Biomechanical dynamic capture system and force measuring table

Abstract

The invention relates to the technical field of motion capture, in particular to a biomechanical motion capture system and a force table, comprising a high-precision sensor, an acquisition module, a processing module, an analysis module and a regulation module, the acquisition module is used for acquiring multi-dimensional force and position data when a subject moves and synchronously recording in real time through the high-precision sensor, and the processing module is used for processing the multi-dimensional force and position data through the regulation module. Performing primary processing on the acquired data through a filtering algorithm to reduce noise interference caused by environment and equipment; the analysis module is used for carrying out multi-scale analysis on the processed data and extracting characteristic parameters reflecting fine action changes so as to improve the capture precision; and the regulation and control module optimizes the feedback control system based on the characteristic parameters, regulates and controls the signal purity and the motion capture error in real time, and ensures the accuracy and the stability of the biomechanical data.

Description

Technical Field

[0001] The present application relates to the field of motion capture technology, particularly to a biomechanical motion capture system and a force platform. Background Art

[0002] In a motion capture system, the biomechanical subsystem usually relies on high-precision devices such as force platforms, treadmills, pressure-sensitive pads, and electromyographs. The biomechanical capture system is an advanced device that combines force measurement and motion capture technology and is widely used in fields such as sports science, rehabilitation medicine, and ergonomics. Through high-precision sensors and camera systems, it can synchronously record the force distribution and posture changes during human movement, providing detailed motion analysis data for researchers.

[0003] However, this system also faces some challenges. On the one hand, the biomechanical data will inevitably be mixed with noise components, such as environmental vibrations or minute perturbations of the sensors themselves. These interferences will result in impure signals and affect the accuracy of experimental results. On the other hand, when capturing the movements of subjects, there may also be problems with large errors in the capture accuracy of subtle changes in the system. For example, high-speed movements or minute posture adjustments may not be accurately recognized.

[0004] Therefore, in order to ensure the data quality and the reliability of analysis results, it is necessary to continuously improve noise regulation techniques and algorithms for enhancing motion capture accuracy in research and technological development, so as to more precisely handle these complex situations. Summary of the Invention

[0005] To solve the problems raised in the above background art, the present application provides a biomechanical motion capture system and a force platform.

[0006] The present application provides a biomechanical motion capture system and a force platform, adopting the following technical solutions: A biomechanical motion capture force platform includes a sensor, a top plate, and a protective base. The top plate is fixedly installed on the top of the protective base. The sensor is arranged on the bottom surface of the top plate. An acquisition module, a processing module, an analysis module, and a regulation module are also arranged between the top plate and the protective base.

[0007] A biomechanical force platform motion capture system includes:

[0008] Collect multi-dimensional force and position data of a subject during movement and synchronously record them in real time through high-precision sensors;

[0009] Preliminarily process the collected data through a filtering algorithm to reduce noise interference caused by the environment and equipment;

[0010] Conduct multi-scale analysis on the processed data to extract characteristic parameters reflecting subtle motion changes to improve capture accuracy;

[0011] Optimize the feedback control system based on the above characteristic parameters to adjust the signal purity and motion capture error in real time, ensuring the accuracy and stability of biomechanical data.

[0012] Preferably, optimizing the feedback control system based on the characteristic parameters extracted by the above analysis module to reflect subtle motion changes further includes:

[0013] Divide the signal purity into several intervals and set different correction factors K for each interval.

[0014] Calculate the current motion capture error E and evaluate the signal purity S;

[0015] Dynamically adjust the correction factor based on the following formula: K = max(Sα * E + Dβ, 0), where S represents the signal purity, E represents the motion capture error, α represents the influence factor with a value range of [0, 1], β is the influence factor of D, and the data stability index D can be obtained by calculating the variance of the data over a period of time. The smaller the variance, the more stable the data, and the larger the D value;

[0016] When K is greater than the set threshold KT, trigger the automatic noise adjustment module for processing to ensure the real-time adjustment of the stability and purity of biomechanical data, and solve the problem of large noise components and motion capture errors in biomechanical data.

[0017] Preferably, the noise regulation and accuracy improvement of biomechanical data based on the above method further include the following steps:

[0018] Obtain the base frequency Fb collected by the high-precision sensor;

[0019] Determine the maximum noise frequency Fn to be filtered according to the signal fluctuation degree;

[0020] Use Fourier transform to separate the main signal, and screen and retain the effective signal segments through the following logical formula: if ω > Fn and A < Am, then remove the signal segment (this frequency component), where ω represents the frequency component, Am represents the maximum amplitude, and A represents the amplitude of each component;

[0021] Construct a new time series data based on the effective signal to reduce the interference caused by the environment and equipment and ensure the data accuracy.

[0022] Preferably, the improvement of the capture accuracy based on the extraction of characteristic parameters of subtle motion changes is specifically reflected in:

[0023] Select the time slice t of a specific motion stage;

[0024] Record the multi-dimensional position coordinate set P(i) of all subjects within this time period, and the corresponding acting force F(i);

[0025] Calculate the velocity V = dP2 / dt2 and angular acceleration at each moment

[0026] Use the following conditions to determine whether a subtle movement change is captured: If V2 + βA ≥ γ (γ is the sensitivity coefficient), perform high-frequency sampling; β represents the proportional gain, V represents the velocity, and A represents the acceleration. This algorithm can improve the system sensitivity and response ability, and can respond more quickly when detecting small human movements.

[0027] Preferably, the method performed to better solve the problem of impure signals is as follows:

[0028] Define an initial weight W0 corresponding to each noise source during the system initialization process;

[0029] Dynamically track the actual influence of each source of noise to obtain the weight Wt;

[0030] Set the ideal signal purity standard I and the current state C, and adjust the weight through iterative calculation: Wi+1 = Wi + (Ci - I) × Q, where Q is the learning rate and Ci is the current purity status measured in each cycle;

[0031] If Wi > W_Lim_max, then limit it to W_Lim_max to ensure that the self-correction mechanism of the system can stably perform the noise reduction function, improve the credibility of the output result, and enhance the user experience.

[0032] Preferably, the specific implementation method for optimizing the capture accuracy is as follows:

[0033] Set the key posture T of the target to capture the target action;

[0034] Continuously obtain the approximate posture At during actual operation;

[0035] Use a machine learning algorithm to compare the difference between the two vectors D = ||At - T||, and give the similarity score S = L - D / L, where L represents the reference length;

[0036] If S < η, η is the threshold, the system automatically triggers a more precise tracking mode. Through this control condition, when the similarity is too low, it can quickly switch to a higher performance working state to improve the detail capture quality and technical stability.

[0037] In summary, the present application includes at least one of the following beneficial technical effects:

[0038] The embodiments of the present disclosure provide a biomechanical force platform motion capture system, including: collecting multi-dimensional force and position data of a subject during movement and synchronously recording them in real time through high-precision sensors; preliminarily processing the collected data through a filtering algorithm to reduce noise interference caused by the environment and equipment; performing multi-scale analysis on the processed data to extract characteristic parameters reflecting subtle motion changes to improve the capture accuracy; optimizing the feedback control system based on the above characteristic parameters to regulate the signal purity and motion capture error in real time to ensure the accuracy and stability of biomechanical data. Through the solution of the embodiments of the present disclosure, it is possible to solve the problem of regulating the noise components in biomechanical data to address the issue of impure signals, and to regulate the capture accuracy of subtle changes in the subject's movements to address the problem of large motion capture errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a structural diagram of the force platform;

[0040] Figure 2 is a flowchart of the biomechanical motion capture system.

[0041] BRIEF DESCRIPTION OF THE DRAWINGS: 1. Top plate; 2. High-precision sensor; 3. Protective base. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] The following details the embodiments of the present application, and the examples of the embodiments are shown in the drawings.

[0043] In the description of this specification, the descriptions referring to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0044] The embodiments of the present application disclose a biomechanical motion capture system and a force platform, referring to Figure 1-2, Next, with reference to the accompanying drawings, the biomechanical motion capture force platform of the present invention will be described, which includes a sensor 2, a top plate 1, and a protective base 3. The top plate is fixedly installed on the top of the protective base, the sensor is arranged on the bottom surface of the top plate, and an acquisition module, a processing module, an analysis module, and a regulation module are also provided between the top plate and the protective base. The biomechanical force platform motion capture system of the present invention can collect, process, analyze, and optimize the motion data of the subject, aiming to ensure the accuracy and stability of the data, especially focusing on regulating the noise components and improving the capture accuracy of the subtle motion changes of the subject to cope with the challenges of actual application scenarios.

[0045] The system first collects the force and position information of the subject during movement from multiple dimensions and synchronously records them in real time through high-precision sensors. For example, in a motion analysis laboratory, a three-dimensional dynamic camera array and a force platform synchronously obtain data such as the limb posture, ground reaction force, and joint movement angle of the subject. The system is built with a time synchronization module to align and integrate multi-modal data with millisecond-level precision, ensuring that the biomechanical parameters reflect the real motion scenario. Different sampling rates and filter settings are designed for different experimental conditions or action tasks. For example, a high-frequency acquisition of 500 Hz is used for gait research, and it is adjusted to 120 Hz for analyzing low-intensity rehabilitation training.

[0046] After the data acquisition is completed, it is immediately transmitted to the data initial filter component for preliminary processing. Advanced adaptive noise cancellation technology is used to combat environmental and mechanical source interference, reducing white noise and other frequency-domain mutation points. A digital band-stop filter is used to remove periodic non-steady-state interferences such as 60 Hz power grid clutter and human heartbeat pulses, and a statistical threshold detection mechanism is used to exclude occasional single shocks. After processing, a relatively smooth and clean time series is obtained, providing a good basis for the training of subsequent deep learning algorithms.

[0047] To extract features from the processed data, the system uses multi-scale analysis techniques and mathematical tools such as wavelet transform, short-time Fourier transform, or local mean decomposition to reveal the internal laws of the original signal. For example, it can accurately locate the rapid foot contact points and the contact surface pressure distribution within a short time, precisely depict the energy aggregation characteristics of tiny gesture actions such as the fingertips touching the table in each sub-band, and at the same time track the trajectories of each key point in the multi-dimensional coordinate system to form a complete posture description system.

[0048] Based on the above characteristic parameters, an optimized feedback control system is established to continuously monitor and improve the output quality. When the measurement error exceeds the allowable range or the signal-to-noise ratio decreases, affecting the diagnostic accuracy, compensation measures are immediately initiated. For example, in clinical tests, if the torque of the ankle joint fluctuates violently due to the patient's inflexible muscle control, the feedback device sends instructions to the main processor to adjust the magnification of the camera's perspective, focus on the ground to track the leg path, and at the same time enable the accelerometer to assist in stabilizing the body posture judgment. If there is interference from an external vibration source, the system quickly evaluates and switches to a backup redundant solution to maintain the normal business process and ensure the high-level operation of the overall performance.

[0049] The entire biomechanical force platform motion capture system is designed with full consideration of ease of use and user interface friendliness. Without affecting scientific nature, the operation steps are simplified, the interaction function is enhanced, which is convenient for researchers to use, maintain and update the software version in daily life. This system is committed to solving the two core problems that biomechanics experts have faced for a long time, namely eliminating measurement noise and improving the resolution of motion analysis, and has been successfully practiced in many well-known sports medicine research institutions at home and abroad.

[0050] The feedback control system is optimized based on the above characteristic parameters, which is divided into four stages: data acquisition, parameter setting, control algorithm adjustment and system feedback. During data acquisition, the force platform sensor obtains the time series signal of the ground reaction force of the human body. For example, when an athlete uses training equipment to run, the data of each step cycle is collected and recorded, and the sampling frequency is not less than 1 kHz to comprehensively cover the force change curve.

[0051] In the parameter setting link, according to performance indicators such as the maximum peak response time, minimum mean square error and stability margin, the key control variables and their value ranges are determined. For example, for the optimal gain K2, it is generally taken as a positive value and does not exceed 50 to avoid overcorrection causing oscillation instability. If the goal is to make the force output smooth during the starting acceleration stage of the athlete, K2 is set according to past successful starting examples; if there is hysteresis, appropriately increasing K2 can improve the dynamic response characteristics without affecting the overall stability.

[0052] The control algorithm adjustment introduces a more refined adjustment factor to adapt to special situations or changing conditions in actual applications. The PID proportional-integral-derivative operation plays a core role, combining historical behavior patterns, real-time change rates and cumulative deviation amounts to dynamically predict the best control measures and achieve ideal target trajectory tracking. Determine the P, I, D coefficients according to the system characteristics. For example, keep P around 4 to quickly reflect transient mutations without overly amplifying the error term.

[0053] The system feedback mechanism analyzes the past training result library with the help of built-in machine learning algorithms, identifies repeated or similar scenario segments, provides guidance and reference for the next decision, realizes closed-loop automatic adjustment, continuously optimizes the movement posture and force distribution, and reduces energy consumption losses. For example, in the case of repeated poor performance in practice, the intelligent module helps the user correct potential defects by adjusting the suggestion intensity or changing the force application part.

[0054] The signal purity is divided into multiple intervals according to the quality level. In the application scenario of a biomechanical force platform, based on the ratio of noise to useful signal in the system output signal, such as the signal of the runner's foot landing force being impure due to environmental interference or sensor characteristics, 90%-100% can be set as the high-purity interval, 80%-90% as the medium-high interval, etc. A correction factor K is assigned to each interval to compensate for the data analysis deviation caused by signal impurity. For example, in the interval of 85%-90%, the G value is taken as 0.97, and for signals with a high-quality level above 95%, the G value approaches 1. The signal is corrected through the formula V = V * K to improve the reliability of the measurement result. For example, when an athlete takes off and is affected by external electromagnetic waves, the original electrical signal weakens by 5%. Through the corresponding K factor compensation, stable and accurate information is output.

[0055] The action capture system of the biomechanical force platform calculates the current action capture error E and evaluates the signal purity S through the following steps: Collect reference motion data, use a calibrated high-precision measurement tool to generate a reference motion pattern, such as marking the positions of human bones; Calculate the dynamic capture error E through to calculate the dynamic capture error. Among them, X ref 、Y ref 、Z ref respectively represent the position coordinates on the X, Y, and Z axes at a certain moment in the reference motion pattern; X obs 、Y obs 、Z obs represent the position coordinates on the three axes at the same moment actually captured. The formula calculates the Euclidean distance between the reference trajectory and the actual trajectory in three-dimensional space to measure the deviation between the two. The acceptable range of the error E is usually between -5 and +5 millimeters, depending on the accuracy requirements of the application scenario; Collect real-time capture data, ensure a stable environment, and continuously record multiple sets of consecutive frames; Quantify the signal purity S using the formula formula to quantify the signal purity S, where Δx i represents the change amount of adjacent two-frame data in a certain feature dimension, and N represents the total number of frames collected. The value range of the signal purity S is between [0,1]. The closer S is to 1, the less the signal is affected by noise and the purer the data; on the contrary, the closer S is to 0, the more noise is included in the signal and the worse the data quality. For example, when collecting human hand motion data, Δx iIt can be the change amount of the angle of a certain joint of the hand in two adjacent frames. By summing up the change amounts of all frames and calculating according to the formula, the signal purity S is obtained.

[0056] The correction factor is dynamically adjusted based on the formula K = max(Sα*E + Dβ, 0). S represents the signal purity, E represents the motion capture error, α is the influence factor, with a value range of [0, 1], β is the influence factor of D, D is the data stability index, which is obtained by calculating the data variance within a certain period of time. The smaller the variance, the larger the D value. When S = 0, another set of data verification processes is started to check the working state of the sensor and the continuity of the collected data to determine the quality of the data. For example, when processing biomechanical experiment data, if the external noise causes S = 0.6, E = 0.4, and α = 3, K is approximately 0.087; if S = 0, the max function is introduced to ensure a minimum return of 0.

[0057] When K is greater than the set threshold KT, the automatic noise adjustment module is triggered. Define the noise evaluation parameter CS = σ / σ_ideal, where σ is the standard deviation of the current data noise, and σ_ideal is the theoretically expected ideal noise level, with an optimal value of approximately 0.01 unit. The closer CS is to 1, the better the system performance. The default value of KT is 1.5 - 2, which is optimized according to actual tests. When K is greater than KT, the system is at risk of instability or high noise, and the automatic noise adjustment function is activated. Signal processing techniques such as frequency domain selective reduction and mathematical morphology methods are used to filter the noise and output a smoothed result to ensure the purity and stability of the biomechanical parameters.

[0058] The steps for regulating the noise and improving the accuracy of biomechanical data are as follows: After receiving the original data, a low-pass filter is used to remove high-frequency noise. The cut-off frequency f_c of the filter is default set between 50Hz and 200Hz to suppress fast fluctuations that are not physiologically relevant; the preliminarily filtered data is passed to the noise feature recognition unit, and machine learning algorithms are used to identify and quantify the attributes of the remaining noise, and the abnormal parts are marked; advanced noise reduction processing is performed, and techniques such as wavelet transform are introduced. The threshold T0 is adjusted according to the requirements of the application field to balance the noise reduction efficiency and data fidelity rate, with a value range of 0.01 - 0.1; a regression equation is constructed using the multiple fitting method to correct the system deviation error E, and the error correction coefficient Xk is determined through multiple rounds of comparison experiments, generally ranging from 0.9 to 1.1; the improvement results are comprehensively evaluated, measured by indicators such as the signal-to-noise ratio SNR and the root mean square error RMSE. For example, after improvement, the SNR value jumps from 8dB to 30dB, meeting high-standard requirements.

[0059] The steps to obtain the base frequency Fb collected by a high-precision sensor are as follows: Select or confirm the sensor model and basic technical parameters, considering sensitivity, accuracy, etc.; Determine the expected application range and the influence of environmental variables, and analyze the influence of the environment on Fb. For example, a biomechanical force platform is affected by temperature and humidity; Calculate the base frequency through the formula Fb = T1 / S, where T1 is the measurement time and S is the sampling interval. It is necessary to reasonably balance the relationship between the two. For example, if a force platform can complete 2048 high-quality measurements per second, an appropriate S value can be selected to calculate Fb.

[0060] The steps to determine the maximum noise frequency Fn to be filtered according to the signal fluctuation degree are as follows: Collect the original signal to be analyzed, such as the pressure and motion data when a subject is walking; Calculate the standard deviation to evaluate the total variance component of the original signal;

[0061] The formula is sigma = sqrt{frac{1}{N}sum_{i = 1}^{N}(x_i - mu)^2}; Set the threshold according to the standard deviation; Determine the maximum acceptable noise frequency Fn based on the threshold. Fn should be close to but not exceed the distortion limit; Apply a low-pass filter to the original signal to verify the effect. For example, if the standard deviation exceeds the given range, it indicates that it is affected by external noise. Fn can be set to an integer value slightly smaller than the frequency at which the peak appears to reduce the false alarm rate and maintain the signal-to-noise ratio.

[0062] Use Fourier transform to separate the main signal, and screen and retain the effective signal segment through logical formulas. Convert the original time-domain acquisition signal into a frequency-domain representation through fast Fourier transform, and decompose it into different frequency components and their corresponding amplitudes. The frequency threshold Fn determines the filtering limit. In general, the optimal range for human motion analysis is about 3 to 5 Hz. The maximum amplitude Am limits the maximum allowable amplitude. The formula is If(ω>Fn&&A<Am), and the frequency components that meet the conditions will be removed. For example, when tracking the foot shock wave of an athlete, set Fn = 4 Hz to remove the noise with a frequency higher than 4 Hz and an amplitude less than Am, and improve the accuracy of target information recognition.

[0063] Use the sensor to collect the original signal in real time to form the initial time series data. Subsequently, adopt technologies such as digital filtering for denoising preprocessing, such as combining high-pass and low-pass filters, to retain the meaningful frequency segments within the human body's dynamic range (generally 0 - 10 Hz), and adjust the parameter thresholds according to the task type. Then identify and mark the time positions and characteristic quantities of important physical phenomena for feature extraction. For example, in gait analysis, locate the peak point and zero-crossing point of the ground contact moment. After that, correct the signal amplitude and time axis, and adopt the Min-Max normalization method. The formula is where x is the original signal value, and min and max are the minimum and maximum values in the original signal respectively). Finally, reorganize the filtered and corrected time series in ways such as dividing by time window. By calculating statistics such as the mean and variance of the data within the window, a stable and reliable long-term trend expression pattern is formed, enhancing the accuracy and reliability of the data. Compact. See if it meets your requirement for simplicity. If you have other ideas, such as adding or deleting certain content, you can tell me at any time.

[0064] The steps to improve the capture accuracy based on the extraction of characteristic parameters of subtle motion changes are as follows: The sensor acquires and stores the original motion data; Screen and analyze the parameter set reflecting the human body's dynamic characteristics from the collected data, such as the range of heel strike force (P_f) in gait [5, 500] Newtons, the range of angle difference (A_diff) [-45°, 45°]. Optimize P_f to 200 to 350 Newtons and A_diff to about ±10 degrees for different populations; Use machine learning algorithms to construct a classifier to process the feature space, distinguish multiple walking modes, and continuously fine-tune the algorithm parameters to reduce the error rate; Through test verification, improve the accuracy of motion tracking and enhance the reliability and practicality of biomechanical research.

[0065] The steps to select the time slice t0 of a specific motion stage are as follows: Determine the research object and lock the part of the body activity or the motion behavior. For example, when studying the running posture correction of athletes, determine that the hip joint is the focus of attention; Set the start and end points of motion capture and select an appropriate time interval. For example, when recording the walking process, select the time from when the foot touches the ground to when it is lifted again; Through the formula t0 = t e -t s (where t e represents the end moment of motion capture, and t s represents the start moment of motion capture) to confirm the appropriate time range. If the video frame rate is f frames per second, according to the video frame rate, use (n is the number of frames) to convert the number of frames corresponding to the start and end moments into time to adjust the start and end moments, ensure the capture of key data points, avoid noise pollution and not miss key events.

[0066] Record the multi-dimensional position coordinate set P(i) of all subjects during this period, as well as the corresponding acting force F(i). The steps are as follows: Assign an identity code to each subject to ensure accurate data correlation; Collect the position information P(i) of the subject in real time. The capturing device records the three-dimensional space coordinate points (x, y, z) at a high frame rate, with the coordinate range {-2 < x < 2}, {-2 < y < 2}, {z >= 0} (the z coordinate in the treadmill experiment is restricted by the movement plane); Read the acting force F(i) through the sensor array. The force value reflects the interaction relationship between the human body and the contact surface. The formula is F = mg + Δv * t. After optimization, the ideal range is about 300N - 800N (calculated based on an ordinary adult of 70KG); Integrate the obtained timestamp, identity number, position parameters, and acting force information to generate a list for analysis.

[0067] Calculate the velocity V = dP / dt and angular acceleration at each moment The steps are as follows: Capture the position vector P (array of [X, Y, Z] coordinates) of the human body or experimental object in real time through the sensor; The parameter t is a discrete time series, and the sampling frequency is, for example, 200Hz (Δt = 0.005s); Take the first derivative of the position P with respect to time t to obtain the velocity V, with the unit m / s; Take the first derivative of the velocity V again to obtain the angular acceleration A, with the unit m / s 2 For example, to analyze the gait pattern, collect continuous displacement data {Pi}, calculate the instantaneous velocity {Vi} and angular acceleration {Ai}, and construct a model of the change trajectory of the center of gravity of the human body.

[0068] Use the condition v 2 + βA ≥ γ, where γ is the sensitivity coefficient, to determine whether a subtle movement change is captured, with high-frequency sampling. Initialize and define the velocity V, acceleration A, and proportional gain β. The values of V and A are determined by the sensor resolution and range. For example, the higher the sensor resolution, the higher the numerical accuracy of the measured velocity V and acceleration A, and its range limits the maximum and minimum values that V and A can measure. β is generally set between 0 and 1. Among them, the range of the sensitivity coefficient γ is adjusted according to the environment and application scenario, generally between 1 and 5. The formula adds the square value of the velocity to the product of the proportional coefficient and the acceleration. When the result is greater than or equal to γ, high-frequency sampling is triggered. For example, in the study of human walking posture changes, at the moment when the heel touches the ground and the posture is transformed, the system measures the data and inputs it into the formula. If the result exceeds, immediately increase the acquisition rate to the rate that meets the requirements of fine movement analysis (such as from 50 frames per second to 200 frames per second, etc., and the specific rate depends on actual needs) to obtain detailed gait data, and select appropriate β (such as 0.8) and γ values to improve the attention level to fine movement changes.

[0069] To better solve the problem of signal impurity, the following methods are implemented: pre-filtering, applying high-pass or low-pass filtering to the original sensor electrical signal to cut down on useless frequency components and retain the vibration characteristics of specific frequency bands during changes in the motion state; signal synchronization optimization, adjusting the small time differences at each measurement point to synchronize all acquisition channels with a fixed time reference point, such as selecting the moment when the tester touches the first pressure sensing unit as the zero moment; application of a noise reduction algorithm, based on technologies such as wavelet transform, selecting a formula model according to the type of noise distribution, and adjusting the decomposition level L (1 - 5) and the threshold T (adjusted according to the background noise power spectral density) to remove residual noise. After this processing, in the gait experiment, the fluctuation amplitude of the torque curve is on average reduced by 30% compared to before the transformation, proving the effectiveness of the solution;

[0070] In the system initialization stage, initial weights W0 are defined separately for each noise source. The subscript "0" here represents the initial moment, and W0 is used to characterize the estimated degree of influence of each noise source on the signal in the initial state.

[0071] During the system operation process, it is necessary to dynamically track the actual influence of each noise source on the signal, and then obtain the real-time weight Wt of each noise source at time t z where "t z " represents any moment during the system operation process. z

[0072] At the same time, an ideal signal purity standard I is set, which is a pre-determined expected signal purity target value. The purity status of the current signal is measured every cycle, and the measurement result is quantified into a value that can reflect the signal purity degree, and this value is the current state C.

[0073] The system adjusts the weights through the following iterative formula: Wi+1 = Wi + (Ci - I) × Q. Here, Wi is the weight value calculated in the i-th step, Ci is the current purity status measured in the i-th cycle, and Q is the learning rate. The value range of the learning rate Q will determine the speed of weight adjustment. When Q is larger, the weight adjustment speed is faster, and the system should respond more quickly to changes in signal purity, but it may lead to an unstable adjustment process; when Q is smaller, the weight adjustment speed is slower, and the system adjustment process is relatively stable, but it is less sensitive to changes in signal purity. Through this iterative formula, the system can continuously optimize the weights of the noise sources according to the difference between the signal purity status measured every cycle and the ideal purity standard, so as to better adapt to signal changes and improve the signal processing effect;

[0074] The specific implementation method for optimizing the capture accuracy is as follows:

[0075] ​First, set the key posture T of the target action to be captured. Here, T is a vector that represents the position, angle, etc. of each key feature point of the target action in the ideal state, and it is the standard posture we expect to capture. For example, in human motion capture, T can be the precise angle combination of each joint of the human body under a specific dance action.

[0076] During the actual operation process, the system continuously obtains the approximate posture At. Similarly, At is also a vector, which is the posture information captured by the system at the current moment and is similar to the target action. Due to various interference factors in the actual situation, such as sensor errors and inaccurate action execution, At is usually only close to the target posture T and will not be exactly the same.

[0077] Next, use machine learning algorithms to compare the differences between these two vectors At and T. Specifically, when calculating, the Euclidean distance is used to measure the degree of difference between them, that is, D = ||At - T||. The Euclidean distance is a common vector distance measurement method, which can intuitively reflect the proximity of two vectors in space. The smaller the value of D, the more similar At and T are.

[0078] Then, according to the calculated difference D, a similarity score S is given, and the calculation formula is S = L - D / L. Here, the reference length L is a preset constant, which represents the maximum possible difference between two completely dissimilar postures in theory. For example, in a limited dynamic space, L can be the Euclidean distance between the two postures with the largest difference among all possible postures. Through such calculations, the value range of the similarity score S is between [0, 1]. S approaching 1 indicates a high similarity between At and T; S approaching 0 indicates a low similarity.

[0079] Finally, set a threshold η. η is a value between 0 and 1, and it is the critical value for judging whether to improve the capture accuracy. If the calculated similarity score S < η, it means that the similarity between the currently captured approximate posture At and the target posture T is low. At this time, the system will automatically trigger a higher-precision tracking mode. Through such control conditions, when it is detected that the captured posture is significantly different from the target posture, the system can quickly switch to a higher-performance working state, invest more computing resources or use a more refined algorithm for tracking, thereby improving the detail capture quality and technical stability to ensure that the key features of the target action can be captured more accurately.

[0080] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. Biomechanical motion capture system, including: A high-precision sensor (2), a data acquisition module, a processing module, an analysis module, and a regulation module, characterized in that the data acquisition module acquires multi-dimensional force and position data of a subject during movement and synchronously records them in real time through the high-precision sensor (2); The processing module preliminarily processes the acquired data through a filtering algorithm to reduce noise interference caused by the environment and equipment; The analysis module performs multi-scale analysis on the processed data and extracts characteristic parameters reflecting subtle movement changes to improve the capture accuracy; The regulation module optimizes the feedback control system based on the above characteristic parameters, and regulates the signal purity and motion capture error in real time to ensure the accuracy and stability of biomechanical data.

2. The biomechanical motion capture system according to claim 1, characterized in that: Optimizing the feedback control system based on the characteristic parameters reflecting subtle movement changes extracted by the above analysis module further includes: Dividing the signal purity into several intervals and setting different correction factors K for each interval; Calculating the current motion capture error E and evaluating the signal purity S; Dynamically adjusting the correction factor based on the following formula: K = max(Sα * E + Dβ, 0), where S represents the signal purity, E represents the motion capture error, α represents the influence factor with a value range of [0, 1], β is the influence factor of D, and the data stability index D can be obtained by calculating the variance of data over a period of time. The smaller the variance, the more stable the data, and the larger the D value; When K is greater than the set threshold KT, the automatic noise adjustment module is triggered for processing to ensure the stability and purity of the system's real-time regulation of biomechanical data, and to solve the problems of large noise components and motion capture errors in biomechanical data.

3. The biomechanical motion capture system according to claim 2, characterized in that: Noise regulation and accuracy improvement of biomechanical data further include the following steps: Obtaining the base frequency Fb collected by the high-precision sensor (2); Determining the maximum noise frequency Fn to be filtered according to the signal fluctuation degree; Using Fourier transform to separate the main signal, and screening and retaining the effective signal segments through the following logical formula: if ω > Fn and A < Am, then remove this signal segment; where ω represents the frequency component, Am represents the maximum amplitude, and A represents the amplitude of each component; Constructing a new time series data based on the effective signal to reduce interference caused by the environment and equipment and ensure data accuracy.

4. The biomechanical motion capture system according to claim 3, characterized in that: Improving the capture accuracy based on the extraction of characteristic parameters of subtle movement changes is specifically reflected in: Selecting a time slice t of a specific motion stage; Recording the multi-dimensional position coordinate set P(i) of all subjects and the corresponding acting force F(i) during this time period; Calculate the velocity V = dP2 / dt2 and angular acceleration at each moment Using the following conditions to judge whether subtle movement changes are captured: if V2 + βA ≥ γ, high-frequency sampling; β represents the proportional gain, V represents the velocity, and A represents the acceleration.

5. The biomechanical motion capture system according to claim 4, characterized in that: The method executed to better solve the problem of impure signals is: In the system initialization stage, initial weights W0 are defined for each noise source respectively, and the initial weights W0 are used to represent the estimated degree of the influence of each noise source on the signal in the initial state; During the operation of the system, dynamically track the actual influence of each noise source on the signal and obtain the real-time weight Wt corresponding to each noise source; At the same time, an ideal signal purity standard I is set, where the ideal signal purity standard I is a predetermined target value of the desired signal purity, and the purity of the current signal is measured in each cycle to obtain the current state C; The weight is adjusted by the following iterative formula: Wi+1=Wi+(Ci-I)×Q, where Wi is the weight value calculated in the i-th step, Q is the learning rate, and its value range determines the speed of weight adjustment. Ci is the current pure state measured in the i-th cycle.

6. The biomechanical motion capture system according to claim 5, characterized in that: After the weight value Wi is calculated using the iterative formula, the comparison module in the system compares Wi with the maximum weight limit value W_Lim_max pre-stored in the system parameter table. If Wi is greater than W_Lim_max, the weight adjustment execution module in the system assigns Wi to W_Lim_max. This operation ensures that the system's self-correction mechanism can stably perform the noise reduction function. The maximum weight limit value W_Lim_max is a threshold set to prevent the weight value from increasing excessively during the adjustment process, causing system instability.

7. The biomechanical motion capture system according to claim 6, characterized in that: The specific implementation method of optimizing capture accuracy is as follows: The goal is to set the key posture T that captures the target action; In actual operation, the approximate posture At is continuously obtained; Use machine learning algorithms to compare the difference between two vectors D = ||At-T|| and give a similarity score S = LD / L, where L represents the reference length; If S<η, η is the threshold, the system automatically triggers a more precise tracking mode. Through this control condition, it can quickly switch to a higher performance working state when the similarity is too low, thereby improving the quality of detail capture and technical stability.

8. A biomechanical motion capture force platform, using the biomechanical motion capture system described in claim 7, characterized in that: The invention comprises a high-precision sensor (2), a top plate (1) and a protective base (3), wherein the top plate (1) is fixedly mounted on the top of the protective base (3), the high-precision sensor (2) is arranged on the bottom surface of the top plate (1), and the acquisition module, the processing module, the analysis module and the control module are arranged between the top plate (1) and the protective base (3).

Images