Athlete movement analysis and training optimization method and system based on big data

Abstract

The present invention discloses a method and system for athlete motion analysis and training optimization based on big data, which relates to the field of sports training technology. The method involves deploying multiple array sensors to collect a trainee's motion data, including static and dynamic data, and simultaneously obtaining the trainee's medical images. A basic model of the trainee is constructed based on the motion data, including a body model and a motion model. A big database of athletes is acquired to construct an athlete database, and a shadow model is generated based on the coupling of the athlete database and the body model. The degree of overlap between the basic model and the shadow model is calculated, and an optimal functional unit is generated based on the overlap. The optimal functional unit is used to guide the training of the remaining body functional units to obtain training data. The present invention constructs a basic model of the trainee and a shadow model of the athlete, setting a scientific target benchmark for personalized training. The data-driven personalized model construction ensures that the pertinence and operability of training optimization are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of sports training technology, and in particular to a method and system for athlete movement analysis and training optimization based on big data. Background Art

[0002] Currently, the field of sports training is facing an increasing number of new opportunities from technological advancements. Traditional sports training models often rely on coaches' subjective experience and trial and error, making it difficult to fully realize the potential of each athlete. With the rapid development of sensing technology, computer vision, biomechanical modeling, and other fields, the use of data-based methods to objectively analyze athletes' movements and optimize personalized training has become a key research direction in this field.

[0003] Some existing intelligent training systems, such as motion capture devices, electromyography detectors, and pressure plates, can indeed provide some basic biomechanical data on athletes during training. However, this data is often fragmented, making it difficult to construct a complete athlete profile. Furthermore, these systems primarily focus on single-source data collection and lack in-depth data analysis and closed-loop optimization. The development of training plans still relies on the coach's subjective judgment, lacking specificity and scientificity. Furthermore, the training targets of these systems are typically targeted at ideal or standard movements, making them difficult to tailor to individual needs. Therefore, even if training outcomes improve to a certain extent, they fail to fully unleash each athlete's maximum competitive potential.

[0004] Therefore, it is of great significance to develop a method for athlete movement analysis and training optimization based on big data. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for athlete movement analysis and training optimization based on big data to address the shortcomings of the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for athlete movement analysis and training optimization based on big data, comprising:

[0007] Deploy multi-array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtain the trainee's medical images;

[0008] The basic model of the trainer based on the motion data includes the body model and the motion model;

[0009] Obtain athlete big data to build an athlete big database, and generate a shadow model based on the coupling of the athlete big database and the body model;

[0010] The degree of overlap between the basic model and the shadow model is calculated, and an optimal functional unit is generated based on the overlap. The training of the remaining body functional units is guided based on the optimal functional unit to obtain training data.

[0011] In a preferred embodiment, the steps of deploying a multi-array sensor to collect the trainee's motion data, including static data and dynamic data, and simultaneously obtaining the trainee's medical images are as follows:

[0012] The multi-array sensor includes an inertial measurement unit, a force perception sensor, a visual capture sensor, and a surface electromyography sensor, while also collecting medical images of the trainee based on medical equipment;

[0013] Based on multi-array sensors and medical imaging, static data of trainees including bone structure parameters, range of motion limits and soft tissue characteristics are acquired;

[0014] Based on multi-array sensors, the trainee's dynamic data including kinematic data, kinetic data, electromyographic activation data and spatiotemporal characteristic data are acquired in real time;

[0015] The kinematic data include joint angles, joint angular velocities, functional unit end trajectories, and functional unit velocities; the dynamic data include joint torques, joint contact forces, joint power, and pressure centers; the electromyography activation data include raw sEMG signals, activation levels, and activation timing; and the spatiotemporal feature data include motion cycle parameters, phase coordination, and motion smoothness.

[0016] In a preferred embodiment, the steps of constructing a basic model of the trainee based on the motion data, including a body model and a motion model, are as follows:

[0017] A body model of the trainee is constructed based on the trainee's static data. The body model is composed of multiple functional units, where the functional units include the trainee's physiological constraints, bone structure, joint degrees of freedom, muscle attachment points, and tendon elasticity parameters;

[0018] Build motion models based on body models and dynamic data:

[0019] ,

[0020] in, Indicates the i The moment of inertia of each functional unit, Indicates the i Coriolis force and centripetal moment of the functional unit, Indicates the i The gravitational moment of each functional unit, Indicates the i External force conversion of functional units, represents the Jacobian matrix;

[0021] The parameter identification algorithm is used to solve the motion model to obtain the trainee's mechanical data.

[0022] In a preferred embodiment, the steps of acquiring athlete big data to construct an athlete big database, constructing optimal motion data based on the athlete big database, and coupling the optimal motion data with the body model to generate a shadow model are as follows:

[0023] Constructing a large athlete database including athlete body database, athlete movement database and athlete training database;

[0024] Real-time acquisition of athlete big data is stored in the athlete big database, and a similarity matching engine is built. Based on the similarity matching engine, the trainer's body model is matched with the athlete's body database to generate similar groups;

[0025] A shadow model is generated by matching the athlete movement database and the athlete training database based on similar groups.

[0026] In a preferred embodiment, the steps of matching the trainee's body model with the athlete's body database based on the similarity matching engine to generate similar groups are:

[0027] Build a similarity matching engine as:

[0028] ,

[0029] in, , , is the dimension weight, represents the Euclidean distance of body shape, represents the anatomical cosine similarity, represents the physiological DTW distance;

[0030] Feature extraction is performed on the static data of the body model to obtain body features including body shape feature vectors, anatomical feature vectors and physiological feature sequences. The body features are input into the similarity matching engine to retrieve the athlete's body data with the highest matching degree as the similarity group.

[0031] In a preferred embodiment, the step of generating a shadow model by matching the athlete movement database and the athlete training database based on similar groups is as follows:

[0032] Retrieving optimal athlete movement data and optimal athlete training data from an athlete movement database and an athlete training database based on physiological constraints using similar groups;

[0033] Introducing optimal athlete movement data and training data into similar groups to generate a shadow model, and generating dynamic data for the shadow model;

[0034] Among them, the dynamic data of the shadow model includes kinematic targets, dynamic targets and neuromuscular control targets. The dynamic targets include the optimal range of joint angles, ideal trajectory templates, ideal functional unit speeds and acceleration curves. The dynamic targets include joint torque ratios and power output timing. The neuromuscular control targets include movement phase optimization and joint coordination optimization.

[0035] In a preferred embodiment, the steps of calculating the degree of overlap between the basic model and the shadow model, generating the optimal functional unit based on the overlap, and guiding the training of the remaining body functional units based on the optimal functional unit to obtain training data are:

[0036] Construct the three-dimensional matching equation between the base model and the shadow model:

[0037] ,

[0038] in, , , represents an adjustable parameter, Indicates the i The trajectory similarity of functional units, Indicates the trainer i The power efficiency of each functional unit, Represents the shadow model i The power efficiency of each functional unit, Indicates the i The joint coordination entropy of each functional unit;

[0039] The basic model is divided into multiple functional chains including lower limb propulsion chain, core stability chain, upper limb manipulation chain, coordination chain and energy conduction chain;

[0040] Based on the three-dimensional matching degree equation, multiple functional units with the highest matching degree in multiple functional chains are obtained as optimal functional units;

[0041] Based on the optimal functional unit and neuromuscular association map, the strong associated muscle groups are identified as weak functional units;

[0042] Generate training data for weak functional units based on dynamic data of shadow models.

[0043] The present invention also provides an athlete movement analysis and training optimization system based on big data, including:

[0044] Data acquisition module: deploys multiple array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtains the trainee's medical images;

[0045] Basic model generation module: connected to the data acquisition module, builds the basic model of the trainee based on the motion data, including the body model and motion model;

[0046] Shadow model generation module: connects with the data acquisition module, obtains athlete big data to build an athlete database, and generates shadow models based on the athlete database coupled with the body model;

[0047] Training data generation module: connected to the shadow model generation module, calculates the overlap between the basic model and the shadow model, generates the optimal functional unit based on the overlap, guides the training of the remaining body functional units based on the optimal functional unit, and obtains training data.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0049] 1. This invention constructs a highly refined and personalized "base model" for each trainee. Based on this, a targeted "shadow model" is established through big data analysis as an optimization benchmark. Traditional training often relies on experience or general models, which makes it difficult to capture individual differences. This method first utilizes multi-array sensors and medical imaging to achieve comprehensive and high-precision collection of the trainee's static and dynamic data. This multi-dimensional data is used to construct a detailed "body model" and a "motion model" based on biomechanical principles. Using a complex similarity matching engine, the system intelligently selects "similar groups" that highly match the trainee's physiological characteristics from a large database of athletes. From this, a "shadow model" representing the individual's optimal potential is generated. This shadow model represents the optimal movement techniques and training plans of top athletes with similar physical characteristics to the trainee, setting a scientific target benchmark for personalized training. The data-driven personalized model construction ensures that the targeted and operational nature of training optimization is greatly improved.

[0050] 2. The present invention identifies the trainee's optimal functional units and weak functional units in each functional chain by calculating the three-dimensional matching degree between the trainee's basic model and the shadow model in key biomechanical parameters. For these weak links, the system can further combine neuromuscular association analysis to accurately locate the key muscle groups that affect training efficiency. Finally, relying on the dynamic data of the shadow model, targeted training plans and data are generated for these weak functional units; precise training intervention based on data analysis has greatly improved the targetedness and effectiveness of training, which can not only significantly improve sports techniques, but also effectively prevent sports injuries. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0052] Figure 1 is a flow chart of the method of the present invention;

[0053] Figure 2 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] Example 1, please refer to Figure 1 As shown, the athlete movement analysis and training optimization method based on big data described in this embodiment includes:

[0056] S1. Deploy multiple array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtain the trainee's medical images;

[0057] S2. Build a basic model of the trainee based on the motion data, including a body model and a motion model;

[0058] S3, obtaining athlete big data to build an athlete big database, and generating a shadow model based on the athlete big database coupled with the body model;

[0059] S4. calculating the degree of overlap between the basic model and the shadow model, generating an optimal functional unit based on the degree of overlap, and guiding the training of the remaining body functional units based on the optimal functional unit to obtain training data;

[0060] As described in steps S1-S4 above, the field of sports training is currently facing an increasing number of new opportunities from technological advancements. Traditional sports training models often rely on coaches' subjective experience and trial and error, making it difficult to fully realize the potential of each athlete. With the rapid development of sensing technology, computer vision, biomechanical modeling, and other fields, using data-based methods to objectively analyze athletes' movements and optimize personalized training has become a key research direction in this field.

[0061] Some existing intelligent training systems, such as motion capture devices, electromyography detectors, and pressure plates, can indeed provide some basic biomechanical data on athletes during training. However, this data is often fragmented, making it difficult to construct a complete athlete profile. Furthermore, these systems primarily focus on single-source data collection and lack in-depth data analysis and closed-loop optimization. The development of training plans still relies on the coach's subjective judgment, lacking specificity and scientificity. Furthermore, the training targets of these systems are typically targeted at ideal or standard movements, making them difficult to tailor to individual needs. Therefore, even if training outcomes improve to a certain extent, they fail to fully unleash each athlete's maximum competitive potential.

[0062] The present invention constructs a highly refined and personalized "base model" for each trainee. Based on this, a targeted "shadow model" is established through big data analysis as an optimization benchmark. Traditional training often relies on experience or universal models, which makes it difficult to capture individual differences. This method first utilizes multi-array sensors and medical imaging to achieve comprehensive and high-precision collection of the trainee's static and dynamic data. The multi-dimensional data is used to construct a detailed "body model" and a "motion model" based on biomechanical principles. The system uses a complex similarity matching engine to intelligently screen "similar groups" that highly match the trainee's physiological characteristics from a large database of athletes, and uses this to generate a "shadow model" representing the individual's optimal potential. This shadow model represents the optimal movement techniques and training plans of top athletes with similar physical characteristics to the trainee, setting a scientific target benchmark for personalized training. The data-driven personalized model construction ensures that the targetedness and operability of training optimization are greatly improved.

[0063] By calculating the three-dimensional matching degree between the trainee's basic model and the shadow model in key biomechanical parameters, the trainee's optimal functional units and weak functional units in each functional chain are identified. For these weak links, the system can further combine neuromuscular correlation analysis to accurately locate the key muscle groups that affect training efficiency. Finally, relying on the dynamic data of the shadow model, targeted training plans and data are generated for these weak functional units. Precise training intervention based on data analysis has greatly improved the targetedness and effectiveness of training, which can not only significantly improve sports techniques, but also effectively prevent sports injuries.

[0064] In one embodiment, the step S1 of deploying a multi-array sensor to collect motion data of a trainee, including static data and dynamic data, and simultaneously acquiring medical images of the trainee comprises:

[0065] S11, multi-array sensor including inertial measurement unit, force perception sensor, visual capture sensor and surface electromyography sensor, while collecting medical images of the trainee based on medical equipment;

[0066] S12. Obtain the trainee's static data including bone structure parameters, range of motion limits, and soft tissue characteristics based on multi-array sensors and medical imaging;

[0067] S13, based on multi-array sensors, real-time acquisition of the trainee's dynamic data, including kinematic data, kinetic data, electromyographic activation data, and spatiotemporal feature data;

[0068] S14, wherein the kinematic data includes joint angles, joint angular velocities, functional unit end trajectories, and functional unit velocities; the dynamic data includes joint torques, joint contact forces, joint powers, and pressure centers; the myoelectric activation data includes raw sEMG signals, activation levels, and activation timing; and the spatiotemporal feature data includes motion cycle parameters, phase coordination, and motion smoothness;

[0069] As described in the above steps S11-S14, the multi-array sensor system can select a 9-axis IMU with 200Hz sampling to ensure an accuracy of ±0.5° through the integration of inertial measurement units deployed at the key joint positions of the trainee. The embedded mechanical sensing sensor includes a piezoelectric plate and an insole pressure distribution system. The optical motion capture system uses 8 infrared cameras with a sampling frequency of 250Hz to ensure that the mark point error is <0.2mm. The surface electromyography sensor uses dual differential electrodes 2000Hz and 20-450Hz bandpass filtering. Synchronous medical imaging equipment collects the full-dimensional data of the trainee; in static data extraction, the CT reconstructed bone mesh is aligned with the optical static calibration framework, and combined with the tendon elastic modulus and joint range of motion test results measured by ultrasound to generate a personalized data set containing bone parameters, activity limits and soft tissue characteristics; dynamic During data acquisition, multi-source synchronization is achieved through hardware trigger signals. Kinematic data include joint angles calculated by IMU and optical fusion (such as knee flexion of 62.3°), angular velocity processed by fourth-order Butterworth filtering (cutoff frequency 8Hz), and end-of-hand trajectories (spatial error ±1.1mm). Dynamic data include joint torques calculated by inverse dynamics (such as ankle plantar flexion torque of 2.8Nm / kg), contact force (hip joint peak value of 3.2 times body weight), and pressure center trajectory (sampling interval 2ms). Myoelectric activation data include raw sEMG signals, activation degree after RMS rectification, and force timing identified by the differential threshold method. Spatiotemporal feature data are extracted through phase space reconstruction to extract motion cycle phase (such as gait stance period accounting for 58.7%), cross-joint coordination index (hip-knee angular velocity correlation coefficient >0.85), and acceleration derivative smoothness.

[0070] In one embodiment, the step S2 of constructing a basic model of the trainee based on the motion data, including a body model and a motion model, includes:

[0071] S21. Constructing a body model of the trainee based on the trainee's static data, where the body model is composed of multiple functional units, wherein the functional units include the trainee's physiological constraints, skeletal structure, joint degrees of freedom, muscle attachment points, and tendon elasticity parameters;

[0072] S22. Constructing a motion model based on the body model and dynamic data:

[0073] ,

[0074] S23, among which, Indicates the i The moment of inertia of each functional unit, Indicates the i Coriolis force and centripetal moment of the functional unit, Indicates the i The gravitational moment of each functional unit, Indicates the i External force conversion of functional units, represents the Jacobian matrix;

[0075] S24, using a parameter identification algorithm to solve the motion model and obtain the trainee's mechanical data;

[0076] As described in the above steps S21-S24, a modular body model is constructed based on static data (such as CT-reconstructed femur length of 432 mm and shoulder joint internal rotation / external rotation freedom of ±80°), and the human body is divided into biomechanical functional units such as the lower limb propulsion chain (including ankle, knee and hip joint freedom), the core stability chain (lumbar spine-pelvic stability parameters), etc. Each functional unit integrates bone geometry such as the patellar surface curvature radius of 32.1±0.8 mm, muscle attachment point coordinates such as the proximal rectus femoris attachment point being 12.3 mm away from the anterior inferior iliac spine, and tendon nonlinear elastic models such as the Achilles tendon stiffness coefficient k=2.1 kN / m. The motion model uses the functional unit to decouple the dynamic equations, and the parameter identification algorithm is used to solve the motion model to obtain the trainee's mechanical data. For example, when a basketball player performs take-off training, when the athlete bends his knee 60°, the knee angle acceleration The mass matrix includes the moment of inertia of the lower leg. The Coriolis force term C calculates the additional torque generated by the knee joint rotation (120° / s). The gravity moment G is determined by the center of mass position of the body model (38 cm above the knee). The external force term maps the ground reaction force (1800N) to the knee joint space. Based on the above data, the recursive least squares method is used to dynamically solve the unknown quantities in the equation to obtain the peak torque of the knee extensor group. , and updates the tendon stiffness in real time. The Achilles tendon k value increases from 1.8 to 2.3kN / m during the take-off and accumulation period. This model can accurately quantify the mechanical interactions between functional units, such as identifying the hip-knee power conduction delay in the lower limb propulsion chain, and provide a timing optimization target for explosive power training.

[0077] In one embodiment, the step S3 of acquiring athlete big data to build an athlete big database, building optimal motion data based on the athlete big database, and coupling the optimal motion data with the body model to generate a shadow model includes:

[0078] S31, constructing a large athlete database including an athlete body database, an athlete movement database, and an athlete training database;

[0079] S32: Acquire the athlete big data in real time and store it in the athlete big database, build a similarity matching engine, match the trainee's body model with the athlete body database based on the similarity matching engine, and generate similar groups;

[0080] S33, generating a shadow model by matching an athlete sports database and an athlete training database based on similar groups;

[0081] As described in steps S31-S33 above, a comprehensive athlete database is established, comprising three key sub-databases: an athlete body database (recording physiological, physical, biomechanical, and injury data), an athlete movement database (collecting technical movement, strength, speed, and competition performance data), and an athlete training database (covering process data such as training plans, load, recovery, and nutrition). Simultaneously, an advanced similarity matching engine is constructed, utilizing data mining techniques to analyze the athlete's physical data and accurately match the trainee's body model with the athlete body database, thereby generating groups with similar characteristics. Finally, based on these generated similar groups, the athlete movement database and training database are matched, and the optimal movement data is selected. This data is then coupled with the trainee's body model to construct a shadow model. This shadow model not only reflects the trainee's athletic potential and performance but also provides data support for personalized training programs, optimizing training effectiveness and promoting overall improvement in the athlete's competitive level.

[0082] In one embodiment, the step S32 of matching the trainee's body model with the athlete's body database based on the similarity matching engine to generate similar groups includes:

[0083] S321. Constructing a similarity matching engine:

[0084] ,

[0085] S322, among which, , , is the dimension weight, represents the Euclidean distance of body shape, represents the anatomical cosine similarity, represents the physiological DTW distance;

[0086] S323, extracting features from the static data of the body model to obtain body features including a body shape feature vector, an anatomical feature vector, and a physiological feature sequence, inputting the body features into a similarity matching engine, and retrieving the body data of the athletes with the highest matching degree as a similar group;

[0087] As described in the above steps S321-S323, the system analyzes the static features of the trainee through multimodal data fusion technology: based on CT / MRI images, the iterative closest point algorithm (ICP) is used to align the bone point cloud with the sensor coordinate system, and a body feature vector containing 17-dimensional parameters is extracted, such as the absolute length of the femur 483.2mm and the relative Achilles tendon / calf length ratio of 0.42; relying on the medical image segmentation model, 23 biomechanical landmarks are quantified to generate anatomical feature vectors, such as the coordinates of the acetabulum rotation center [132.5, 87.3, -15.8] mm and the anterior cruciate ligament stiffness value of 182N / mm; combined with the six-degree-of-freedom robotic arm passive traction protocol, a physiological feature sequence is constructed at 10Hz sampling within the range of 0-90° joint motion, such as the knee flexion The peak stiffness at 30° bending is 285Nm / rad and its time derivative curve. After the above feature vectors are input into the dynamic weighted similarity matching engine, the engine loads the preset weight template according to the sport type and calculates the three-dimensional similarity in parallel. The body shape dimension quantifies the difference through Euclidean distance. For example, the L2 distance between the trainer's arm span of 198cm and the database sample of 201cm is 0.08. The anatomical dimension uses cosine similarity to evaluate the structural alignment. For example, the dot product operation of the scapula inclination vector obtains a similarity of 0.992. The physiological dimension uses dynamic time regularization. For example, the optimal bending path cost of the knee joint stiffness curve is 19.3. Finally, a comprehensive similarity score is synthesized. The dimension weights in the similarity matching engine are dynamically adjusted based on the sport type. For example, for confrontational projects, the improvement The value of The value of anaerobic burst is increased The core advantage of this step is that it achieves precise biomechanical matching through a multi-dimensional dynamic matching engine, effectively solving the training risks caused by "similar body shapes but incompatible force chains" in traditional experience matching, and providing a safe and available top-level biomechanical blueprint for the shadow model.

[0088] In one embodiment, the step S33 of generating a shadow model by matching the athlete movement database and the athlete training database based on similar groups includes:

[0089] S331, using similar groups based on physiological constraints to retrieve optimal athlete motion data and optimal athlete training data from an athlete motion database and an athlete training database;

[0090] S332, introducing the optimal athlete movement data and training data into a similar group to generate a shadow model, and generating dynamic data of the shadow model;

[0091] S333, wherein the dynamic data of the shadow model includes kinematic targets, dynamic targets, and neuromuscular control targets. The dynamic targets include the optimal range of joint angles, ideal trajectory templates, ideal functional unit speeds, and acceleration curves. The dynamic targets include joint torque ratios and power output timing. The neuromuscular control targets include motion phase optimization and joint coordination optimization.

[0092] As described in the above steps S331-S333, the system first intelligently retrieves the optimal data from the athlete motion database and training database based on the physiological constraints of similar groups (such as the joint range of motion limit ±10%, ligament stiffness threshold, etc.); through the biomechanical compatibility filter (for example, excluding sprint take-off data with knee flexion >140° to prevent the risk of acetabular impact), efficient movement patterns and related training plans that are adapted to the trainee's anatomical structure are screened out, and then the spatiotemporal normalization algorithm is used to couple the optimal data with the body model of the similar group. In actual implementation, the motion trajectories of different sampling rates are unified to the standard time axis through B-spline curve fitting technology, and the dynamic coordinate system is aligned with the skeleton chain as the reference, finally generating a high-fidelity shadow model and outputting three types of dynamic targets. Objectives: Kinematic goals include the optimal range of joint angles, such as 105°±3° elbow flexion when shooting a basketball, and ideal trajectory templates, such as a parabolic shooting path based on NURBS curve interpolation. Dynamic goals are refined to joint torque ratios, such as a 4:3:3 hip-knee-ankle torque distribution ratio during the sprint kick-off phase and power output timing (peak power window ≤ 50ms). Neuromuscular control goals are achieved through phase optimization matrices, such as the high jump takeoff leg muscle activation sequence: gastrocnemius → quadriceps → gluteus maximus, with a phase difference of ≤ 30ms, and joint coordination optimization functions. The joint coordination optimization function is constructed using the Lyapunov exponent to quantify the fluctuation tolerance of the core stability chain. For example, the shadow model generated for a pole vaulter explicitly requires that the pelvic rotation angular acceleration during the run-up phase reach The ratio of the shoulder joint external rotation torque to the trunk lateral flexion torque at the moment of pole contact is locked at 1.25±0.05. The data comes from the cross-database feature fusion of three World Championship medalists in a similar group, ensuring that each target has the dual guarantee of anatomical feasibility and championship performance genes.

[0093] In one embodiment, the step S4 of calculating the degree of overlap between the base model and the shadow model, generating an optimal functional unit based on the degree of overlap, and guiding the training of the remaining body functional units based on the optimal functional unit to obtain training data includes:

[0094] S41. Construct a three-dimensional matching equation between the base model and the shadow model:

[0095] ,

[0096] S42, among which, , , represents an adjustable parameter, Indicates the i The trajectory similarity of functional units, Indicates the trainer i The power efficiency of each functional unit, Represents the shadow model i The power efficiency of each functional unit, Indicates the i The joint coordination entropy of each functional unit;

[0097] S43, divide the basic model into multiple functional chains including lower limb propulsion chain, core stability chain, upper limb control chain, coordination chain and energy conduction chain;

[0098] S44, obtaining multiple functional units with the highest matching degree in multiple functional chains as optimal functional units based on the three-dimensional matching degree equation;

[0099] S45. Based on the optimal functional unit and neuromuscular association map, identify the strong associated muscle groups as weak functional units;

[0100] S46. Generate training data for weak functional units based on dynamic data of the shadow model;

[0101] As described in steps S41-S46 above, a three-dimensional matching equation is constructed between the base model and the shadow model. This equation reflects the comprehensive impact of different parameters on the model matching. The adjustable parameters are used to adjust the importance of each factor on the final matching score. The trajectory similarity is quantified by the dynamic time warping algorithm to measure the difference in the motion path of the functional unit terminal. and Represents the trainer and the shadow model in iThe force efficiency of each functional unit is defined as the ratio of output power to metabolic energy consumption. The collaborative volatility of related joints is calculated based on Shannon entropy theory to obtain the joint coordination entropy, which is used to quantify the degree of coordination between the joints of the functional unit during movement. Then the basic model is divided into multiple functional chains, including the lower limb propulsion chain, core stability chain, upper limb control chain, coordination chain and energy conduction chain. Among them, the lower limb propulsion chain refers to the functional chain formed by the leg muscle groups and their mutual cooperation during movement, especially in running, jumping and other explosive movements. This chain mainly includes the gluteus maximus, anterior thigh muscles, hamstrings, calf muscles, etc.; the core function of the lower limb propulsion chain is to overcome the ground reaction force by providing thrust to achieve the body's forward movement or jump; the core stabilization chain refers to the muscle group surrounding the trunk and pelvis, mainly including the abdominal muscles, back muscles and trapezius muscles, etc. The key task of this chain is to maintain the stability of the central area of ​​​​the body and support posture control during body movement. The role of the core stabilization chain is to provide a stable foundation to help the effective function of the upper and lower limbs. By enhancing the strength and endurance of the core area, it can optimize sports performance, reduce the risk of injury, and improve sports skills, such as turning, throwing, and other movements that require high stability; the upper limb control chain involves the muscle system of the shoulders, arms and hands, mainly composed of the deltoid, biceps, triceps, serratus anterior and grip muscles. The main task of this chain The main function of the coordination chain is to perform precise movement control, force application and power transmission, which is usually used in throwing, pushing, pulling and other movements; the coordination chain refers to the coordination ability of various parts of the body during movement, ensuring effective cooperation between the limbs and making the movement smooth. This chain involves the central nervous system's ability to control movement, including limb coordination of the trunk, upper and lower limbs, and head. Especially in complex and fast movements such as dance and ball sports, the effectiveness of the coordination chain depends not only on strength and flexibility, but also on good reaction time and body awareness to ensure that the movement pattern can be adjusted quickly and properly in a dynamic environment; the energy conduction chain focuses on how energy is effectively transferred between different parts of the body to ensure that the force generated by the muscle groups can be correctly converted into movement. This chain involves the movement of the joints and the contraction mechanism of the muscles, and can be considered an efficient force transmission system.A good energy conduction chain not only improves exercise efficiency but also reduces energy loss during exercise, thereby enhancing athletic performance. Based on the matching equation, the functional unit with the highest matching degree within each functional chain is identified as the optimal functional unit. Pre-stored neuromuscular association maps are then used to identify weak functional units that are strongly associated with the optimal functional unit, such as the gastrocnemius-soleus complex associated with the optimal ankle joint module in the lower limb propulsion chain. Finally, training data for the weak functional unit is generated based on the dynamic data of the shadow model, including kinematic targets, kinetic targets, and neuromuscular control targets. Within the dynamic data of the shadow model, performance data of the weak functional unit when completing a specific task is precisely located. Key dynamic data includes electromyographic signal patterns, kinematic data, kinetic data, coordination indicators, and energy conduction efficiency. The extracted dynamic data of the weak functional unit in the shadow model is converted into quantifiable and measurable training target parameters. These parameters should be slightly lower than the shadow model level but higher than the trainee's current level. The trainee then specifically stimulates and trains the weak functional unit, and continuously compares it with the shadow model during training to observe the training effect.

[0102] Example 2, please refer to Figure 2 As shown, the athlete movement analysis and training optimization system based on big data described in this embodiment includes:

[0103] Data acquisition module: deploys multiple array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtains the trainee's medical images;

[0104] Basic model generation module: connected to the data acquisition module, builds the basic model of the trainee based on the motion data, including the body model and motion model;

[0105] Shadow model generation module: connects with the data acquisition module, obtains athlete big data to build an athlete database, and generates shadow models based on the athlete database coupled with the body model;

[0106] Training data generation module: connected to the shadow model generation module, calculates the overlap between the basic model and the shadow model, generates the optimal functional unit based on the overlap, guides the training of the remaining body functional units based on the optimal functional unit, and obtains training data.

[0107] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. Athlete movement analysis and training optimization method based on big data, characterized by: Deploy multi-array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtain the trainee's medical images; The basic model of the trainer based on the motion data includes the body model and the motion model; Obtain athlete big data to build an athlete big database, and generate a shadow model based on the coupling of the athlete big database and the body model; Calculating the degree of overlap between the basic model and the shadow model, generating an optimal functional unit based on the overlap, and guiding the training of the remaining body functional units based on the optimal functional unit to obtain training data; The steps of constructing a basic model of the trainee based on the motion data, including a body model and a motion model, are as follows: A body model of the trainee is constructed based on the trainee's static data. The body model is composed of multiple functional units, where the functional units include the trainee's physiological constraints, bone structure, joint degrees of freedom, muscle attachment points, and tendon elasticity parameters; Build motion models based on body models and dynamic data: ， in, Indicates the i The moment of inertia of each functional unit, Indicates the i The Coriolis force and centripetal moment of each functional unit, Indicates the i The gravitational moment of each functional unit, Indicates the i External force conversion of functional units, represents the Jacobian matrix; The parameter identification algorithm is used to solve the motion model to obtain the trainee's mechanical data.

2. The method for athlete movement analysis and training optimization based on big data according to claim 1, characterized in that: The steps of deploying multiple array sensors to collect the trainee's motion data, including static data and dynamic data, and simultaneously obtaining the trainee's medical images are as follows: The multi-array sensor includes an inertial measurement unit, a force perception sensor, a visual capture sensor, and a surface electromyography sensor, while also collecting medical images of the trainee based on medical equipment; Based on multi-array sensors and medical imaging, static data of trainees including bone structure parameters, range of motion limits and soft tissue characteristics are acquired; Based on multi-array sensors, the trainee's dynamic data including kinematic data, kinetic data, electromyographic activation data and spatiotemporal characteristic data are acquired in real time; The kinematic data include joint angles, joint angular velocities, functional unit end trajectories, and functional unit velocities; the dynamic data include joint torques, joint contact forces, joint power, and pressure centers; the electromyography activation data include raw sEMG signals, activation levels, and activation timing; and the spatiotemporal feature data include motion cycle parameters, phase coordination, and motion smoothness.

3. The method for athlete movement analysis and training optimization based on big data according to claim 1, characterized in that: The steps of acquiring athlete big data to build an athlete big database, building optimal motion data based on the athlete big database, and coupling the optimal motion data with the body model to generate a shadow model are as follows: Constructing a large athlete database including athlete body database, athlete movement database and athlete training database; Real-time acquisition of athlete big data is stored in the athlete big database, and a similarity matching engine is built. Based on the similarity matching engine, the trainer's body model is matched with the athlete's body database to generate similar groups; A shadow model is generated by matching the athlete movement database and the athlete training database based on similar groups.

4. The method for athlete movement analysis and training optimization based on big data according to claim 3, characterized in that: The steps of matching the trainee's body model with the athlete's body database based on the similarity matching engine to generate similar groups are as follows: Build a similarity matching engine as: ， in, , , is the dimension weight, represents the Euclidean distance of body shape, represents the anatomical cosine similarity, represents the physiological DTW distance; Feature extraction is performed on the static data of the body model to obtain body features including body shape feature vectors, anatomical feature vectors and physiological feature sequences. The body features are input into the similarity matching engine to retrieve the athlete's body data with the highest matching degree as the similarity group.

5. The method for athlete movement analysis and training optimization based on big data according to claim 3, characterized in that: The steps of matching the athlete sports database and the athlete training database based on similar groups to generate a shadow model are as follows: Retrieving optimal athlete movement data and optimal athlete training data from an athlete movement database and an athlete training database based on physiological constraints using similar groups; Introducing optimal athlete movement data and training data into similar groups to generate a shadow model, and generating dynamic data for the shadow model; Among them, the dynamic data of the shadow model includes kinematic targets, dynamic targets and neuromuscular control targets. The dynamic targets include the optimal range of joint angles, ideal trajectory templates, ideal functional unit speeds and acceleration curves. The dynamic targets include joint torque ratios and power output timing. The neuromuscular control targets include movement phase optimization and joint coordination optimization.

6. The method for athlete movement analysis and training optimization based on big data according to claim 1, characterized in that: The steps of calculating the degree of overlap between the basic model and the shadow model, generating the optimal functional unit based on the degree of overlap, and guiding the training of the remaining body functional units based on the optimal functional unit to obtain training data are as follows: Construct the three-dimensional matching equation between the base model and the shadow model: ， in, , , represents an adjustable parameter, Indicates the i The trajectory similarity of functional units, Indicates the trainer i The power efficiency of each functional unit, Represents the shadow model i The power efficiency of each functional unit, Indicates the i The joint coordination entropy of each functional unit; The basic model is divided into multiple functional chains including lower limb propulsion chain, core stability chain, upper limb manipulation chain, coordination chain and energy conduction chain; Based on the three-dimensional matching degree equation, multiple functional units with the highest matching degree in multiple functional chains are obtained as optimal functional units; Based on the optimal functional unit and neuromuscular association map, the strong associated muscle groups are identified as weak functional units; Generate training data for weak functional units based on dynamic data of shadow models.

7. A system for athlete motion analysis and training optimization based on big data, for implementing the method for athlete motion analysis and training optimization based on big data according to any one of claims 1 to 6, characterized in that: Data acquisition module: deploys multiple array sensors to collect the trainee's motion data, including static and dynamic data, and simultaneously obtains the trainee's medical images; Basic model generation module: connected to the data acquisition module, builds the basic model of the trainee based on the motion data, including the body model and motion model, and builds the trainee's body model based on the trainee's static data. The body model is composed of multiple functional units, where the functional units include the trainee's physiological constraints, bone structure, joint degrees of freedom, muscle attachment points, and tendon elasticity parameters; Build motion models based on body models and dynamic data: ， in, Indicates the i The moment of inertia of each functional unit, Indicates the i The Coriolis force and centripetal moment of each functional unit, Indicates the i The gravitational moment of each functional unit, Indicates the i External force conversion of functional units, represents the Jacobian matrix; Use parameter identification algorithm to solve the motion model to obtain the trainee's mechanical data; Shadow model generation module: connects with the data acquisition module, obtains athlete big data to build an athlete database, and generates shadow models based on the athlete database coupled with the body model; Training data generation module: connected to the shadow model generation module, calculates the overlap between the basic model and the shadow model, generates the optimal functional unit based on the overlap, guides the training of the remaining body functional units based on the optimal functional unit, and obtains training data.

Images