A running intelligent evaluation and diagnosis system

Abstract

The application relates to a running intelligent evaluation and diagnosis system, belonging to the field of human motion analysis, which comprises a hardware device and system software, wherein the hardware device comprises a plantar force measurement unit and a non-marker point motion capture unit; the system software comprises a plantar force measurement unit control module, an acquisition module, a data processing module, an intelligent evaluation and diagnosis module, a data display module and a historical comparison module. The system takes key feature parameters of running as evaluation indexes, establishes four-level evaluation standards and training suggestions corresponding to each index, clearly and intuitively presents each index and evaluation results to runners in various modes such as video, graphics and tables, realizes quantitative evaluation, risk diagnosis and personalized improvement guidance of running technology, effectively reduces the blindness of running, and improves the scientific nature of training. The application is convenient to use, does not need to paste point markers, is simple and efficient to operate, and can complete the whole process of running physical examination from testing, analysis to output of professional suggestions within several minutes.

Description

Technical Field

[0001] This invention relates to a running intelligent assessment and diagnostic system, belonging to the field of human motion analysis. Background Technology

[0002] Running, as one of the most convenient daily forms of exercise to improve physical fitness, offers numerous benefits when maintained long-term: it effectively enhances cardiorespiratory endurance, improves lung function, and strengthens the cardiovascular system; it effectively lowers blood lipid levels, cholesterol, and LDL levels, reducing or slowing the onset of fatty liver; it can regulate mood, reduce the incidence of depression, and improve sleep; regular running causes rhythmic contraction and relaxation of muscles throughout the body, increasing muscle fiber and protein content, resulting in a more toned and robust physique and improved immunity. Due to the various benefits of running, the number of people participating in running is increasing as people's health awareness grows, with the rapidly developing marathon being a prime example. However, although everyone runs, not everyone runs correctly. For both beginners and experienced runners, proper running form is crucial. Proper running form allows the body to develop and strengthen along the correct path, rather than becoming increasingly distorted and deformed. It reduces the risk of injury during running, enabling the body to form a reasonable protective mechanism. While improving physical fitness, it also prevents sports injuries and maximizes the overall coordination and output distribution of the body, thus scientifically and economically helping runners gradually improve their performance. In contrast, poor running form will only lead to increasingly distorted forms over long-distance running, causing injuries and various other problems.

[0003] For the average runner, many technical and physical issues cannot be effectively identified and corrected solely through subjective ability. A more professional comprehensive running ability assessment system is needed to provide objective and detailed running advice. Currently, there are four main methods of running assessment on the market: First, runners can assess each other or have a running coach evaluate them visually. This is a qualitative evaluation method, but it requires an experienced running coach and is highly subjective. Second, most mobile phones or cameras have slow-motion video recording capabilities to capture running posture. A simpler approach is to directly slow down the video for evaluation, while a more complex approach involves downloading a motion analysis app or software and importing the video for analysis. The first method of running posture assessment requires a higher level of expertise and often requires professionals to complete. In addition, factors such as the clarity of the video, the shooting angle, and the frame rate have a significant impact on the analysis results, making it difficult to obtain accurate results. The third method uses wearable devices such as sports watches, wristbands, or sensors for running posture assessment. The advantage of this method is its simplicity and ease of use, but the disadvantage is that the available data is limited, making comprehensive analysis difficult. The fourth method involves conducting running posture assessments in a professional sports biomechanics laboratory. This method often requires cumbersome on-site testing and professional post-processing of data, which is time-consuming and labor-intensive.

[0004] For runners, there is a need for a comprehensive, accurate, user-friendly, and professional running assessment system to answer the three core questions about running: "How well am I running?", "What problems are there in my running?", and "How can I improve?". Summary of the Invention

[0005] This invention aims to solve the problems of inaccurate data, cumbersome operation, low efficiency, and inability to accurately quantify running posture and force exertion and provide scientific improvement solutions in existing running assessment systems. It proposes an intelligent running assessment and diagnosis system.

[0006] The technical solution of the present invention:

[0007] A running intelligent assessment and diagnostic system, including hardware devices and system software;

[0008] The hardware device includes a plantar force measurement unit and a markerless motion capture unit;

[0009] The foot force measurement unit uses a force measurement treadmill or a force measurement track made up of multiple force measurement plates to collect foot force data during the runner's running process;

[0010] The markerless motion capture unit includes four or more industrial cameras deployed around the foot force measurement unit to synchronously collect kinematic image data of the runner during the running process.

[0011] The system software includes a plantar force measurement unit control module, an acquisition module, a data processing module, an intelligent assessment and diagnosis module, a data display module, and a historical comparison module;

[0012] The foot plantar force measurement unit control module is connected to the acquisition module and is used to realize the power-on, power-off, start, stop and speed control of the foot plantar force measurement unit;

[0013] The acquisition module is connected to the plantar force measurement unit and the markerless motion capture unit respectively, and is used to realize personnel information management, hardware device communication interaction, acquisition parameter configuration and acquisition start and stop control.

[0014] The data processing module is signal-connected to the acquisition module and includes a force processing module, an attitude processing module, and a comprehensive processing module. The force processing module is used to perform periodic division, effective data filtering, parameter calculation, and standardization processing on the plantar force data. The attitude processing module is used to perform joint point recognition, three-dimensional reconstruction, and kinematic parameter calculation on the kinematic image data. The comprehensive processing module is used for attitude periodic division, calculation of key running feature parameters, and output of key running feature parameters, feature curves, and feature values ​​for the entire run.

[0015] The intelligent assessment and diagnosis module is connected to the data processing module and is used to compare and evaluate key feature parameters with preset evaluation standards, and output rating results, problem risks, causes and corresponding training suggestions.

[0016] Both the data display module and the historical comparison module are connected to the intelligent evaluation and diagnosis module. The data display module is used for the visualization of test data and evaluation results, and the historical comparison module is used for multi-dimensional comparative analysis of multiple sets of test data.

[0017] Specifically, the foot force measurement unit adopts a force-measuring treadmill or a force-measuring track spliced ​​together from multiple force-measuring plates. The collected foot force data includes force data in the XYZ three directions and pressure center data, with a sampling rate of no less than 1000 times / second. The markerless motion capture unit includes four or more industrial cameras deployed around the foot force measurement unit. The camera shooting frame rate is no less than 100 frames / second, the image resolution is no less than 1920*1080, and it is equipped with an AI recognition model that can automatically identify 25 joint points of the human body in the image.

[0018] Specifically, the acquisition module supports three working modes: manual acquisition, fixed-duration acquisition, and automatic trigger acquisition. The manual acquisition is controlled by manual triggering to start and stop the acquisition. The fixed-duration acquisition is completed automatically according to the preset acquisition delay and acquisition duration. The automatic trigger acquisition is started and stopped automatically by photoelectric switch signals at both ends of the force measurement track.

[0019] Specifically, the processing flow of the force measurement module is as follows:

[0020] The plantar force data is converted to the support reaction force world coordinate system to generate curves for normal force Fz-t, horizontal force Fy-t, and lateral force Fx-t, and invalid pressure center data with normal force below the zero threshold are removed.

[0021] The timing of single-foot contact and push-off is determined by using the zero-point threshold of normal force, thus completing the division of the running cycle; the timing of the end of the buffer is determined by the moment Fy=0, which is closest to the midpoint of the contact period in the horizontal force Fy-t curve, thus dividing the contact period into the braking phase and the propulsion phase.

[0022] Valid running cycles were selected based on the average value of the peak normal force during the ground contact period and the rule of three times the standard deviation, and invalid data with fewer than three ground contact times or abnormal peak values ​​were removed.

[0023] It calculates cadence, ground contact time, track stride, treadmill stride, and track speed, and standardizes the force measurement data in the time dimension. After multi-cycle averaging, it outputs single-cycle average force measurement data.

[0024] Specifically, the processing flow of the attitude processing module is as follows:

[0025] The AI ​​joint point recognition model extracts the two-dimensional coordinates of 25 human joint points from the acquired images, and then obtains the three-dimensional coordinates of the joint points through multi-view three-dimensional reconstruction.

[0026] The three-dimensional coordinates of the joints are translated and transformed to make the attitude coordinate system coincide with the force measurement center coordinate system of the plantar force measurement unit;

[0027] The attitude data is upsampled to the same sampling rate as the force measurement data by spline interpolation, and the motion parameters of the human body's joints, angular velocity, angular acceleration and center of mass are calculated.

[0028] Based on the effective period of the force measurement data, the attitude data is divided into periods and standardized. After multi-period averaging, the single-period average attitude data is output.

[0029] Specifically, the integrated processing module, based on standardized single-cycle average force measurement data and single-cycle average posture data, calculates the joint forces and moments of the hip, knee, and ankle joints of the lower limbs throughout the entire cycle using inverse dynamics. At the same time, it calculates the key characteristic parameters of the entire running process, including vertical amplitude of the center of gravity, lateral displacement of the center of gravity, ground contact mode, landing point, inversion / exversion angle of the foot, inversion / exversion angle of the foot, joint angle characteristics, trunk forward tilt angle, head forward tilt angle, and lower limb stiffness, and extracts the corresponding characteristic curves and feature values ​​of key time nodes.

[0030] Specifically, the integrated processing module has a built-in human body center of mass calculation submodule. Based on the corrected human body inertial parameters, it first calculates the mass and center of mass coordinates of each body segment of the head, torso, and limbs according to the runner's gender and weight, and then calculates the three-dimensional coordinates of the whole body center of mass by weighted summation of body segment mass.

[0031] Specifically, the preset evaluation criteria of the intelligent assessment and diagnosis module are divided into four levels: A, B, C, and D. Each level corresponds to a preset numerical range, with the A-level values ​​taken from elite athlete data. The output training suggestions include movement difficulty, movement type, movement name, movement key points, training volume, training intensity, training frequency, training cycle, and multi-view movement videos demonstrated by professional athletes.

[0032] Specifically, the data display module displays personnel and test information, multi-view videos, a three-dimensional human running model, key parameters and rating results, mechanical and kinematic curves and characteristic values, diagnostic problems and training suggestions; the historical comparison module supports the comparison of multiple sets of data of the runner or data of different runners, and the comparison forms include curve comparison and characteristic value bar chart comparison.

[0033] The beneficial effects of this invention are:

[0034] This invention discloses a smart running assessment and diagnostic system. The system uses key running characteristic parameters as evaluation indicators, is based on extensive test data from marathon runners, and incorporates the experience of top domestic running experts and coaches. It establishes a four-level evaluation standard and training suggestions for each indicator, presenting runners with clear and intuitive results through videos, graphs, and tables. The system data-driven assessment of running posture and power generation provides targeted training suggestions based on different problems identified in runners, helping them reduce aimless running and achieve efficient, injury-free running. The system is user-friendly, requiring only a few minutes to complete the entire "running health check" process, from testing to receiving recommendations.

[0035] Compared to simple posture assessment or wearable sensor assessment, this system has the following advantages: First, it combines high-sampling-rate, high-precision force measurement equipment with a motion capture system. By acquiring runner kinematic and plantar dynamic data, it can accurately calculate the joint forces and torques of the hip, knee, and ankle joints in the lower limbs, which helps to accurately assess the runner's injury risk. This is something that simple posture assessment or wearable sensor assessment cannot do. Second, the high sampling rate of the force measurement equipment allows time parameters such as cycle time and ground contact time calculated based on the force measurement data to be accurate to 1ms. In contrast, pure posture assessment or wearable sensor assessment often only has a sampling rate of 100 / s to 200 / s, and time parameters are only accurate to 10ms to 5ms. Using force measurement equipment not only provides higher accuracy in time parameters, but also in time-related posture angle data and even joint torque data. Compared to professional biomechanical laboratory assessments, the advantage of this system lies in its ease of use. Runners no longer need to perform tedious preparations such as applying reflective dots, which usually takes more than an hour, or wait for experts to conduct step-by-step manual assessments. Through intelligent means, runners only need to run on the treadmill for one minute or spend a few seconds running across the force testing track, and wait two or three minutes to obtain their own running assessment results, understand their own problems and corresponding suggestions, which greatly improves the efficiency of assessment. Attached Figure Description

[0036] Figure 1 This is a flowchart of the system workflow of the present invention;

[0037] Figure 2 This is a flowchart of the force measurement and processing module.

[0038] Figure 3 This is a flowchart of the integrated processing module. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described below with reference to specific embodiments shown in the accompanying drawings.

[0040] like Figure 1 , Figure 2 , Figure 3 As shown, this invention discloses a running intelligent assessment and diagnosis system, which consists of two main parts: hardware and system software. The hardware includes: 1) a plantar force measurement unit, which uses a force-measuring treadmill or a force-measuring track composed of multiple force-measuring plates to acquire data on the runner's foot force exertion during running; and 2) a markerless motion capture unit, which includes four or more industrial cameras deployed around the plantar force measurement unit to simultaneously acquire kinematic image data of the runner during running. The system software includes a plantar force measurement unit control module, a data acquisition module, a data processing module, an intelligent assessment and diagnosis module, a data display module, and a historical comparison module.

[0041] The foot force measurement unit collects runner's foot force data, including XYZ three-directional force data and pressure center data, with a sampling rate of no less than 1000 times / second.

[0042] The markerless motion capture unit consists of four or more industrial cameras, arranged around the perimeter of the force-measuring treadmill or force-measuring track. The camera capture frame rate is no less than 100 frames per second, and the image resolution is no less than [missing information]. The markerless technology used can automatically identify 25 joints of the human body in the image by AI.

[0043] The foot force measuring unit control module is connected to the acquisition module and is used to realize the power-on, power-off, start, stop and speed control functions of the foot force measuring unit.

[0044] The data acquisition module is connected to the plantar force measurement unit and the markerless motion capture unit, respectively, for personnel information management, hardware device communication and interaction, and configuration and start / stop control of acquisition parameters. Personnel information input and retrieval covers information such as name, contact information, gender, age, height, and weight. Acquisition parameter configuration includes start / stop acquisition control, acquisition duration setting, acquisition delay setting, automatic trigger setting, and test number setting. There are three acquisition modes: manual acquisition, fixed-duration acquisition, and automatic trigger acquisition. Manual data acquisition mode refers to the operator manually clicking the "Start Data Acquisition" and "Stop Data Acquisition" buttons to control the start and end of data acquisition. Fixed-duration data acquisition mode means the operator only controls the start of data acquisition; the end is determined by the system based on pre-set "data acquisition delay" and "data acquisition duration." For example, a "data acquisition delay" set to 5 seconds and a "data acquisition duration" set to 4 seconds means that after the operator clicks the "Start Data Acquisition" button, the system starts data acquisition after 5 seconds and stops after 4 seconds. Automatic trigger data acquisition mode is mainly used for force measurement tracks. Photoelectric switches are installed at both ends of the track and connected to the system. When a runner passes the photoelectric switch at the starting line, the switch sends a start signal, and the system begins data acquisition upon receiving the signal. Similarly, when the runner passes the photoelectric switch at the finish line, the system stops data acquisition. After data acquisition, it is automatically sent to the data processing module for processing.

[0045] The data processing module is signal-connected to the acquisition module and includes a force processing module, an attitude processing module, and a comprehensive processing module. The force processing module is used to perform periodic division, effective data filtering, parameter calculation, and standardization processing on the plantar force data. The attitude processing module is used to perform joint point recognition, three-dimensional reconstruction, and kinematic parameter calculation on the kinematic image data. The comprehensive processing module is used for attitude periodic division, calculation of key running feature parameters, and output of key running feature parameters, feature curves, and feature values ​​for the entire run.

[0046] The main processes of the force measurement processing module include running cycle division, effective cycle selection, calculation of key running characteristic parameters, and force measurement data standardization and averaging. The specific execution steps are as follows:

[0047] S1. Generate resultant force data from the raw force measurement data:

[0048] The force measurement data is transformed from the balance force coordinate system to the support reaction world coordinate system. Data points are generated at 1 ms intervals, starting at time 0, to produce the normal force Fz-t curve, horizontal force Fy-t curve, and lateral force Fx-t curve. Note that invalid values ​​in the pressure core data are removed when the normal force is below the zero-point threshold.

[0049] S2, Running Cycle Division:

[0050] S2.1 Determine the contact time and push-off time from the Fz-t curve:

[0051] The zero-point threshold of the normal force Fz determines the single-foot contact time (the first point greater than the zero-point threshold) and the push-off time (the point after the contact time but before the zero-point threshold). Taking the right foot contact as an example, from the starting point of the force measurement data, the right foot contact time, right foot push-off time, left foot contact time, left foot push-off time, right foot re-contact time, and so on, until the end of the data. The period from right foot contact to right foot re-contact is divided into a complete running cycle. The period from right foot contact to right foot push-off is the right foot contact period; the period from right foot push-off to left foot contact is the two-foot airborne period; the period from left foot contact to left foot push-off is the left foot contact period; and the period from left foot push-off to right foot re-contact is the two-foot airborne period. Thus, the data is divided into N running cycles, and these N running cycles can be further divided into N left foot contact periods and N right foot contact periods.

[0052] S2.2 Determine the end time of buffering using the horizontal force Fy-t curve:

[0053] During each contact period, there are one or more (generally occurring during the braking phase) moments of Fy=0 on the horizontal force Fy-t curve. The Fy=0 moment closest to the middle of the contact period is defined as the end of the buffer. The period from the contact moment to the end of the buffer is the braking phase of the contact period, and the period from the end of the buffer to the push-off moment is the propulsion phase of the contact period.

[0054] S3. Valid Data Filtering:

[0055] 1) Determine if the number of ground touches is greater than or equal to three (whether there are at least three ground touch moments). If the number of ground touches is less than three, stop the calculation and output: The number of ground touches is less than three, and a complete running cycle cannot be obtained.

[0056] 2) Extract Fzmax and its corresponding time t for each contact period. Fzmax The Fzmax of the foot is determined based on the Fz-t curve during each ground contact period. If the heel strikes the ground, there is a double peak. Due to the impact upon landing, the first peak may be greater than the second peak. In this case, the second peak value is taken as Fzmax. Fzmax usually occurs around the middle of the ground contact period.

[0057] 3) Valid Period Screening: Step 1: Calculate the average and standard deviation of Fzmax for all feet. Step 2: Determine if the difference between the first touchdown Fzmax and the average Fzmax is greater than three times the standard deviation. If not, the touchdown data is considered valid, and the valid period is defined from the moment that foot touches the ground. If it is greater than three times the standard deviation, the touchdown data is considered invalid (usually because the foot did not fully land on the first force plate of the track), and the valid period is defined from the moment the next foot touches the ground. Step 3: Determine if the difference between the last touchdown Fzmax and the average Fzmax is greater than three times the standard deviation. If not, the touchdown data is considered valid and can be included in the valid period definition. If it is greater than three times the standard deviation, the touchdown data is considered invalid (usually because the foot did not fully land on the last force plate of the track) and is not included in the valid period definition. However, note that the moment of touchdown is valid.

[0058] 4) After the above steps, valid force measurement period data are obtained. All subsequent calculations are performed within the range of valid force measurement data.

[0059] S4. Extract and calculate the force characteristic value:

[0060] 1) Extract all ground contact time, buffer end time, and push-off time for subsequent calculations and provide video images to identify the ground contact time, buffer end time, and push-off time of all feet;

[0061] 2) Determine whether the right foot or the left foot touches the ground at each contact moment: t = contact moment (given by force measurement). If Z LAnkle (Z14) > Z RAnkle (Z11) indicates that if the left foot is in the air, then the right foot will touch the ground. LAnkle (Z14) < Z RAnkle (Z11) indicates that the right foot is in the air at this time, and the left foot touches the ground. Combining the aforementioned effective data filtering steps, all effective left foot Fzmax and right foot Fzmax are obtained, and the average value of left foot Fzmax and right foot Fzmax is calculated.

[0062] 3) Based on the Fy-t curve and the aforementioned phase division within the ground contact period, extract the maximum braking force for all effective left-foot braking phases and calculate its average value; extract the maximum braking force for all effective right-foot braking phases (positive values ​​in the coordinate system of this embodiment) and calculate its average value; extract the maximum propulsion force for all effective left-foot propulsion phases (negative values ​​in the coordinate system of this embodiment) and calculate its average value; extract the maximum propulsion force for all effective right-foot propulsion phases and calculate its average value.

[0063] 4) Based on the Fx curve and the aforementioned ground contact period division, extract the maximum lateral force of the left foot (the point with the largest absolute value) within all valid ground contact periods, and calculate its average value (with a positive or negative sign when calculating the average value); extract the maximum lateral force of the right foot (the point with the largest absolute value) within all valid ground contact periods, and calculate its average value (with a positive or negative sign when calculating the average value).

[0064] 5) Braking impulse: Calculated separately for the left and right feet. First, calculate all positive Fy values ​​and time t during each contact braking phase, then average them to obtain the braking impulse.

[0065] 6) Thrust impulse: Calculated separately for the left and right feet. First, calculate all negative Fy integrals with time t during each ground contact propulsion phase, and then average them to obtain the thrust impulse.

[0066] S5. Gait parameter calculation:

[0067] 1) Step frequency. First step: Separate the left and right feet, and calculate the cycle time of all left and right feet within the effective range; the first left foot cycle time:

[0068] The first left-foot cycle time = t 左脚第二次触地 -t 左脚第一次触地 ;

[0069] The second left-foot cycle time = t 左脚第三次触地 -t 左脚第二次触地 ;

[0070] Calculate all left and right foot cycle times in the same way. Second step: Calculate the average left and right foot cycle times. Third step: Calculate left and right foot cadence; Left foot cadence = 120 / average left foot cycle time, Right foot cadence = 120 / average right foot cycle time. Fourth step: Calculate average cadence; Average cadence = (left foot cadence + right foot cadence) / 2. Note: If fewer ground contact results in only left or right foot cadence, then average cadence = either left or right foot cadence.

[0071] 2) Stride length (stride length). Stride length is calculated using the center of pressure (Ly) in the forward direction. Step 1: Separate left and right feet, and extract the center of pressure (Ly) coordinates in the forward direction for all left and right feet within the effective range at the ground contact time t=Fzmax; Step 2: Calculate the stride length of the left and right feet:

[0072] If the first foot touches the ground is the left foot, then,

[0073] First left step length = -(Ly 左脚第二个触地期内Fzmax -Ly 右脚第一个触地期内Fzmax );

[0074] Second left step length = -(Ly 左脚第三个触地期内Fzmax -Ly 右脚第二个触地期内Fzmax ); ... ...

[0076] First right foot length = -(Ly 右脚第一个触地期内Fzmax -Ly 左脚第一个触地期内Fzmax );

[0077] Second right foot length = -(Ly 右脚第二个触地期内Fzmax -Ly 左脚第二个触地期内Fzmax );

[0078] Calculate all left and right foot lengths in the same way;

[0079] If the first foot touches the ground is the right foot, then,

[0080] First left step length = -(Ly 左脚第一个触地期内Fzmax -Ly 右脚第一个触地期内Fzmax );

[0081] Second left step length = -(Ly 左脚第二个触地期内Fzmax -Ly 右脚第二个触地期内Fzmax ); ... ...

[0083] First right foot length = -(Ly 右脚第二个触地期内Fzmax -Ly 左脚第一个触地期内Fzmax );

[0084] Second right foot length = -(Ly 右脚第三个触地期内Fzmax -Ly 左脚第二个触地期内Fzmax );

[0085] Calculate all left and right foot lengths in the same way; Step 3: Calculate the average left and right foot lengths; Step 4: Calculate the average stride length; Average stride length = (average left foot length + average right foot length) / 2; Step 5: Calculate the average stride length in special cases. If the number of ground touches is low, resulting in only the average left or right foot length, then the average stride length = the average left or right foot length.

[0086] 3) Treadmill stride: Since the speed of a treadmill is already given, the stride can be calculated from the speed and cadence. Step 1: Obtain the treadmill speed. The unit is km / h; Step 2: Calculate the average left foot length and the average right foot length, the average left foot length ,

[0087] Average right foot length ;

[0088] Step 3: Calculate the average stride length; Average stride length = (average left stride length + average right stride length) / 2.

[0089] 4) Runway speed. (km / h) = Track stride / 1000 / 60 × cadence 3.6, where the runway stride is measured in mm.

[0090] Time to contact with ground = (t) 蹬离时刻 -t 触地时刻 ) / 1000+0.001, where the unit of time t is ms and needs to be converted to s. Step 1: Based on the previously obtained contact time and push-off time, and which foot is on the ground between the two times, all valid contact times of the left foot and right foot can be determined; Step 2: Calculate the average contact time of the left foot and the average contact time of the right foot.

[0091] S6. Force measurement data standardization and force measurement curve averaging:

[0092] To facilitate comparison, the force measurement data are standardized. Support reaction force data needs to be divided by the human body weight BW (mg, g=9.8), thus changing the unit of support reaction force to BW; impulse data needs to be divided by the human body mass m, thus changing the unit of impulse to... For the Ft curve, in addition to standardizing the vertical axis force, the horizontal axis time t also needs to be standardized. Specifically, the horizontal axis is changed from absolute time in seconds to relative time in % of the cycle time, that is, 0% of the cycle time corresponds to the start of the cycle (ground contact), and 100% of the cycle time corresponds to the end of the cycle (ground contact again). The number of standardized points for each force measurement cycle is 101 (through interpolation), that is, each cycle is divided into 100 equal intervals, represented by 0-100. Averaging is to average the N standardized cycle data to obtain the single-cycle average force measurement data. Two points need to be noted here: 1) Before standardization, that is, before interpolation, the forces in all three directions during the take-off period are all taken as zero, in order to avoid introducing errors into possible take-off moment calculations later. 2) To facilitate comparison between different people, the standardized and averaged support reaction force curves are currently uniformly arranged with the left foot forward and the right foot backward. Therefore, if the test involves the right foot in front and the left foot behind, the data for each foot in each cycle needs to be swapped before standardization. Specifically, the left foot grounding phase + left foot takeoff phase in the latter half should be moved to the front, and the right foot grounding phase + right foot takeoff phase in the first half should be moved to the back. Then, standardization (interpolation) and averaging should be performed.

[0093] S7. Force Measurement Output Content:

[0094] The output content refers to the post-processed data that needs to be presented on the interface or kept for later use and stored in a file, excluding the original data. Specifically, it includes: 1) all valid ground contact times, buffer end times, and push-off times; 2) left foot cadence, right foot cadence, and average cadence; 3) average left foot length, average right foot length, and average stride length; 4) speed; 5) average ground contact time of the left foot and average ground contact time of the right foot; 6) average Fzmax of the left foot and average Fzmax of the right foot; 7) average maximum braking force, average maximum propulsion force, and average maximum lateral force of the left and right feet; 8) average braking impulse of the left and right feet and average propulsion impulse; 9) averaged and standardized Fx-% cycle, Fy-% cycle, and Fz-% cycle curves.

[0095] The attitude processing module mainly includes two processes: 2D joint recognition and 3D reconstruction, as well as subsequent coordinate transformation, upsampling, initial kinematic data calculation, attitude data standardization and averaging. The specific execution steps are as follows:

[0096] S1, 2D Joint Recognition and 3D Reconstruction:

[0097] 2D joint recognition uses the acquired kinematic image data as input, and outputs the two-dimensional coordinates of 25 joints of the human body after being identified by the AI ​​joint recognition model; 3D reconstruction optimizes and solves the two-dimensional coordinates of human joints from multiple angles to obtain the 3D coordinates of human joints.

[0098] S2. Processing of raw attitude coordinate data:

[0099] S2.1 Coordinate Transformation: To facilitate subsequent calculation of joint moments, the origin of the attitude coordinate system needs to coincide with the force measurement center of the plantar force measurement unit. Since the coordinate systems are aligned, it is only necessary to translate the three-dimensional coordinates of the joint points in the attitude coordinate system: X = X 原始 -X 偏移

[0100] Y=Y 原始 -Y 偏移

[0101] Y=Z 原始 -Z 偏移 ;

[0102] Among them, X 偏移 Y 偏移 Z 偏移 The coordinates are determined during calibration and can be changed in the configuration file. Subsequent calculations are all performed using the transformed coordinates.

[0103] S2.2. The upsampling frame rate of attitude data (usually 100 frames / s) is lower than the sampling rate of force measurement data (1000 times / second). At higher speeds, the difference in attitude data between two frames is significant. When the frame rate cannot be increased, interpolation (spline interpolation) is used to increase the number of attitude data points to 1000 times / second, perfectly matching the sampling rate of the force measurement data. For example, with 100 frames / second of attitude data, the original time series is 0, 10ms, 20ms, 30ms…, which needs to be interpolated to time series 0, 1, 2, 3…10, 11, 12, 13…20, 21, 22…30.

[0104] S3, Kinematic parameter calculation:

[0105] The entire kinematic data is calculated, including: posture angles; three-dimensional coordinates of the body's center of mass; angular velocity and angular acceleration calculated from the posture angles using the finite difference method; and XYZ velocities and accelerations calculated from the three-dimensional coordinates of the body's center of mass.

[0106] S4. Attitude Period Division:

[0107] The kinematic data for the entire process is divided into N effective periods using the effective ground contact times of force measurements. This facilitates the subsequent extraction of feature time values ​​and the standardization and averaging of the data. It's important to note that, except for the head and trunk, both the upper and lower limbs are divided into left and right sides. The period data for the left limb angles (including angular velocity and angular acceleration) is from the left foot's ground contact to its re-contact; the period data for the right limb angles (including angular velocity and angular acceleration) is from the right foot's ground contact to its re-contact. Therefore, if the left foot is in front and the right foot is behind within this period, the calculation of the right limb angles (including angular velocity and angular acceleration) requires a left-right reversal, similar to the force measurement method. Conversely, if the right foot is in front and the left foot is behind within this period, the calculation of the left limb angles (including angular velocity and angular acceleration) requires a left-right reversal, similar to the force measurement method.

[0108] S5. Extract and calculate kinematic parameters:

[0109] S5.1, Attitude Feature Values:

[0110] 1) Extract the values ​​of the head, torso, and left and right limb angles (including angular velocity and angular acceleration) at the moment of ground contact, the end of the buffer, and the moment of push-off for each effective cycle, as well as the maximum value, minimum value, and range of motion (maximum value - minimum value). 2) Average the N cycle values ​​to calculate the average value of the posture characteristic value.

[0111] S5.2, Vertical amplitude of the center of gravity:

[0112] 1) Extract the maximum and minimum Z-coordinates of the body's center of mass within each cycle; 2) Calculate N... Zmax-Zmin; 3) Calculate the average .

[0113] S5.3 Lateral displacement of the center of gravity:

[0114] 1) Extract the maximum and minimum X-coordinates of the body's center of mass within each cycle; 2) Calculate N... Xmax-Xmin; 3) Calculate the average .

[0115] S5.4, Ground contact method:

[0116] 1) Based on the valid left and right foot contact times, calculate each left and right foot contact pattern.

[0117] Left foot: (Z) 21 -Z 19 -10mm < (Z) indicates heel touches the ground; -10mm < (Z) 21 -Z 19 <10mm, full palm contact; (Z)21 -Z 19 >10mm, forefoot strikes the ground;

[0118] Right foot: (Z) 24 -Z 22 -10mm < (Z) indicates heel touches the ground; -10mm < (Z) 24 -Z 22 <10mm, full palm contact; (Z) 24 -Z 22 >10mm, forefoot strikes the ground;

[0119] 2) The left foot takes the most frequent contact with the ground as the left foot's contact method, and the right foot takes the most frequent contact with the ground as the right foot's contact method.

[0120] S5.5, Landing point location:

[0121] 1) Based on the valid left and right foot contact times, calculate the landing point for each foot contact:

[0122] Left: The moment the left foot touches the ground (Y) BC -Y 14

[0123] Right: Right foot touches the ground at moment Y BC -Y 11 .

[0124] 2) Calculate the average value of the landing positions of the left and right feet.

[0125] S5.6, Foot inversion / pronation, i.e. the average value of the inversion / pronation angle of the foot during the ground contact period. Therefore: 1) Extract the inversion / pronation angle of the left foot during the ground contact period of each left foot and the inversion / pronation angle of the right foot during the ground contact period of each right foot within the effective range; 2) Take the average of the inversion / pronation angles of the left foot during the ground contact period of all left feet and take the average of the inversion / pronation angles of the right foot during the ground contact period of all right feet.

[0126] S5.7, Foot inward and outward toeing, i.e. the average value of the inward and outward toeing angles of the foot during the ground contact period. Therefore: 1) Extract the inward and outward toeing angles of the left foot during the ground contact period of each left foot and the inward and outward toeing angles of the right foot during the ground contact period of each right foot within the effective range; 2) Take the average of the inward and outward toeing angles of the left foot during all the ground contact periods of the left foot and take the average of the inward and outward toeing angles of the right foot during all the ground contact periods of the right foot.

[0127] S5.8 Head forward tilt, which is the average value of the head forward tilt angle throughout the entire cycle. Therefore, the average value of all head forward tilt angles within the effective range can be calculated directly.

[0128] S5.9, Forward tilt of the trunk, which is the average value of the forward tilt angle of the trunk throughout the entire cycle. Therefore, the average value of all forward tilt angles of the trunk within the effective range can be calculated directly.

[0129] S5.10 Lower limb stiffness: According to the definition of lower limb stiffness, the hip-to-ankle distance is considered as the lower limb length, and the length changes as follows:

[0130] ;

[0131] Lower limb stiffness = Fzmax / length change.

[0132] 1) Based on the effective contact time of the left and right feet, Fzmax and t obtained from the force measurement Fzmax Calculate the stiffness of each left lower limb and the stiffness of each right lower limb separately;

[0133] 2) The stiffness of all left lower limbs is averaged, and the stiffness of all right lower limbs is averaged.

[0134] S6. Attitude data curve standardization and averaging:

[0135] This refers to the previously calculated attitude angles, angular velocities, angular accelerations, and the XZ coordinates of the body's center of mass, as well as the velocities and accelerations in the three directions. Similar to the standardization of the force measurement curve, the attitude curve is standardized for easier comparison. The vertical axis of the curve remains unchanged, while the horizontal axis time t is standardized. Specifically, the horizontal axis is changed from absolute time in seconds to relative time in % of the cycle time, meaning 0% of the cycle time corresponds to the start of the cycle (ground contact), and 100% of the cycle time corresponds to the end of the cycle (ground contact again). Each cycle time is standardized with 101 points (through interpolation), meaning each cycle is divided into 100 equal intervals, represented by 0-100. Also similar to force measurement, attitude data averaging involves averaging N standardized cycle data to obtain single-cycle average attitude data.

[0136] S7, Attitude Output Content:

[0137] The output content refers to the post-processed data that needs to be presented on the interface or stored in a file for later use, excluding the original coordinate data. Specifically, it includes: 1) the values ​​of the head, torso, and left and right limb angles (including angular velocity and angular acceleration) at the moment of ground contact, the end of the buffer, and the moment of push-off, as well as the maximum value, minimum value, and range of the periodic motion. Note that the data here are all averaged results; 2) other parameters including the vertical amplitude of the center of gravity, the lateral displacement of the center of gravity, the ground contact method, the landing point, the inversion / outversion of the feet, the inversion / inversion of the feet, the forward tilt of the head, the forward tilt of the torso, and the stiffness of the lower limbs. The data here are all averaged results; 3) the standardized and averaged posture data curves, including posture angles, angular velocities, angular accelerations, and the XYZ coordinates of the body's center of mass, as well as the velocities and accelerations in the three directions; 4) the posture angles, angular velocities, angular accelerations, and the XYZ coordinates of the body's center of mass, as well as the velocities and accelerations in the three directions, calculated throughout the entire process (after upsampling).

[0138] The main processes of the integrated processing module include kinematic data calculation, posture period division, kinematic data standardization, calculation of force and torque at the hip, knee, and ankle joints of the lower limbs, data averaging, calculation of key running characteristic parameters, and extraction of characteristic curves and feature values. The specific execution steps are as follows: 1) Kinematic data calculation: Based on the single-cycle average posture data, calculate the characteristic values ​​of the angles, angular velocities, and angular accelerations of each joint in the head, trunk, and upper and lower limbs throughout the entire cycle, including the values ​​at the ground contact time, the end of the cushioning, and the push-off time, as well as the maximum and minimum values ​​of the cycle and the range of motion; 2) Posture period division: Divide the kinematic data into cycles based on the filtered cycle times given by the force measurement; 3) Kinematic data standardization; 4) Calculation of force and torque at the hip, knee, and ankle joints of the lower limbs: Based on the standardized single-cycle average force measurement data and single-cycle average posture data, calculate the force and torque of the hip, knee, and ankle joints of the lower limbs through inverse dynamics calculation methods. 5) Data averaging: averaging of kinematic data, lower limb hip, knee, and ankle joint forces and moments; 6) Calculation of key running characteristic parameters: key running characteristic parameters include cadence, ground contact time, stride length, vertical amplitude of center of gravity, landing point, ground contact mode, inward foot eversion angle during ground contact, outward foot occlusion angle during ground contact, knee joint angle at ground contact, thigh swing amplitude, left and right arm swing amplitude, forward and backward arm swing amplitude, trunk forward lean angle, head forward lean angle, maximum normal support reaction force, maximum joint forces and moments of the hip, knee, and ankle joints, etc.; 7) Extraction of characteristic curves and characteristic values: characteristic curves refer to the standardized upper and lower limb joint angles, support reaction forces, and lower limb hip, knee, and ankle joint forces and moments curves. Characteristic values ​​refer to the values ​​of these curves at ground contact time, buffer time, and ground departure time, as well as the maximum value, minimum value, and range of the cycle (maximum value minus minimum value), etc.

[0139] The intelligent assessment and diagnosis module is signal-connected to the data processing module. It compares and evaluates key feature parameters against preset evaluation standards, outputting rating results, problem risks, causes, and corresponding training suggestions. The module's built-in preset evaluation standards are divided into four levels: A, B, C, and D, each corresponding to a preset numerical range. The values ​​for level A are taken from elite athlete data. The output training suggestions include movement difficulty, movement type, movement name, key points of the movement, training volume, training intensity, training frequency, training cycle, and multi-view movement videos demonstrated by professional athletes.

[0140] Both the data display module and the historical comparison module are connected to the intelligent assessment and diagnosis module. The data display module is used for the visual presentation of test data and assessment results, displaying content including personnel and test information, multi-view videos, a 3D running model of the human body, key parameters and rating results, biomechanical and kinematic curves and characteristic values, diagnostic problems, and training suggestions. The historical comparison module is used for multi-dimensional comparative analysis of multiple sets of test data, supporting comparisons of multiple sets of data from the runner or data from different runners. The comparison formats include curve comparisons and characteristic value bar chart comparisons. Through the display interface, runners can clearly and intuitively understand their assessment results, existing problems, causes, and corresponding training suggestions. Through historical comparison, runners can understand the improvement of their running technique or their comparison with others.

Claims

1. A running intelligent assessment and diagnostic system, characterized in that, This includes hardware devices and system software; The hardware device includes a plantar force measurement unit and a markerless motion capture unit; The foot force measurement unit uses a force measurement treadmill or a force measurement track made up of multiple force measurement plates to collect foot force data during the runner's running process; The markerless motion capture unit includes four or more industrial cameras deployed around the foot force measurement unit to synchronously collect kinematic image data of the runner during the running process. The system software includes a plantar force measurement unit control module, an acquisition module, a data processing module, an intelligent assessment and diagnosis module, a data display module, and a historical comparison module; The foot plantar force measurement unit control module is connected to the acquisition module and is used to realize the power-on, power-off, start, stop and speed control of the foot plantar force measurement unit; The acquisition module is connected to the plantar force measurement unit and the markerless motion capture unit respectively, and is used to realize personnel information management, hardware device communication interaction, acquisition parameter configuration and acquisition start and stop control. The data processing module is signal-connected to the acquisition module and includes a force processing module, an attitude processing module, and a comprehensive processing module. The force processing module is used to perform periodic division, effective data filtering, parameter calculation, and standardization processing on the plantar force data. The attitude processing module is used to perform joint point recognition, three-dimensional reconstruction, and kinematic parameter calculation on the kinematic image data. The comprehensive processing module is used for attitude periodic division, calculation of key running feature parameters, and output of key running feature parameters, feature curves, and feature values ​​for the entire run. The intelligent assessment and diagnosis module is connected to the data processing module and is used to compare and evaluate key feature parameters with preset evaluation standards, and output rating results, problem risks, causes and corresponding training suggestions. Both the data display module and the historical comparison module are connected to the intelligent evaluation and diagnosis module. The data display module is used for the visualization of test data and evaluation results, and the historical comparison module is used for multi-dimensional comparative analysis of multiple sets of test data.

2. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The foot pressure measurement unit uses a pressure-measuring treadmill or a pressure-measuring track composed of multiple pressure plates. The collected foot force data includes force data in the XYZ three directions and pressure center data, with a sampling rate of no less than 1000 times / second. The markerless motion capture unit includes four or more industrial cameras deployed around the foot pressure measurement unit, with a camera frame rate of no less than 100 frames / second and an image resolution of no less than [missing information]. Equipped with an AI recognition model, it can automatically identify 25 joints of the human body in an image.

3. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The acquisition module supports three working modes: manual acquisition, fixed-duration acquisition, and automatic trigger acquisition. The manual acquisition is controlled by manual triggering to start and stop the acquisition. The fixed-duration acquisition is completed automatically according to the preset acquisition delay and acquisition duration. The automatic trigger acquisition is started and stopped automatically by photoelectric switch signals at both ends of the force measurement track.

4. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The processing flow of the force measurement module is as follows: The plantar force data is converted to the support reaction force world coordinate system to generate curves for normal force Fz-t, horizontal force Fy-t, and lateral force Fx-t, and invalid pressure center data with normal force below the zero threshold are removed. The timing of single-foot contact and push-off is determined by using the zero-point threshold of normal force, thus completing the division of the running cycle; the timing of the end of the buffer is determined by the moment Fy=0, which is closest to the midpoint of the contact period in the horizontal force Fy-t curve, thus dividing the contact period into the braking phase and the propulsion phase. Valid running cycles were selected based on the average value of the peak normal force during the ground contact period and the rule of three times the standard deviation, and invalid data with fewer than three ground contact times or abnormal peak values ​​were removed. It calculates cadence, ground contact time, track stride, treadmill stride, and track speed, and standardizes the force measurement data in the time dimension. After multi-cycle averaging, it outputs single-cycle average force measurement data.

5. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The processing flow of the attitude processing module is as follows: The AI ​​joint point recognition model extracts the two-dimensional coordinates of 25 human joint points from the acquired images, and then obtains the three-dimensional coordinates of the joint points through multi-view three-dimensional reconstruction. The three-dimensional coordinates of the joints are translated and transformed to make the attitude coordinate system coincide with the force measurement center coordinate system of the plantar force measurement unit; The attitude data is upsampled to the same sampling rate as the force measurement data by spline interpolation, and the motion parameters of the human body's joints, angular velocity, angular acceleration and center of mass are calculated. Based on the effective period of the force measurement data, the attitude data is divided into periods and standardized. After multi-period averaging, the single-period average attitude data is output.

6. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The integrated processing module, based on standardized single-cycle average force measurement data and single-cycle average posture data, calculates the joint forces and moments of the hip, knee, and ankle joints of the lower limbs throughout the entire cycle using inverse dynamics. It also calculates key characteristic parameters of the entire running process, including vertical amplitude of center of gravity, lateral displacement of center of gravity, ground contact mode, landing point, inversion / exversion angle of the foot, inversion / exversion angle of the foot, joint angle characteristics, trunk forward tilt angle, head forward tilt angle, and lower limb stiffness, and extracts the corresponding characteristic curves and feature values ​​of key time nodes.

7. The intelligent running assessment and diagnosis system according to claim 6, characterized in that, The integrated processing module has a built-in human body center of mass calculation submodule. Based on the corrected human body inertial parameters, it first calculates the mass and center of mass coordinates of each body segment of the head, torso, and limbs according to the runner's gender and weight, and then calculates the three-dimensional coordinates of the whole body center of mass by weighted summation of body segment mass.

8. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The preset evaluation criteria of the intelligent assessment and diagnosis module are divided into four levels: A, B, C, and D. Each level corresponds to a preset numerical range, with the A-level values ​​taken from elite athlete data. The output training suggestions include movement difficulty, movement type, movement name, key points of the movement, training volume, training intensity, training frequency, training cycle, and multi-view movement videos demonstrated by professional athletes.

9. The intelligent running assessment and diagnosis system according to claim 1, characterized in that, The data display module displays personnel and test information, multi-view videos, a three-dimensional human running model, key parameters and rating results, mechanical and kinematic curves and characteristic values, diagnostic problems and training suggestions; the historical comparison module supports the comparison of multiple sets of data of the runner or data of different runners, and the comparison forms include curve comparison and characteristic value bar chart comparison.

Images