Physical training test method and system

Abstract

The invention provides a physical training test method and system, and the method comprises the steps: synchronously obtaining motion capture data through a multi-modal data collection device, the motion capture data comprising optical motion data or surface electromyogram signals; performing data preprocessing on the original motion capture data, wherein the data preprocessing comprises coordinate system unification processing and timestamp calibration processing; performing multi-stage data validity verification processing on the preprocessed motion capture data to obtain a verification result; in response to the verification result, representing that the data is valid, when the motion capture data is optical motion data, mapping the motion capture data into joint kinematics parameters based on a biomechanical model, and generating physical fitness motion evaluation parameters according to the joint kinematics parameters; according to the invention, scientificity and effectiveness of physical training can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of physical training variable testing, and more particularly, to a physical training testing method and system. Background Art

[0002] In the field of physical training variable testing, there are two main technical approaches. Dynamic analysis solutions based on wearable sensors suffer from imperfect data verification mechanisms. When sensors lose connection or experience signal interference, the system cannot effectively identify and filter abnormal data, resulting in erroneous data entering the computational pipeline. These solutions also overly rely on the evaluation of a single dynamic parameter and lack the ability to analyze key kinematic characteristics such as joint angle deviation. This makes it difficult to identify movement deformations such as knee buckling and spinal hyperextension that occur in athletes during compound movements like squats and bench presses.

[0003] While computer vision-based optical motion capture technology can capture joint motion trajectories, it has significant shortcomings in multimodal data fusion. When the system simultaneously collects electromyographic signals and optical data, the lack of a precise timestamp synchronization mechanism causes timing misalignment between the data streams collected by different sensors, seriously affecting the accuracy of cross-modal data analysis. Furthermore, existing optical solutions only focus on verifying the integrity of single-frame data and lack effective detection methods for data interruptions during continuous motion (such as the barbell stalling phase during weightlifting).

[0004] Existing technologies suffer from three common flaws: First, the data verification system is overly simplistic, performing only basic threshold determinations and failing to establish a progressive, multi-level verification process; second, the biomechanical parameter conversion algorithm is too rudimentary to accurately map raw data to joint kinematic parameters; and finally, the system lacks the ability to deeply mine historical assessment data, making it impossible to establish an early warning mechanism for movement normative trends. These issues severely restrict the scientific nature and reliability of sports training assessment. Summary of the Invention

[0005] The purpose of this application is to provide a physical training test method and system, which has the advantage of constructing a progressive multi-level verification process to improve the accuracy of data validity judgment and avoid erroneous data from entering subsequent processing links.

[0006] The present application provides a physical training test method, and the technical solution is as follows: a physical training test method, characterized in that it includes: synchronously acquiring motion capture data through a multimodal data acquisition device, the motion capture data including optical motion data or surface electromyography signals; performing data preprocessing on the original motion capture data, the data preprocessing including coordinate system normalization processing and timestamp calibration processing; performing multi-level data validity verification processing on the preprocessed motion capture data to obtain a verification result; in response to the verification result indicating that the data is valid, when the motion capture data is optical motion data, mapping the motion capture data into joint kinematic parameters based on a biomechanical model, and generating physical motion evaluation parameters based on the joint kinematic parameters; wherein the multi-level data validity verification processing includes: primary verification: verifying that the total number of data frames of the motion capture data reaches a preset frame number threshold; intermediate verification: verifying that the time interval between adjacent data frames is less than or equal to a preset time threshold; advanced verification: confirming that the proportion of invalid data points in the motion capture data is lower than a preset proportion threshold, and the length of continuous valid data segments is greater than a preset continuous frame threshold; the multi-level data validity verification processing adopts progressive verification logic, and terminates subsequent verification when the current level verification fails.

[0007] Furthermore, the present application also proposes that generating physical action evaluation parameters includes: performing frame processing on motion capture data to obtain an action frame sequence; for each action frame in the action frame sequence, performing: verifying that the coverage rate of the human body motion area boundary box is greater than a preset coverage rate threshold; performing bone joint positioning processing to obtain a bone joint position coordinate set; generating motion trajectory data based on the bone joint position coordinate set; performing speed analysis processing on the motion trajectory data to obtain an instantaneous speed vector; performing angle deviation calculation processing on the motion trajectory data to obtain a joint angle deviation value; generating dynamic indicators based on the instantaneous speed vector and the joint angle deviation value; screening action frames whose joint angle deviation values ​​are less than a preset angle threshold as key action frames; performing feature extraction processing on the key action frames to obtain motion feature data; and generating physical action evaluation parameters based on the motion feature data.

[0008] Furthermore, the present application also proposes that the speed analysis processing includes: performing time differential processing on the motion trajectory data to generate an original speed sequence; performing filtering processing on the original speed sequence to generate a smoothed speed sequence; performing acceleration calculation processing on the smoothed speed sequence to generate an acceleration sequence; extracting the maximum acceleration value of the acceleration sequence as an explosive force indicator; and determining the smoothed speed sequence as an instantaneous speed vector.

[0009] Furthermore, the present application also proposes that the feature extraction processing includes: performing skeleton point detection processing on key action frames to obtain skeleton point coordinates; performing timing alignment processing on the skeleton point coordinates to generate an aligned skeleton sequence; performing electromyographic signal integration processing on the aligned skeleton sequence to generate muscle activation timing data; and determining the muscle activation timing data as motion feature data.

[0010] Furthermore, the present application also proposes that, when the motion capture data is a surface electromyographic signal: the data validity verification process includes calculating the ratio of signal power to noise power; generating physical movement evaluation parameters based on the motion capture data includes: performing baseline correction processing and power frequency filtering processing on the surface electromyographic signal to obtain a clean electromyographic signal; performing feature extraction processing on the clean electromyographic signal to generate an electromyographic feature sequence; performing fatigue calculation processing on the electromyographic feature sequence to obtain a muscle fatigue index; and determining the muscle fatigue index as a physical movement evaluation parameter.

[0011] Furthermore, the present application also proposes that the feature extraction processing also includes: performing lower limb joint detection processing on key action frames to obtain a lower limb bone joint position coordinate set, the lower limb bone joint position coordinate set includes hip joint coordinates, knee joint coordinates and ankle joint coordinates; performing cross-frame trajectory alignment processing on the lower limb bone joint position coordinate set to generate a lower limb motion trajectory; performing speed analysis processing on the lower limb motion trajectory: calculating the displacement change per unit time to generate a lower limb instantaneous speed sequence; performing smoothing processing on the lower limb instantaneous speed sequence; extracting the ratio of the acceleration segment duration to the deceleration segment duration; integrating the explosive power index and the ratio of the acceleration segment duration to the deceleration segment duration into motion feature data.

[0012] Furthermore, the present application also proposes that before executing the acquisition of motion capture data, it also includes: receiving a test start instruction containing a geographic location identifier; obtaining the current device location coordinates based on a positioning device; calculating the Euclidean distance between the current device location coordinates and the center coordinates of a preset training area; verifying that the Euclidean distance is less than or equal to a preset area radius threshold; in response to the verification being passed, activating a data acquisition device that matches the test type and sending an acquisition parameter configuration instruction.

[0013] Furthermore, the present application also proposes that obtaining motion capture data includes: collecting three-dimensional acceleration and angular velocity data through an inertial measurement unit; obtaining motion capture data includes: collecting three-dimensional coordinate data of infrared reflective marker points through an optical motion capture system; obtaining motion capture data includes: performing continuous inter-frame motion vector analysis based on a video acquisition device to generate a two-dimensional motion trajectory.

[0014] Furthermore, the present application also proposes that it also includes: establishing a historical evaluation database that stores athlete identification, test time and physical movement evaluation parameters; executing the currently generated physical movement evaluation parameters with the historical data set with the same athlete identification: calculating the time series change rate of the joint angle deviation value, and generating a movement deformation warning when the change rate exceeds the preset change threshold; performing linear regression modeling on the explosive power index, and outputting the regression coefficient sign change detection result; calculating the recommended recovery period based on the exponential attenuation curve of the muscle fatigue index; generating a biomechanical report containing the joint angle deviation value change rate, the regression coefficient sign change detection result and the recommended recovery period.

[0015] Furthermore, the present application also proposes a physical training test system, characterized in that it includes: a data acquisition module for acquiring motion capture data; a data verification module for performing data validity verification processing on the motion capture data to obtain a verification result; a core processing module for generating physical movement evaluation parameters based on the motion capture data in response to the verification result indicating that the data is valid; wherein the data acquisition module includes: an inertial measurement unit interface, configured to receive three-dimensional motion data; an optical capture parser, configured to process infrared marker point coordinates; and a video analysis engine, configured to calculate inter-frame motion vectors.

[0016] From the above, it can be seen that the physical training test method and system provided by this application constructs a progressive verification process through multi-level data validity verification processing, and combines the biomechanical model to achieve accurate mapping of joint kinematic parameters, which has the advantages of improving the accuracy of data validity judgment and avoiding erroneous data from entering subsequent processing links. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which: Figure 1 A flow chart of a physical fitness training test method provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0018] The technical solutions of this application will be described clearly and completely below, in conjunction with the accompanying drawings. It should be understood that the described embodiments represent only a portion of the embodiments of this application, and not all of them. The components of this application, generally described and illustrated in the drawings herein, may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but rather merely represents selected embodiments of this application. All other embodiments derived by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used solely to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0019] In traditional existing physical training and testing systems, the data verification mechanism uses only a single-level basic threshold judgment, which cannot effectively identify abnormal interruptions or sensor failures in motion capture data, resulting in invalid data segments entering the subsequent calculation chain. Multimodal data collected by different sensors lack a unified spatiotemporal reference, resulting in coordinate system deviations and timing misalignments when cross-device data is fused, affecting the calculation accuracy of biomechanical parameters. The conversion process from raw data to joint motion parameters relies on simple linear calculations, and no kinematic feature mapping based on the biomechanical model has been established, making it difficult to accurately reflect movement normative indicators.

[0020] For example, when using an optical motion capture system to evaluate squat movements, ten consecutive frames of data were lost due to limb occlusion of the infrared markers. The existing verification mechanism only checked whether the total number of frames met the requirements, but did not verify the length of the continuous valid data segment, and mistakenly input the motion trajectory containing data holes into the calculation module. During the surface electromyography signal acquisition process, poor sensor contact caused a sudden drop in signal power. Traditional methods did not calculate the signal-to-noise ratio threshold, resulting in the electromyography feature sequence containing high-frequency interference. During multimodal data fusion, due to the lack of timestamp calibration processing, there was a millisecond-level timing deviation between the optical data and the electromyography signal, which made the correlation analysis between joint angles and muscle activation states invalid.

[0021] If these issues are not addressed, invalid data segments will lead to peak errors in the calculation of joint kinematic parameters, causing motion assessment parameters to deviate from the true biomechanical state. Inconsistent spatiotemporal benchmarks across modal data can cause distortion in multi-source information fusion, reducing the credibility of physical fitness test results. Parameter mapping without biomechanical model support cannot capture deformation characteristics of movements such as knee buckling, affecting the accuracy of sports injury warning mechanisms. Insufficient data verification granularity will also cause the system to continuously process low-quality data, increasing computing resource consumption and reducing the efficiency of real-time feedback.

[0022] In this regard, the present application proposes a physical training test method, comprising: S1. synchronously acquiring motion capture data through a multimodal data acquisition device, where the motion capture data includes optical motion data or surface electromyography signals; S2. performing data preprocessing on the original motion capture data, the data preprocessing including coordinate system normalization and time stamp calibration; S3, performing multi-level data validity verification processing on the pre-processed motion capture data to obtain a verification result; S4. In response to the verification result indicating that the data is valid, when the motion capture data is optical motion data, mapping the motion capture data into joint kinematic parameters based on a biomechanical model, and generating physical motion evaluation parameters based on the joint kinematic parameters; Among them, the multi-level data validity verification processing includes: primary verification: verifying that the total number of data frames of the motion capture data reaches the preset frame number threshold; intermediate verification: verifying that the time interval between adjacent data frames is less than or equal to the preset time threshold; advanced verification: confirming that the proportion of invalid data points in the motion capture data is lower than the preset proportion threshold, and the length of the continuous valid data segment is greater than the preset continuous frame threshold; the multi-level data validity verification processing adopts progressive verification logic, and terminates subsequent verification when the current level verification fails.

[0023] A multimodal data acquisition device refers to a hardware device capable of simultaneously collecting different types of motion data. Specifically, this can be implemented using a combination of an inertial measurement unit, an optical motion capture system, or a surface electromyography sensor. This device is used to synchronously acquire optical motion data or surface electromyography signals, addressing the dimensionality issues of a single data source. Coordinate system normalization involves converting data collected by different sensors into a unified spatial coordinate system. This can be achieved using a homogeneous coordinate transformation matrix to eliminate positioning errors caused by differences in the coordinate systems of different devices. Timestamp calibration involves aligning the time series of multi-source data. This can be achieved using a network time protocol or hardware synchronization signal trigger to ensure temporal consistency across modalities. Multi-level data validation refers to a hierarchical data quality check mechanism executed in a pre-set logical sequence. This can be implemented using a progressive process consisting of primary validation to determine the total amount of data, intermediate validation to check the sampling interval, and advanced validation to assess data continuity, preventing invalid data from entering subsequent computational processes. A biomechanical model refers to a joint kinematic parameter calculation model based on human anatomy. This can be implemented using an inverse kinematics algorithm or rigid-body dynamics equations to map raw motion data into biomechanical parameters such as joint angles and angular velocities.

[0024] This application combines a progressive multi-level data verification mechanism with a biomechanical model to build a full-link quality control system from raw data acquisition to biomechanical parameter generation. The multi-level verification mechanism filters invalid data segments layer by layer through progressive screening of primary, intermediate, and advanced verifications, solving the problem of coarse granularity in verification of existing technologies. At the same time, the application of biomechanical models realizes the precise conversion of raw data into joint kinematic parameters, making up for the lack of evaluation dimensions caused by the traditional solution's reliance on simple calculations.

[0025] This application uses a multimodal data acquisition device to synchronously acquire motion capture data, including optical motion data or surface electromyography signals. Data preprocessing is performed on the raw motion capture data, including coordinate system normalization and timestamp calibration. Coordinate system normalization converts data collected by different devices into a unified coordinate system, eliminating positioning errors caused by coordinate system differences. Timestamp calibration aligns the timing of data across devices, overcoming the problem of multi-source data fusion failure.

[0026] Multi-level data validity verification is performed on the pre-processed motion capture data. Primary verification verifies whether the total number of data frames reaches the preset frame number threshold to ensure that the data collection volume meets the analysis requirements. Intermediate verification checks whether the time interval between adjacent data frames is less than or equal to the preset time threshold to eliminate sampling rate anomalies. Advanced verification confirms whether the proportion of invalid data points is lower than the preset proportion threshold and whether the length of continuous valid data segments is greater than the preset continuous frame threshold, effectively identifying invalid data segments caused by sensor loss or motion interruption. Multi-level verification uses progressive logic, terminating subsequent verification when the current level verification fails, preventing invalid data from entering the computing chain.

[0027] When the validation results indicate valid data representation and the motion capture data is optical, the motion capture data is mapped to joint kinematic parameters based on a biomechanical model. This biomechanical model considers the effects of multi-joint linkage, enabling precise conversion of raw data into biomechanical parameters. Physical performance evaluation parameters are generated from these joint kinematic parameters, integrating kinematic features such as joint angle deviation to detect motion deformation.

[0028] As a preferred embodiment, the solution of this application is specifically implemented as follows: The multimodal data acquisition device includes an optical motion capture system and a surface electromyography (SEM) acquisition device. The optical motion capture system, composed of multiple high-speed cameras, captures three-dimensional motion data by tracking reflective markers attached to key points on the human body. The SEM acquisition device, composed of multi-channel EMG sensors, collects muscle surface potential signals.

[0029] During data preprocessing, coordinate normalization is first performed. The coordinates of the marker points collected by the optical system are converted from the camera coordinate system to the global coordinate system to eliminate differences in camera perspectives. Surface electromyographic signals are spatially filtered to remove baseline drift caused by skin impedance variations. Timestamp calibration is then performed to align the sampling times of the optical and electromyographic data with the system clock, ensuring temporal consistency across multiple sources.

[0030] The multi-level data validity verification process is divided into three stages. Primary verification checks whether the total number of data frames meets the preset minimum frame number requirement, for example, no less than 1000 frames. Intermediate verification calculates the time interval between adjacent frames to verify whether it is less than or equal to a preset threshold, such as 10 milliseconds. Advanced verification first counts the number of invalid data points in each frame and calculates whether the proportion of invalid points is lower than a preset threshold, such as 5%. It then checks whether the length of the continuous valid data segment is greater than the preset continuous frame threshold, such as 100 frames. The verification process uses progressive logic. If any level of verification fails, the subsequent verification is terminated and the data invalid result is returned.

[0031] If verification passes and the data type is optical motion data, parameter mapping is performed based on the biomechanical model. First, a human skeletal chain model is constructed, defining joint degrees of freedom and motion constraints. Then, using an inverse kinematics algorithm, the 3D coordinates of the marker points are converted into a joint angle sequence. Angular velocity and angular acceleration are calculated based on the joint angle sequence to obtain a complete set of joint kinematic parameters. Finally, physical performance evaluation metrics, such as joint range of motion and coordination, are generated based on the kinematic parameters.

[0032] Through the above scheme, this application has constructed a multi-level data verification system to effectively identify abnormal data at different levels. The progressive verification mechanism can distinguish between occasional noise and systematic failures, and prevent invalid data segments from entering subsequent calculation links. The coordinate system normalization and timestamp calibration in the data preprocessing stage ensure the accuracy of multimodal data fusion and overcome the problem of inconsistent spatiotemporal benchmarks of data across devices. The parameter mapping based on the biomechanical model breaks through the limitations of traditional linear calculation methods, can accurately reflect the kinematic characteristics of joints, and provides a reliable basis for detecting movement deformation. The overall solution improves the data quality of physical training tests and the accuracy of biomechanical parameter conversion, providing more reliable technical support for movement evaluation.

[0033] The present application further proposes that generating physical action evaluation parameters includes: performing frame segmentation processing on motion capture data to obtain an action frame sequence; for each action frame in the action frame sequence, performing: verifying that the coverage rate of the human body motion area boundary box is greater than a preset coverage rate threshold; performing bone joint positioning processing to obtain a bone joint position coordinate set; generating motion trajectory data based on the bone joint position coordinate set; performing speed analysis processing on the motion trajectory data to obtain an instantaneous speed vector; performing angle deviation calculation processing on the motion trajectory data to obtain a joint angle deviation value; generating dynamic indicators based on the instantaneous speed vector and the joint angle deviation value; screening action frames whose joint angle deviation values ​​are less than a preset angle threshold as key action frames; performing feature extraction processing on the key action frames to obtain motion feature data; and generating physical action evaluation parameters based on the motion feature data.

[0034] Frame segmentation segments continuous motion data into a sequence of discrete action frames with temporal correlation. A preset coverage threshold can be set to ensure that the bounding box covers at least 85% of the human body area. This is verified using an image recognition algorithm to calculate the percentage of valid pixels within the bounding box. Joint localization uses a deep neural network model to output a three-dimensional coordinate set containing key points such as the hip, knee, and ankle, with coordinate accuracy controlled within ±2 mm. Motion trajectory data is generated by calculating the change in Euclidean distance between joint coordinates between adjacent frames to form a time-space displacement curve. Velocity analysis uses the central difference method to calculate instantaneous velocity, for example, calculating the rate of change of displacement within a 10 millisecond time window. A Butterworth low-pass filter is used to eliminate high-frequency noise. Angle deviation calculation compares actual joint angles with ideal angles in a standard motion database and generates deviation values ​​using the cosine similarity algorithm. The preset angle threshold can be set to 5 degrees. Key action frame screening processes the joint angle deviation values ​​of all action frames and retains frames with deviations below the threshold. For example, frames with excessive knee inward angles in a squat are eliminated. The feature extraction process performs principal component analysis on the retained key frames to extract feature vectors such as motion trajectory curvature, velocity change rate, and muscle activation timing.

[0035] After the motion capture data passes multi-level validity verification, frame segmentation converts the continuous data stream into a sequence of discrete action frames, ensuring a clear time base for subsequent processing. Each action frame is first verified for complete coverage of the human body region, eliminating invalid frames due to target deviation. After establishing a three-dimensional spatial reference through high-precision bone and joint positioning, the motion trajectory data generation module captures the continuous displacement changes of joint points to form spatial motion features. The velocity analysis module and the angle deviation calculation module process the motion trajectory data in parallel. The former quantifies the instantaneous dynamic characteristics of the movement, while the latter detects joint kinematic anomalies. Dynamic indices are generated by fusing the velocity vector modulus with the angle deviation value. For example, explosive power is calculated as the product of peak velocity and joint stability coefficient. A key action frame screening mechanism dynamically adjusts the valid data range based on the angle deviation threshold to eliminate noise interference caused by movement deformation. Finally, the feature extraction module integrates multi-dimensional kinematic and dynamic features to generate comprehensive evaluation parameters that include movement standardization, explosive power intensity, and stability. This solution effectively improves the comprehensiveness and accuracy of evaluation parameters through spatiotemporal feature fusion and data cleaning mechanisms.

[0036] As a preferred embodiment, the solution of this application is specifically implemented as follows: The motion capture data is frame-processed to obtain an action frame sequence. For each action frame in the action frame sequence, it is first verified whether the coverage rate of the bounding box of the human motion area is greater than a preset coverage rate threshold.

[0037] Specifically, a deep learning model can be used to identify key points of the human body and extract the three-dimensional coordinates of joints including the head, shoulders, elbows, wrists, hips, knees and ankles.

[0038] Motion trajectory data is generated based on the bone joint position coordinate set. By connecting the corresponding joint points of adjacent frames, a continuous motion trajectory curve is formed.

[0039] Perform velocity analysis on the motion trajectory data to obtain the instantaneous velocity vector. Calculate the displacement change of the joint position between adjacent frames and divide it by the time interval to obtain the instantaneous velocity of each joint point.

[0040] The angle deviation calculation is performed on the motion trajectory data to obtain the joint angle deviation value. Based on the standard action posture, the difference between the joint angle in the actual action and the standard angle is calculated.

[0041] Generate a dynamics index based on the instantaneous velocity vector and joint angle deviation. For example, the peak instantaneous velocity and the mean angle deviation can be combined to form a comprehensive scoring index.

[0042] Filter action frames with joint angle deviation values ​​less than the preset angle threshold as key action frames. You can set the angle threshold to 10 degrees to retain key frames with relatively standard postures.

[0043] Perform feature extraction on key action frames to obtain motion feature data. Extract timing features such as joint position, velocity, acceleration, etc. in key frames.

[0044] Generating physical movement assessment parameters based on motion feature data. Machine learning algorithms can be used to train assessment models based on historical data, outputting quantitative metrics such as movement standardization, explosive power, and stability.

[0045] Through the above technical solutions, this application realizes multi-dimensional motion feature extraction and evaluation. By introducing instantaneous velocity vector analysis, the changing characteristics of the explosive power of the movement are captured. Combined with the calculation of joint angle deviation, the standardization of the movement is fully reflected. By screening key action frames, the interference of invalid data on the evaluation results is effectively reduced. The final generated physical movement evaluation parameters comprehensively consider the standardization, explosive power and stability of the movement, improving the accuracy and comprehensiveness of the evaluation.

[0046] The present application further proposes that the speed analysis processing includes: performing time differentiation processing on the motion trajectory data to generate an original speed sequence; performing filtering processing on the original speed sequence to generate a smoothed speed sequence; performing acceleration calculation processing on the smoothed speed sequence to generate an acceleration sequence; extracting the maximum acceleration value of the acceleration sequence as an explosive force indicator; and determining the smoothed speed sequence as an instantaneous speed vector.

[0047] Among them, the time differential processing generates the original velocity sequence by calculating the ratio of the displacement difference between adjacent time points to the time interval, for example, using the central difference method or the backward difference method, and the time interval can be set to 0.01 seconds to 0.05 seconds. The filtering processing can use a low-pass filter or a moving average filter, and the cutoff frequency is preferably 10Hz to 15Hz to suppress high-frequency noise and retain effective motion characteristics. The acceleration calculation processing is achieved by performing time differentiation on the smoothed velocity sequence again, for example, using a second-order difference algorithm, and the time differential interval is consistent with the velocity sequence generation interval. The maximum acceleration value is extracted by traversing the peak detection of the acceleration sequence, for example, using the sliding window method to identify the local maximum value, and the window length is set to 5 to 10 data points. When the smoothed velocity sequence is used as an instantaneous velocity vector, its data length is consistent with the original velocity sequence to ensure the time sequence correspondence.

[0048] Specifically, motion trajectory data undergoes temporal differentiation to generate a raw velocity sequence, which may contain high-frequency components caused by sensor noise or limb tremor. Low-pass filtering is used to remove noise components above a preset cutoff frequency. For example, a Butterworth filter with a 12Hz cutoff frequency is used to process velocity data sampled at 100Hz, eliminating abnormal fluctuations. The smoothed velocity sequence is then secondarily differentiated to generate an acceleration sequence. Because raw velocity fluctuations are suppressed, no sudden changes are introduced into the acceleration calculation. Maximum acceleration is extracted during the peak phase of power output during the movement, such as capturing the extreme acceleration at takeoff in a vertical jump. Finally, the filtered velocity sequence is output as the instantaneous velocity vector, preserving the dynamic characteristics of the motion trajectory while providing a stable input for subsequent joint angle deviation calculations. This processing flow, coordinated with prior data validation, ensures that analysis is performed only on consecutive valid data segments, avoiding filter edge effects caused by data interruptions, thereby systematically improving the accuracy and calculation stability of explosive power metrics.

[0049] As a preferred embodiment, time differentiation processing is performed on the motion trajectory data to generate an original velocity sequence. Specifically, the central difference method can be used to calculate the displacement change rate between adjacent time points to obtain a velocity vector sequence containing three directional components: x, y, and z.

[0050] Furthermore, a filtering process is performed on the original velocity sequence to generate a smoothed velocity sequence. For example, a Butterworth low-pass filter can be applied with a cutoff frequency set to 10 Hz to remove high-frequency noise and retain the main features of the motion.

[0051] Thus, the acceleration sequence is generated by performing acceleration calculation processing on the smoothed velocity sequence. In specific implementation, a second-order central difference method can be used to calculate the rate of change of velocity at adjacent time points to obtain the acceleration sequence.

[0052] The maximum acceleration value of the acceleration sequence is extracted as the explosive force indicator. This can be achieved by traversing the acceleration sequence, comparing and recording the maximum value.

[0053] Finally, the smoothed velocity sequence is determined as the instantaneous velocity vector. This step directly uses the filtered velocity data without additional calculation.

[0054] Through the above technical solution, the present application can effectively eliminate high-frequency noise and abnormal fluctuations in the original velocity sequence, improving the accuracy of the instantaneous velocity vector. At the same time, the acceleration is calculated based on the smoothed velocity sequence, reducing the acceleration mutation error caused by data noise, thereby improving the reliability of the explosive power index. In addition, through multi-stage data processing, this solution not only retains the temporal characteristics of the motion trajectory, but also improves the stability of the subsequent dynamic index calculation, providing a more accurate and reliable data basis for physical training evaluation.

[0055] The present application further proposes that feature extraction processing includes: performing skeleton point detection processing on key action frames to obtain skeleton point coordinates; performing timing alignment processing on the skeleton point coordinates to generate an aligned skeleton sequence; performing electromyographic signal integration processing on the aligned skeleton sequence to generate muscle activation timing data; and determining the muscle activation timing data as motion feature data.

[0056] Among them, the skeleton point detection processing can adopt a posture estimation algorithm based on convolutional neural network to locate the two-dimensional or three-dimensional coordinates of the hip joint and knee joint by regressing the key point heat map. The timing alignment processing eliminates the inter-frame time offset of the skeleton point coordinates through the dynamic time warping algorithm, for example, limiting the displacement change of the skeleton points in adjacent frames to the preset time window length. The electromyographic signal integration processing uses the aligned skeleton sequence as the time reference, performs a sliding window integration calculation on the synchronously collected electromyographic signals, and sets the integration period to 50 milliseconds to 200 milliseconds to quantify the activation intensity of the rectus femoris or vastus lateralis. The time reference of the skeleton point coordinates and the electromyographic signal is synchronized through a unified timestamp to ensure the timing matching of the muscle activation state and the skeleton motion trajectory during the integration processing.

[0057] Skeletal point detection extracts coordinate data of the human lower limb joints from key action frames, providing a spatial positioning basis for subsequent analysis. Timing alignment corrects for timing deviations in skeletal point coordinates through interpolation compensation or kinematic constraint models. For example, cubic spline interpolation is used to fill in the coordinate data of missing frames, ensuring continuous motion trajectories across frames. Myoelectric signal integration combines skeletal motion trajectories with muscle electrophysiological signals. By calculating the area integral of the myoelectric signal within a preset time window, a timing curve reflecting the degree of muscle activation is generated. Muscle activation timing data incorporates both the spatial characteristics of the skeletal motion trajectory and the temporal characteristics of muscle co-activation. For example, in a squat, changes in knee joint angle and quadriceps femoris activation intensity show a temporal positive correlation. The resulting physical movement assessment parameters can accurately quantify movement standardization, detect deforming movements such as knee buckling, and enhance the biomechanical credibility of the assessment results.

[0058] Preferably, the feature extraction process includes the following steps: first, a skeleton point detection process is performed on the key action frame to obtain the coordinates of the skeleton points. Specifically, a deep learning model can be used to locate the skeleton points of the key action frame and extract the two-dimensional or three-dimensional coordinate information of the main joints of the human body. Secondly, a time alignment process is performed on the skeleton point coordinates to generate an aligned skeleton sequence. For example, a dynamic time warping algorithm can be used to align the coordinates of the skeleton points at different time points to eliminate the problem of inconsistent timing caused by differences in sampling frequency or movement speed. Thirdly, an electromyographic signal integration process is performed on the aligned skeleton sequence to generate muscle activation timing data. Specifically, the synchronously collected surface electromyographic signal can be time-matched with the aligned skeleton sequence, and then the electromyographic signal can be rectified and smoothed, and the integral value of its envelope is calculated as a quantitative indicator of the degree of muscle activation. Finally, the muscle activation timing data is determined as motion feature data. Thus, the obtained motion feature data contains both the spatial information of skeletal movement and the temporal characteristics of muscle contraction.

[0059] Through the above technical solutions, this application solves the problems of timing consistency and muscle activation state quantification in motion feature data extraction. Through skeletal point detection, timing alignment, and electromyographic signal integration processing, dynamic correlation analysis between skeletal motion trajectory and muscle force state is achieved. This multi-dimensional biomechanical feature extraction method improves the accuracy and comprehensiveness of physical movement evaluation parameters, providing a more reliable data foundation for subsequent movement quality evaluation and training guidance.

[0060] The present application further proposes that when the motion capture data is a surface electromyographic signal, the data validity verification process includes calculating the ratio of signal power to noise power; generating physical movement evaluation parameters based on the motion capture data includes performing baseline correction processing and industrial frequency filtering processing on the surface electromyographic signal to obtain a clean electromyographic signal, performing feature extraction processing on the clean electromyographic signal to generate an electromyographic feature sequence, performing fatigue calculation processing on the electromyographic feature sequence to obtain a muscle fatigue index, and determining the muscle fatigue index as the physical movement evaluation parameter.

[0061] During the data validity verification stage, the ratio of signal power to noise power can be calculated using a sliding window method. For example, the local signal-to-noise ratio is calculated with a window length of 200 milliseconds. When the ratio is lower than the preset threshold, the signal quality is determined to be substandard. The baseline correction process eliminates the zero-point drift of the device through a moving average filter, such as a high-pass filter with a time constant of 0.5 seconds. The power frequency filtering process can be configured as a 50Hz or 60Hz notch filter, which is selected according to the frequency of the local power system. The feature extraction process can use the root mean square value calculation method, such as calculating the root mean square value sequence of the electromyographic signal within a 500 millisecond time window. The fatigue calculation process can establish a model based on the median frequency offset. For example, when the median frequency decreases by more than 15% of the baseline value, it is determined that the muscle has entered a fatigue state.

[0062] Specifically, the surface EMG signal is first verified for signal-to-noise ratio. When the ratio of signal power to noise power reaches a preset threshold, it enters the data processing process. Baseline correction eliminates the DC offset component in the signal, ensuring that the EMG signal fluctuates around a zero baseline. Power frequency filtering performs frequency domain filtering to address power system interference, retaining the effective components of the EMG signal between 20 and 450 Hz. The clean EMG signal undergoes time domain feature extraction to generate a characteristic sequence representing the intensity of muscle activation. This sequence is input into the fatigue calculation model, and the muscle fatigue index is generated by analyzing the amplitude attenuation slope or spectral migration. This index can quantitatively reflect the dynamic changes in muscle contraction ability during exercise. For example, when the fatigue index reaches 0.7, it indicates that the muscle has entered the compensatory contraction stage. While ensuring signal quality, the entire process achieves accurate assessment of muscle functional status through multi-stage signal processing and feature modeling.

[0063] As a preferred embodiment, when the motion capture data is a surface electromyography signal, the ratio of the signal power to the noise power is first calculated to perform data validity verification processing. Specifically, the short-time Fourier transform method can be used to perform time-frequency analysis on the original electromyography signal, extract the power spectral density of the signal main frequency band and the background noise frequency band, and calculate the ratio of the two as a signal-to-noise ratio index. For example, the 20-450Hz frequency band can be used as the signal main frequency band, and the 0-20Hz and 450-500Hz frequency bands can be used as the background noise frequency band. When the signal-to-noise ratio index is higher than a preset threshold, the data is determined to be valid.

[0064] For validated surface EMG signals, baseline correction and power frequency filtering are further performed to obtain clean EMG signals. Baseline correction uses a high-pass filter to remove low-frequency drift, with a cutoff frequency set to 10Hz. Power frequency filtering uses a notch filter to eliminate 50Hz / 60Hz power frequency interference.

[0065] Perform feature extraction on the clean EMG signal to generate an EMG feature sequence. A sliding window method can be used with a window length of 256ms and a 50% overlap. Within each window, time domain features (such as RMS value and mean absolute value) and frequency domain features (such as median frequency and mean power frequency) are extracted.

[0066] Fatigue calculations are performed on the EMG feature sequence to generate a muscle fatigue index. A fatigue assessment model can be established based on the changing trend of the median frequency. For example, a linear regression method can be used to fit the slope of the median frequency over time, and the absolute value of the slope can be used as the muscle fatigue index.

[0067] Finally, the calculated muscle fatigue index is determined as a physical movement evaluation parameter for subsequent sports performance analysis and training guidance.

[0068] Through the above technical solutions, this application realizes a specialized processing flow for surface electromyographic signals. Signal quality degradation problems are effectively identified through signal-to-noise ratio calculation, and baseline correction and power frequency filtering ensure the acquisition of high-quality electromyographic signals. Feature extraction and fatigue calculation make full use of the time-frequency characteristics of electromyographic signals to generate evaluation parameters that can objectively reflect the functional status of muscles. This multi-dimensional signal processing and feature analysis method significantly improves the accuracy and reliability of physical movement assessment based on electromyographic signals, and provides a more refined and personalized guidance basis for sports training.

[0069] The present application further proposes performing lower limb joint detection processing on key action frames to obtain a lower limb bone joint position coordinate set, which includes hip joint coordinates, knee joint coordinates and ankle joint coordinates; performing cross-frame trajectory alignment processing on the lower limb bone joint position coordinate set to generate a lower limb motion trajectory; performing speed analysis processing on the lower limb motion trajectory: calculating the displacement change per unit time to generate a lower limb instantaneous speed sequence; performing smoothing processing on the lower limb instantaneous speed sequence; extracting the ratio of the acceleration segment duration to the deceleration segment duration; integrating the explosive power index and the ratio of the acceleration segment duration to the deceleration segment duration into motion feature data.

[0070] Lower limb joint detection processing can adopt a joint point regression model based on deep learning, for example, through convolutional neural networks to identify the two-dimensional or three-dimensional coordinates of the hip, knee, and ankle joints from key action frames. Cross-frame trajectory alignment processing can adopt a time series interpolation algorithm to perform linear or nonlinear interpolation on the coordinate data between adjacent frames to eliminate trajectory breaks caused by sampling rate fluctuations. In the velocity analysis processing, the unit time can be set to 10 milliseconds to 50 milliseconds, and the displacement change is obtained by calculating the Euclidean distance difference between the coordinates of two adjacent frames. Smoothing processing can use a sliding average filter, and the window length is set to 3 to 5 data points. The ratio of the duration of the acceleration segment to the duration of the deceleration segment can be divided by the sign change point of the first-order derivative of the velocity curve. For example, when the acceleration changes from positive to negative, it is determined to be the starting point of the deceleration segment.

[0071] From the action frame sequence generated by the frame segmentation process, frames with joint angle deviations less than a preset angle threshold are selected as key action frames. Lower-limb joint detection extracts coordinate data for the hip, knee, and ankle joints, forming a coordinate set containing spatial position information. Cross-frame trajectory alignment connects the discrete joint coordinates in chronological order to generate a continuous lower-limb motion trajectory curve. Velocity analysis calculates the quotient of the joint displacement difference and the time interval between adjacent frames to generate a raw instantaneous velocity sequence. This is then filtered using a Gaussian filter to eliminate velocity fluctuations caused by sensor noise. The smoothed velocity sequence is then used to identify the demarcation point between the acceleration and deceleration phases by finding local maxima. For example, in a vertical jump, the ratio of the acceleration phase during takeoff to the deceleration phase during landing can be quantified to range from 1.5:1 to 2.5:1. Finally, the explosive power index and this ratio are input into the evaluation model to generate a composite feature vector reflecting lower-limb explosive power and movement coordination. By quantifying the duration ratio of the acceleration and deceleration phases, it is possible to identify asymmetries in motion control, such as the difference in the collaborative working state of the hip extensor group during basketball takeoff and the knee flexor group during landing, thereby improving the biomechanical characterization ability of motion feature data.

[0072] As a preferred embodiment, the feature extraction process also includes performing lower limb joint detection on key action frames. This process can be implemented using a deep learning model, such as using a convolutional neural network to perform semantic segmentation on the image and identify the positions of the hip, knee, and ankle joints. The obtained lower limb joint position coordinate set contains the three-dimensional spatial coordinates of these three joints.

[0073] Perform cross-frame trajectory alignment on the lower limb joint position coordinates. This step can use the Kalman filter algorithm to predict the possible position of each joint in the next frame and fuse it with the actual detection results to generate a smooth and continuous lower limb motion trajectory.

[0074] Velocity analysis is performed on the generated lower limb motion trajectory. First, the displacement change per unit time is calculated to generate a lower limb instantaneous velocity sequence. A 10 millisecond time interval can be selected as the unit time interval. The instantaneous velocity is calculated by dividing the position difference between two adjacent time points by the time interval.

[0075] Then, the lower limb instantaneous velocity series is smoothed. A sliding average filter can be used, with a window size of 5 data points selected, and the average velocity within the window is calculated as the smoothed velocity at the current moment.

[0076] The ratio of the duration of the acceleration segment to the duration of the deceleration segment can be extracted by judging the changing trend of the smoothed speed sequence, defining the stage of speed increase as the acceleration segment and the stage of speed decrease as the deceleration segment, accumulating their durations respectively, and calculating the ratio.

[0077] The explosive power index and the ratio of the acceleration duration to the deceleration duration are combined to form the motion feature data. The peak acceleration value can be selected as the explosive power index, and together with the aforementioned ratio, it forms a two-dimensional feature vector as the final motion feature data.

[0078] Through the above technical solution, this application realizes the refined analysis of lower limb joint movement. As a result, it can effectively capture the dynamic changes of lower limb joints in continuous movements and improve the accuracy of the assessment of explosive power indicators and movement coordination.

[0079] The present application further proposes that before executing the acquisition of motion capture data, a test start instruction containing a geographic location identifier is received; the current device location coordinates are obtained based on a positioning device; the Euclidean distance between the current device location coordinates and the center coordinates of a preset training area is calculated; it is verified that the Euclidean distance is less than or equal to a preset area radius threshold; in response to the verification being passed, a data acquisition device matching the test type is activated and an acquisition parameter configuration instruction is sent.

[0080] The geolocation identifier in the test start command can be implemented using a geofence code or a site ID, which is used to correlate the physical environment parameters of the pre-set training area. The positioning device can use a GPS module or an indoor UWB positioning system, with a positioning accuracy range of 0.1-1.5 meters to accommodate different site sizes. The center coordinates of the pre-set training area are generated using site mapping data or calculated using standard sports field dimensions, such as the center of a track or field or the center circle of a basketball court. Euclidean distance calculations utilize spatial geometry in a three-dimensional coordinate system, specifically including longitude, latitude, and altitude. The pre-set area radius threshold is dynamically adjusted based on the sport type; for example, it is set to 3 meters for a sprint test and 1 meter for a weightlifting test. The data acquisition device activation logic matches the sensor combination based on the test type; for example, the running test activates the accelerometer and optical marker collector, while the weightlifting test activates the surface electromyography sensor and pressure sensor. Acquisition parameter configuration commands include adjusting the sampling rate, setting the filter cutoff frequency, and switching sensor ranges. For example, the optical capture system sampling rate is set to 120Hz for indoor sites and 60Hz for outdoor sites to reduce ambient light interference.

[0081] When a test start command is received, the embedded geolocation identifier is first parsed to determine the target training area. A positioning device acquires the device's three-dimensional coordinate data in real time and calculates the spatial distance from the preset center coordinates. This distance value is used to determine whether the device is within the standard training environment. For example, if the calculated distance value is 2.5 meters and the preset threshold is 3 meters, the device's position is considered to meet the requirements. After verification, the system selects the appropriate data acquisition device combination based on the test type, such as a lower limb joint landmark capture device for a squat test or an upper limb electromyography sensor for a throwing test. Simultaneously, acquisition parameter configuration instructions are sent to each activated device. For example, setting the accelerometer range to ±16g on a plastic track and ±8g on a grass track to accommodate different ground reaction forces. This spatial position constraint mechanism ensures that the data acquisition environment is consistent with the physical conditions predefined by the biomechanical model, eliminating deviations in parameters such as ground friction and optical reflectivity caused by site variations, thereby improving the accuracy of subsequent joint kinematic parameter calculations.

[0082] In a preferred embodiment, before acquiring motion capture data, the system first receives a test initiation command containing a geographic location identifier. This command can be sent by an operator via a mobile terminal or automatically triggered by a preset training plan. The command includes predefined training area information, such as "indoor track and field" or "strength training area."

[0083] After receiving the start command, the system obtains the device's current location coordinates using a built-in positioning device. This positioning device can be a GPS module, a Bluetooth beacon, or an indoor positioning system. The acquired coordinate information typically includes longitude, latitude, and altitude.

[0084] The system then calculates the Euclidean distance between the current device's location and the center of the pre-set training area. The center of the pre-set training area is a fixed parameter pre-programmed into the system. The Euclidean distance is calculated using a three-dimensional distance formula.

[0085] The calculated Euclidean distance is compared with a preset area radius threshold for verification. The area radius threshold can be set according to the actual size of different training areas. For example, an indoor training area may be set to 10 meters, while an outdoor training area may be set to 50 meters.

[0086] If the verification passes (i.e., the Euclidean distance is less than or equal to the preset area radius threshold), the system activates the data acquisition device that matches the current test type. For example, for a running test, an inertial measurement unit and an optical motion capture system might be activated, while for a weightlifting test, a surface electromyography device might also be activated.

[0087] The system further sends acquisition parameter configuration instructions to the activated data acquisition device. The configuration instructions may include setting the sampling rate, adjusting the filter parameters, calibrating the dynamic range, etc. to adapt to the characteristics of the current training environment.

[0088] After completing the above steps, the system begins to execute the actual motion capture data acquisition process.

[0089] Through the above technical solution, the present application realizes precise control of the data acquisition environment. This ensures that the acquisition equipment is within the preset standard training area, effectively avoiding the problem of data inconsistency caused by environmental differences. The system eliminates the data distortion that may be caused by non-designated sites or equipment displacement deviations through the geographic location verification mechanism. This method improves the adaptability of the collected data to the preset biomechanical model, and lays a solid foundation for the subsequent joint kinematic parameter mapping and evaluation parameter calculation. Furthermore, by dynamically adjusting the data acquisition parameters according to the position information, the system can adapt to the characteristics of different training environments and improve the quality and consistency of the original data. This environmental adaptation mechanism enhances the applicability and data reliability of the system in a variety of training scenarios.

[0090] This application further proposes that obtaining motion capture data includes: collecting three-dimensional acceleration and angular velocity data through an inertial measurement unit; collecting three-dimensional coordinate data of infrared reflective marker points through an optical motion capture system; and performing continuous inter-frame motion vector analysis based on a video acquisition device to generate a two-dimensional motion trajectory.

[0091] Among them, the inertial measurement unit can be integrated into the wearable device, and its sampling frequency can be set to a range of 50Hz to 200Hz, for example, 100Hz, to capture the instantaneous dynamic characteristics of limb movement. The infrared camera array of the optical motion capture system can be arranged around the test area, collecting the coordinates of the marker points at intervals of 20ms to 50ms, for example, 30ms, to ensure that the spatial positioning error is less than 1mm. The frame rate of the video acquisition device can be configured to be 30fps to 120fps, for example 60fps. The motion vector field is calculated by a pixel block matching algorithm between two adjacent frames to generate a two-dimensional planar motion trajectory.

[0092] Specifically, during data acquisition, the inertial measurement unit's three-axis accelerometer and gyroscope record the subject's linear acceleration and angular velocity in real time, generating time-series data of kinematic parameters. Simultaneously, infrared reflective markers placed at the subject's key joints are captured by an optical camera. A triangulation algorithm generates the three-dimensional coordinates of each marker, which serve as a reference for kinematic parameters. If a marker is obscured or the inertial sensor signal is lost, the video capture device generates a two-dimensional planar motion trajectory by analyzing the pixel displacement of the body's outline in consecutive video frames as a supplementary data source. These three data acquisition methods are synchronized in the temporal dimension using a unified clock source and aligned in the spatial dimension using a pre-set coordinate system transformation matrix. For example, when a subject performs a jump, inertial data captures the peak acceleration at takeoff, optical data records changes in body posture during the flight phase, and video data provides a continuous record of the foot's landing trajectory during landing, even if sensor vibration fails. By integrating data from multiple sources in time and space, a multi-dimensional motion capture system covering dynamics, kinematics, and planar trajectories is formed.

[0093] During the jump test, inertial measurement units (IMUs) were attached to the subject's waist and proximal limbs, collecting 3D acceleration and angular velocity data at a sampling rate of 200 Hz. The optical motion capture system's six infrared cameras, arranged in a circular array around the test area, tracked infrared reflective markers attached to the subject's major joints at a frequency of 120 Hz, generating a 3D coordinate dataset encompassing the hip and knee joints. A video capture device captured the subject's lateral motion at a frame rate of 60 frames per second. An optical flow algorithm was used to calculate motion vectors between five consecutive frames, generating a 2D planar motion trajectory during the foot contact phase. The 3D acceleration data was converted into limb rotation angles using quaternion decoding and spatially aligned with the 3D knee joint coordinates acquired by the optical system. The 2D trajectory generated by the video was then mapped to a 3D coordinate system, forming a multimodal motion capture data stream with synchronized timestamps.

[0094] Through the above technical solutions, this application achieves full-dimensional coverage of motion capture data, the inertial measurement unit supplements the high-speed dynamic parameter acquisition capability, the optical system ensures the accuracy of joint space positioning, and the video analysis maintains the continuity of the motion trajectory when the marker point is blocked.

[0095] The present application further proposes establishing a historical evaluation database that stores athlete identification, test time, and physical movement evaluation parameters; executing the currently generated physical movement evaluation parameters with the historical data set with the same athlete identification: calculating the time series change rate of the joint angle deviation value, and generating a movement deformation warning when the change rate exceeds a preset change threshold; performing linear regression modeling on the explosive power index, and outputting the regression coefficient sign change detection result; calculating the recommended recovery period based on the exponential attenuation curve of the muscle fatigue index; and generating a biomechanical report containing the joint angle deviation value change rate, the regression coefficient sign change detection result, and the recommended recovery period.

[0096] The historical evaluation database uses a storage mechanism that associates athlete IDs with test times, enabling data traceability across time. For example, athlete IDs can be combined with a training number and identity code. Test time accuracy can be set to milliseconds to support high-density data analysis. The time series rate of change of joint angle deviation values ​​is calculated using a sliding window algorithm. The window length can be set to 5-10 test cycles. The preset change threshold is dynamically adjusted based on the sport type. For example, the threshold for weightlifting is set to 0.8 degrees / week, and for sprinting, it is set to 1.2 degrees / week. Linear regression modeling uses the least squares method to fit a trend line of explosive power indicators over time. Sign changes in the regression coefficient are detected by comparing the slope direction of adjacent time periods. For example, an abnormal fluctuation flag is triggered when the regression coefficient changes from positive to negative for three consecutive cycles. The exponential decay curve of the muscle fatigue indicator is generated using a nonlinear fitting algorithm. The decay time constant is calculated based on the integrated value of the electromyographic signal. The recommended recovery period is set to 2-3 times the time constant. For example, for a 48-hour time constant, the recommended recovery period is 96-144 hours.

[0097] During the data storage phase, athlete identification is used to link current test data with historical datasets, for example, using a hash table for fast retrieval. During the trend analysis phase, the time series rate of change of joint angle deviation values ​​is calculated by calculating the ratio of the deviation difference between adjacent test cycles to their time interval. When this ratio exceeds a preset threshold, it indicates an accelerating trend in movement deformation, triggering a warning signal. During the explosive power analysis phase, a linear regression model is fitted to historical explosive power indicators. If the sign of the regression coefficient changes from positive to negative, it indicates a sustained decline in explosive power, and a sign change detection result is output. During the recovery cycle calculation phase, based on the decay characteristics of muscle fatigue indicators, the decay rate is determined through curve fitting, and personalized recovery recommendations are generated based on the exercise load intensity. Finally, by integrating the joint angle change rate, regression coefficient sign change, and recovery cycle data, a biomechanical report is generated containing quantitative indicators and decision-making recommendations. For example, the report can include trend charts, anomaly markers, and a recovery plan. This technical solution achieves a closed-loop process from data storage, longitudinal comparison, and trend warning, addressing the problems of traditional methods in identifying chronic movement deformation and lacking data-driven recovery strategies.

[0098] Furthermore, a database server was deployed at the training base to create a data table structure containing athlete ID, test date, and physical fitness assessment parameters. The data table was indexed with a timestamp and the athlete ID as the primary key. When a new physical fitness assessment parameter was added, historical data was retrieved through the database connection interface, extracting the historical joint angle deviation values ​​for the same ID to form a time series array. A moving average algorithm was used to calculate the slope change rate of the three most recent test data. A warning signal was triggered when the absolute value of the slope exceeded 0.15 radians / cycle. A least squares fit was performed on the explosive power index dataset. If the regression coefficient changed from positive to negative for two consecutive test cycles, it was flagged as an abnormal fluctuation event. The half-life of the muscle fatigue index was calculated, and an exponential function was used to fit the decay curve. When the goodness of fit exceeded 0.85, the half-life was multiplied by a safety factor of 1.2 to determine the recommended recovery period. Finally, the warning signal, abnormal fluctuation event, and recovery period values ​​were compiled into a PDF report and automatically sent to the coach's terminal via an email server.

[0099] The present application further proposes a physical fitness training test system, comprising a data acquisition module, a data verification module and a core processing module. The data acquisition module includes an inertial measurement unit interface, an optical capture parser and a video analysis engine.

[0100] The inertial measurement unit interface is configured to receive three-dimensional motion data. This interface can integrate SPI or I2C communication protocol converters to convert raw byte streams output by devices from different manufacturers into a standardized quaternion format. The optical capture parser has a built-in homogeneous coordinate transformation algorithm that maps the coordinates of infrared markers from the optical system coordinate system to the biomechanical model coordinate system by calculating rotation matrices and translation vectors. The video analysis engine uses optical flow to calculate inter-frame motion vectors and eliminates camera shake interference through feature point matching and motion compensation techniques. The data validation module is configured to perform multi-level validation logic. The primary validation stage checks whether the total number of data frames reaches a preset threshold. The intermediate validation stage verifies timestamp continuity. The advanced validation stage identifies invalid data point distribution patterns. After confirming data validity, the core processing module is configured to input multi-source motion data into the biomechanical model to generate physical movement assessment parameters, including joint angle deviation values ​​and instantaneous velocity vectors.

[0101] When receiving wearable sensor data, the inertial measurement unit interface converts the raw binary stream into a standard data format through a protocol conversion mechanism. For example, when the sensor transmits data at a baud rate of 1200bps, the interface's built-in buffer is configured to reassemble packets into 16-byte packets. When processing infrared marker coordinates, the optical capture parser uses a spatial transformation algorithm to convert the raw 3D coordinates into the biomechanical model coordinate system. This transformation is achieved using a 4×4 homogeneous transformation matrix, whose parameters are dynamically adjusted based on the relative position between the optical system and the biomechanical model. The video analysis engine calculates inter-frame motion vectors using a 30fps video stream as input and employs the Lucas-Kanade algorithm to track feature point displacements and generate 2D motion trajectory data. The data validation module implements progressive validation logic, terminating processing if the total number of data frames falls below 100. It also flags data segments as abnormal if the timestamp interval exceeds 33ms. Data re-collection is triggered if the percentage of invalid data points exceeds 5%. The core processing module calculates joint kinematic parameters in a unified space-time coordinate system by fusing inertial data, optical coordinates, and video trajectories, eliminating the timing misalignment problem caused by sampling rate differences in multi-source data, and ultimately generating evaluation indicators including three-dimensional joint angles and linear velocities.

[0102] Preferably, the data acquisition module receives three-dimensional motion data transmitted by wearable sensors through an inertial measurement unit interface. The data is standardized and encapsulated using the JSON format protocol, converting the raw signals from the accelerometer and gyroscope into physical quantity data in the International System of Units. The optical capture parser performs a homogeneous coordinate transformation on the coordinates of the marker points captured by the infrared camera, mapping the original coordinate system to the pelvic center coordinate system of the biomechanical model using a rotation and translation matrix, eliminating joint positioning deviations caused by differences in camera perspective. The video analysis engine calculates the motion vectors of pixels in consecutive video frames using the Horn-Schunck optical flow algorithm to generate two-dimensional planar motion trajectory data. The data verification module performs timestamp synchronization processing on the three modal data received, using linear interpolation to adjust the sampling frequency differences to a unified frequency of 200Hz. The core processing module fuses multi-source data using an inverse kinematic solution algorithm to generate physical movement assessment parameters such as knee flexion angle and hip rotation speed.

[0103] Through the above technical solution, the present application effectively eliminates the data format differences and coordinate system mismatch problems between multi-source heterogeneous devices, and realizes the precise alignment of the inertial measurement unit, optical marking system and video stream data in the time and space dimensions.

[0104] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A physical training test method, characterized in that: include: Synchronously acquiring motion capture data through a multimodal data acquisition device, wherein the motion capture data includes optical motion data or surface electromyography signals; Performing data preprocessing on the raw motion capture data, wherein the data preprocessing includes coordinate system normalization processing and time stamp calibration processing; Performing multi-level data validity verification processing on the pre-processed motion capture data to obtain a verification result; In response to the verification result indicating that the data is valid, when the motion capture data is optical motion data, mapping the motion capture data into joint kinematic parameters based on a biomechanical model, and generating physical motion evaluation parameters based on the joint kinematic parameters; wherein the multi-level data validity verification process includes: Primary verification: Verify that the total number of frames of motion capture data reaches the preset frame number threshold; Intermediate verification: Verify that the time interval between adjacent data frames is less than or equal to the preset time threshold; Advanced verification: Confirm that the proportion of invalid data points in the motion capture data is lower than the preset proportion threshold, and the length of the continuous valid data segment is greater than the preset continuous frame threshold; The multi-level data validity verification process adopts a progressive verification logic, and terminates subsequent verifications when the current level verification fails.

2. The method according to claim 1, characterized in that Generating physical action evaluation parameters includes: Performing frame processing on the motion capture data to obtain an action frame sequence; For each action frame in the action frame sequence, execute: Verify that the coverage of the bounding box of the human motion area is greater than the preset coverage threshold; Perform bone joint positioning processing to obtain a bone joint position coordinate set; generating motion trajectory data based on the bone joint position coordinate set; Performing velocity analysis on the motion trajectory data to obtain an instantaneous velocity vector; Performing angle deviation calculation processing on the motion trajectory data to obtain a joint angle deviation value; generating a dynamics index according to the instantaneous velocity vector and the joint angle deviation value; Filtering action frames whose joint angle deviation values ​​are less than a preset angle threshold as key action frames; Performing feature extraction processing on the key action frame to obtain motion feature data; The physical action evaluation parameter is generated according to the motion feature data.

3. The method according to claim 2, characterized in that The speed analysis process includes: Performing time differentiation processing on the motion trajectory data to generate an original velocity sequence; Performing filtering on the original speed sequence to generate a smoothed speed sequence; performing acceleration calculation processing on the smoothed velocity sequence to generate an acceleration sequence; extracting the maximum acceleration value of the acceleration sequence as an explosive force index; The smoothed velocity sequence is determined as the instantaneous velocity vector.

4. The method according to claim 2, characterized in that The feature extraction process includes: Performing skeleton point detection processing on the key action frame to obtain skeleton point coordinates; Performing a temporal alignment process on the skeleton point coordinates to generate an aligned skeleton sequence; Performing electromyographic signal integration processing on the aligned skeleton sequence to generate muscle activation timing data; The muscle activation timing data is determined as the motion feature data.

5. The method according to claim 1, wherein When the motion capture data is a surface electromyography signal: The data validity verification process includes calculating the ratio of signal power to noise power; Generating physical action evaluation parameters according to motion capture data includes: Performing baseline correction and power frequency filtering on the surface electromyographic signal to obtain a clean electromyographic signal; Performing feature extraction processing on the clean electromyographic signal to generate an electromyographic feature sequence; Performing fatigue calculation processing on the electromyographic characteristic sequence to obtain a muscle fatigue index; The muscle fatigue index is determined as the physical action evaluation parameter.

6. The method according to claim 2, characterized in that The feature extraction process further includes: Performing lower limb joint detection processing on the key action frame to obtain a lower limb bone joint position coordinate set, wherein the lower limb bone joint position coordinate set includes hip joint coordinates, knee joint coordinates, and ankle joint coordinates; Performing cross-frame trajectory alignment processing on the lower limb bone joint position coordinate set to generate a lower limb motion trajectory; Perform velocity analysis on the lower limb motion trajectory: Calculate the displacement change per unit time to generate the lower limb instantaneous velocity sequence; performing smoothing processing on the lower limb instantaneous velocity sequence; Extract the ratio of the duration of the acceleration segment to the duration of the deceleration segment; The explosive power index and the ratio of the duration of the acceleration segment to the duration of the deceleration segment are integrated into motion characteristic data.

7. The method according to claim 1, characterized in that Before executing the acquisition of motion capture data, the method further includes: receiving a test start instruction including a geographic location identifier; Obtaining the current device location coordinates based on the positioning device; Calculating the Euclidean distance between the current device location coordinates and the center coordinates of the preset training area; Verify that the Euclidean distance is less than or equal to a preset area radius threshold; In response to the verification being passed, a data acquisition device matching the test type is activated and an acquisition parameter configuration instruction is sent.

8. The method according to claim 7, characterized in that The acquiring of motion capture data comprises: Collect three-dimensional acceleration and angular velocity data through the inertial measurement unit; The acquiring of motion capture data comprises: The three-dimensional coordinate data of infrared reflective markers is collected through an optical motion capture system; The acquiring of motion capture data comprises: The video capture device performs continuous inter-frame motion vector analysis to generate a two-dimensional motion trajectory.

9. The method according to claim 8, characterized in that Also includes: Establish a historical evaluation database to store athlete identification, test time, and physical performance evaluation parameters; Compare the currently generated physical performance evaluation parameters with the historical dataset of the same athlete: Calculating the time series change rate of the joint angle deviation value, and generating a motion deformation warning when the change rate exceeds a preset change threshold; Conduct linear regression modeling on explosive power indicators and output regression coefficient sign change detection results; Calculate the recommended recovery period based on the exponential decay curve of muscle fatigue indicators; A biomechanical report including the rate of change of the joint angle deviation value, the regression coefficient sign change detection result and the recommended recovery period is generated.

10. A physical training test system, characterized in that: include: A data acquisition module, used for acquiring motion capture data; A data verification module is used to perform data validity verification processing on the motion capture data to obtain a verification result; A core processing module, configured to generate a physical action evaluation parameter based on the motion capture data in response to the verification result indicating that the data is valid; Wherein, the data acquisition module includes: an inertial measurement unit interface configured to receive three-dimensional motion data; an optical capture parser configured to process infrared marker coordinates; A video analytics engine is configured to calculate inter-frame motion vectors.

Images