with markers placed on the anterior and posterior superior iliac spines. The normal value is between 10 and 15 degrees. A density clustering algorithm uses the angle sequence as input and calculates the fluctuation amplitude and period. The normal fluctuation amplitude is less than 2 degrees, with a period of 0.5 to 2 seconds. If the angle continuously exceeds 20 degrees or fluctuates by more than 5 degrees, the degree of deviation is quantified using Euclidean distance to generate abnormal pelvic tilt parameters. The plantar pressure sensor array collects data in a 16×16 grid with a range of 0 to 600 kPa. A region growing algorithm with a threshold of 50 kPa is used to extract the arch boundary, extending from the forefoot to the calcaneus. The boundary curve is smoothed using cubic spline interpolation, and the longitudinal and transverse arch heights are calculated. Normal longitudinal arch heights range from 15 to 25 mm, and transverse arch heights range from 8 to 12 mm. Significant deviations from these heights are labeled as abnormal arch morphology.

[0081] S1062. A correlation matrix was constructed based on the abnormal pelvic tilt parameters and the abnormal arch morphology parameters. The parameters were grouped and analyzed using a hierarchical clustering method to extract the dominant abnormal factors and evaluate their contribution to the deterioration of balance ability.

[0082] In this step, a correlation matrix is ​​generated by calculating the Pearson correlation coefficient between abnormal pelvic anterior tilt and abnormal arch morphology. The coefficient ranges from -1 to 1, with positive values ​​indicating changes in the same direction and negative values ​​indicating changes in the opposite direction. For example, the correlation coefficient between increased pelvic anterior tilt and decreased arch height may reach 0.8, suggesting a synergistic effect between the two. Hierarchical clustering uses the longest distance method to group parameters with absolute correlation coefficients greater than 0.7, forming a parameter combination signature. Principal component analysis extracts the dominant component of the combination signature. The first principal component typically explains more than 65% of the variance, reflecting the core trend of the abnormal pattern. Within-group variance calculations are used to determine the contribution of each parameter. For example, a 60% contribution of abnormal pelvic anterior tilt indicates that it is the primary influencing factor, while a 30% contribution of abnormal arch height indicates that it is a secondary factor. This analytical method clearly distinguishes the influence of a single abnormality from the cumulative effect.

[0083] In this embodiment of the present invention, the construction of a risk status curve provides a dynamic perspective for anomaly identification. For example, when the score drops from 80 to 40, the sharp increase in the slope of the curve prompts the system to immediately analyze the underlying cause. The density clustering algorithm, using adaptive density thresholds, can effectively separate normal from abnormal fluctuations. For example, it can cluster angles exceeding 20 degrees in a sequence, revealing the persistence of abnormal pelvic posture.

[0084] To further improve analysis accuracy, this step also extracts frequency-domain features of abnormal arch morphology. Fast Fourier transforms are used to calculate the power spectral density of the pressure distribution surface. A normal arch's energy is concentrated below 0.5 Hz, reflecting slow postural adjustment, while an increase in high-frequency components between 2 and 5 Hz indicates unstable support. Wavelet transforms further decompose these high-frequency fluctuations, extracting the time-frequency distribution of abnormal signals. For example, when the arch collapses, the proportion of high-frequency energy increases from 5% to 20%, providing a multi-dimensional basis for quantifying abnormal parameters.

[0085] In an embodiment of the present invention, the combination of the parameter correlation matrix and hierarchical clustering not only reveals the biomechanical correlation between pelvic tilt and arch height, but also provides users with intuitive causal feedback through contribution analysis. For example, when both are abnormal at the same time, the risk score may drop sharply to below 40, indicating a significant superposition effect. This data-based grouping analysis method can effectively locate the root cause of balance degradation and provide scientific guidance for subsequent intervention measures. The clustering threshold can be adjusted according to the characteristics of the subjects. For example, people who are sensitive to arch abnormalities can reduce the correlation coefficient grouping standard to 0.6 to capture more potential associations.

[0086] In an embodiment of the present invention, a method for early warning of the dynamic balance posture of lower limbs and center of gravity based on a physical fitness tester is provided, and risk early warning and real-time feedback are achieved through the following steps.

[0087] S107. Generate a warning signal based on the dynamic balance risk assessment result, correlate and map it with the real-time change trend of the pelvic tilt angle and the arch height, and generate detailed balance degradation warning information based on the potential degradation causes, and present it to the user in real time through the system interface.

[0088] In the embodiment of the present invention, risk assessment data and multi-source sensor information are used to generate intuitive early warning feedback through predictive analysis and feature mapping. When implemented, risk trends are first dynamically monitored and early warning parameters are extracted.

[0089] S1071. Calculate the dynamic threshold range based on the historical data distribution and extract the fluctuation characteristics to generate the first warning parameter. Use a recursive neural network to predict the risk score change trend and generate a graded warning mark. At the same time, calculate the support area change rate from the plantar pressure sensor array data to extract the second warning parameter.

[0090] Specifically, dynamic balance risk assessment uses a 0 to 100 scoring system. Historical data is used to calculate threshold ranges based on normal distribution characteristics. For example, the mean pelvic anterior tilt angle is 12 degrees, with a standard deviation of 1.5 degrees, and the threshold range is set between 9 and 15 degrees. Fluctuation characteristics are extracted using a 0.2-second sliding window. When the rate of change in the angle exceeds 5 degrees per second, it is flagged as an anomaly and the first warning parameter is generated. A recursive neural network uses the risk score sequence and rate of change as input, with a prediction window set to 5 seconds. The network contains 64 hidden neurons, capable of capturing both short-term trends and long-term dependencies. A yellow warning is generated when the predicted score decreases at a rate exceeding 2 minutes per second, an orange warning at 5 minutes per second, and a red warning at 10 minutes per second, distinguished by flash frequencies of 1, 2, and 3 flashes per second, respectively. The plantar pressure sensor array collects data using a 16×16 grid with a sampling frequency of 50 Hz. The support area is calculated using a 50 kPa threshold, with a normal range of 250 to 300 square centimeters. The support area change rate is calculated based on 5 frames of data. If it exceeds 20 square centimeters per second, it is extracted as the second warning parameter to reflect the dynamic adjustment of the arch shape.

[0091] S1072. Construct a mapping relationship table using the first warning identifier and the second warning parameter, generate a warning data packet including abnormal parameters and prompt content, and transmit it to the display terminal via a serial communication protocol to achieve hierarchical visual feedback.

[0092] In this step, a mapping table associates warning indicators with pelvic posture and arch morphology parameters. For example, a yellow warning corresponds to an angle deviation of 9 to 15 degrees or an area change rate exceeding the standard, and is matched with prompt content such as "The pelvic tilt is slightly abnormal, please adjust your posture." The warning data packet encapsulates the timestamp, warning level, parameter value, and recommended measures, and is transmitted using a serial protocol with a baud rate of 115200, 8 data bits, and 1 stop bit. After the display terminal parses the data, the interface is divided into three areas: the real-time warning status on the left is indicated by color and flashing, the pelvic angle and arch support area curve is plotted in the middle, and detailed prompts are displayed on the right. The pop-up window is displayed in the center, and a mild warning has a white translucent background, a moderate warning has a light yellow background, and a severe warning has a light red background to improve recognition.

[0093] In this embodiment of the present invention, the generation and real-time mapping of warning signals ensures that users are promptly aware of the risk of balance degradation. For example, when the pelvic tilt angle exceeds 15 degrees and continues to increase, the system not only triggers an orange warning but also displays the changing trend through a curve, visually indicating the source of the anomaly. The predictive capabilities of the recurrent neural network enable proactive warnings, for example, issuing a warning five seconds before the score drops to 60, buying time for intervention.

[0094] To enhance the practicality of this feedback, a heat map of plantar pressure distribution is incorporated into this step. When the arch support area is abnormally reduced, the heat map highlights the forefoot or heel area in red, alerting the user to a possible shift in center of gravity due to arch collapse. This visual design helps users understand the direct correlation between abnormal parameters and their physical condition. For example, an excessive rate of area change may be associated with fatigue from prolonged standing. Suggested actions such as "rest and elevate your feet" provide specific guidance.

[0095] In an embodiment of the present invention, real-time feedback of warning information is achieved through multi-level data processing and interface optimization. The adaptive adjustment of the dynamic threshold range ensures the individualization of parameter monitoring. For example, for users with flatter arches, the system will relax the area change rate threshold to 25 square centimeters per second to reduce false alarms. The flexibility of the mapping relationship table allows the prompt content to be updated according to actual needs, such as adding a "check insole support" suggestion in sports scenes. The partition layout and pop-up window design of the display interface not only improve the efficiency of information transmission, but also highlight key anomalies through color coding and font size differences to enhance the user experience. The prediction window or flashing frequency can be adjusted according to the application scenario, for example, shortened to 3 seconds in a high dynamic environment to improve the response speed.

[0096] The above only lists some preferred embodiments of the present invention, but the present invention is not limited thereto, and many improvements and modifications can be made. As long as the improvements and modifications are made on the basis of the basic principles of the present invention, they should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for early warning of dynamic balance posture of lower limbs and center of gravity based on a body measurement integrated machine, characterized in that: The method comprises: obtaining three-dimensional skeletal data of a subject in a standard standing posture and a habitual standing posture, collecting pelvic tilt angle and arch height parameters through a built-in sensor of a physical measurement all-in-one machine, generating an initial skeletal model using a three-dimensional skeletal reconstruction algorithm, and obtaining correlation parameters between the pelvis and the arch; extracting pelvic tilt angle change data from the initial skeletal model, combining the arch height parameters to determine the center of gravity position, calculating the center of gravity offset caused by the pelvic position change, and determining the center of gravity forward movement trend according to the time change trend of the offset, comprising: receiving pelvic angle data sent by a skeletal surface equation, processing the pelvic angle data with a Kalman filter to obtain a pelvic tilt angle sequence; obtaining the pelvic tilt angle sequence obtained by a plantar pressure sensor array according to the pelvic tilt angle sequence. The pressure distribution data of the set is collected, and the pressure center coordinate sequence and the arch height value are calculated through the pressure distribution data; the pelvic tilt angle sequence and the arch height value are used to establish the human body center of mass distribution equation, and the initial coordinates of the whole body center of gravity are obtained by linear superposition calculation; for the initial coordinates of the whole body center of gravity, Euler angles are used to describe the three-dimensional posture changes of the pelvis, and the center of gravity displacement vector sequence is calculated through the rigid body kinematic equation. The horizontal plane displacement field distribution of the center of gravity displacement vector sequence is obtained by least squares fitting; the standing posture change is predicted according to the center of gravity forward movement trend. When the standing posture changes, the electromyographic signals of the key muscle groups of the lower limbs are obtained and processed to obtain the lower limb muscle activation pattern. The lower limb muscle activation is determined by combining the relative size and coordination degree of each muscle activation in the activation pattern. The method comprises the following steps: constructing a predicted space trajectory diagram according to the center of gravity displacement sequence data, obtaining a deviation value between the center of gravity projection position and the support base area edge threshold value from the predicted space trajectory diagram; collecting potential signals of muscle groups around the lower limb joints by a surface electromyography sensor, filtering the potential signals by a center frequency bandpass filter, and obtaining a filtered electromyography signal waveform diagram; performing a wavelet transform on the filtered electromyography signal waveform diagram to obtain a wavelet coefficient matrix, calculating a signal mutual correlation coefficient matrix between muscle groups from the wavelet coefficient matrix, and obtaining a muscle group synergy feature vector by eigenvalue decomposition of the mutual correlation coefficient matrix; and obtaining a muscle group synergy feature vector according to the comparison between the muscle group synergy feature vector and standard human body measurement data. The Euclidean distance value is calculated based on the constructed standard feature vector, and a compensation measurement function is established from the Euclidean distance value to obtain a thermal distribution map of the degree of muscle compensation; the pelvic tilt angle, arch height and center of gravity position in the standard standing posture are obtained as individualized baseline data, and a dynamic balance benchmark value reflecting the individual's normal balance ability is generated by statistically analyzing the baseline data; based on the dynamic balance benchmark value, the balance offset index in the standing posture change is calculated by a time series analysis algorithm in combination with the center of gravity forward shift trend and the lower limb muscle compensation pattern, and a dynamic balance risk assessment result is obtained, including: performing a wavelet transform on the center of gravity displacement sequence to obtain a time-frequency spectrum, extracting the dominant frequency and amplitude parameters of the center of gravity movement from the time-frequency spectrum, and constructing a balance state description vector;A recursive neural network is used to process the balance state description vector, and the center of gravity displacement rate curve is calculated through a sliding time window to obtain the balance offset. Envelope extraction is performed on the lower limb electromyographic signals, and the compensatory activation intensity is calculated based on the envelope amplitude. The temporal correlation between muscle groups is solved through the cross-correlation function to obtain the compensatory synergy feature. A feature fusion function is established based on the balance state description vector and the compensatory synergy feature, and the fused feature is normalized. The trend item of the risk feature sequence is extracted through time series analysis. If the trend slope exceeds a preset threshold, the balance risk level is determined. If the dynamic balance risk assessment result exceeds the danger threshold, cluster analysis is performed on the parameters associated with the pelvic tilt angle and the arch height, and the associated parameters are grouped to identify the potential causes of balance ability degradation. Based on the dynamic balance risk assessment result, a corresponding early warning signal is generated. The early warning signal is correlated with the real-time change trend of the pelvic tilt angle and the arch height. Combined with the potential causes of degradation, balance degradation warning information is generated and fed back in real time through the integrated physical test system.

2. The method according to claim 1, characterized in that The method obtains three-dimensional skeletal data of the subject in a standard standing posture and a habitual standing posture, collects pelvic tilt angle and arch height parameters through the built-in sensor of the physical measurement all-in-one machine, generates an initial skeletal model using a three-dimensional skeletal reconstruction algorithm, and obtains correlation parameters of the pelvis and the arch of the foot, including: collecting three-dimensional coordinate data of skeletal feature points, performing noise reduction processing on the three-dimensional coordinate data through a median filter to obtain filtered feature point position data; calculating a pressure center trajectory curve and arch area pressure values ​​based on the filtered feature point position data and plantar pressure distribution data collected by a pressure sensor array; collecting angular velocity data through a three-axis gyroscope sensor array arranged in the pelvic area, and obtaining a pelvic spatial posture angle based on an integral operation of the angular velocity data; establishing a plantar surface equation based on the arch area pressure value, calculating longitudinal arch height and transverse arch height parameters from the plantar surface equation, establishing a group of geometric constraint equations based on the pelvic spatial posture angle and the longitudinal arch height and transverse arch height parameters, and solving the constraint equations by the least squares method to obtain positional relationship parameters between the pelvis and the arch of the foot.

3. The method according to claim 1, characterized in that The method obtains the pelvic tilt angle, arch height and center of gravity position in a standard standing posture as individualized baseline data, and generates a dynamic balance reference value reflecting the individual's normal balance ability by statistically analyzing the baseline data, including: obtaining pelvic marker point data collected by an optical tracker, and obtaining a pelvic tilt angle sequence through three-dimensional spatial coordinate transformation based on the marker point data; receiving a pressure distribution map collected by a plantar pressure sensor array, and calculating an arch support index based on the pressure distribution map, wherein the arch support index is obtained by calculating the longitudinal arch height value and the transverse arch height value of the arch area pressure curve; establishing a center of mass position calculation equation based on the pelvic tilt angle sequence and the arch support index, and obtaining the three-dimensional coordinates of the center of gravity from the calculation equation, wherein the calculation equation includes the segment mass ratio and the spatial coordinate transformation matrix; performing kernel density estimation on the three-dimensional coordinates of the center of gravity to obtain a center of gravity displacement envelope, extracting a displacement direction and amplitude feature sequence from the center of gravity displacement envelope, and obtaining a balance ability reference value based on the feature sequence using a Bayesian estimation method.

4. The method according to claim 1, wherein If the dynamic balance risk assessment result exceeds the danger threshold, cluster analysis is performed on the parameters associated with the pelvic tilt angle and the arch height, and the associated parameters are grouped to identify potential causes of balance ability degradation, including: constructing a risk status curve based on the dynamic balance risk assessment data, calculating the piecewise linear mapping value through the risk status curve, and obtaining the danger threshold interval parameter; using a density clustering algorithm to process the pelvic tilt angle data, calculating the fluctuation amplitude and change period from the angle data sequence, and obtaining the abnormal pelvic tilt parameter through the Euclidean distance metric; calculating the pressure distribution surface based on the plantar pressure sensor array data, and using the region growing algorithm to extract the arch boundary curve from the pressure distribution surface to obtain the abnormal arch morphology parameter; constructing a parameter correlation matrix for the abnormal pelvic tilt parameter and the abnormal arch morphology parameter, using a hierarchical clustering method to group and cluster the correlation matrix, and obtaining the dominant factor of balance ability degradation through intra-group variance calculation.

5. The method according to claim 1, characterized in that Based on the dynamic balance risk assessment results, a corresponding warning signal is generated, and the warning signal is correlated and mapped with the real-time change trend of the pelvic tilt angle and the arch height. In combination with the potential causes of degeneration, balance degradation warning information is generated, and real-time feedback is provided through the physical measurement integrated system, including: obtaining a dynamic threshold range based on historical data distribution, extracting a fluctuation feature sequence from the dynamic threshold range, and obtaining a first warning parameter; predicting and analyzing the first warning parameter through a recursive neural network, and generating a first warning mark of a corresponding level if the rate of decrease of the predictive analysis score exceeds a preset threshold; obtaining a support area change rate curve based on the plantar pressure sensor array, extracting a trend feature sequence from the change rate curve, and obtaining a second warning parameter; establishing a mapping relationship table between the first warning mark and the second warning parameter, matching the warning prompt content through the mapping relationship table, and generating a warning data packet to the display terminal.

Images