Real-time correction method, system and equipment for martial arts in Chinese examination and medium

Abstract

The invention relates to the technical field of education and teaching, and particularly discloses a real-time correction method, system, equipment and medium for Chinese exam martial arts, and the method comprises the steps: obtaining a Chinese exam martial arts standard action library which comprises martial arts routine action sequences conforming to the national Chinese exam martial arts scoring detailed rule, each action sequence is composed of calibrated standard three-dimensional human body joint point trajectory data; in the training process, a multi-view two-dimensional image sequence of a plurality of students executing the same martial art routine is collected in real time through a camera array composed of a plurality of synchronous high-speed cameras arranged in a training field; according to the invention, the limitation of a single sensor is broken through, the AHRS / IMU inertial navigation system and multi-view visual attitude estimation are creatively and deeply fused, advantage complementation is realized, the integrity and stability of motion capture are remarkably improved, and the method is particularly suitable for complex posture changes such as flying, turning, low attitude and the like which are common in martial arts.

Description

Technical Field

[0001] This invention belongs to the field of educational technology, specifically relating to a method, system, equipment, and medium for real-time correction of martial arts movements in the middle school entrance examination. Background Technology

[0002] The martial arts movements in the middle school entrance examination have clear technical specifications and scoring standards. Students are required to complete standardized movements in specific routines within a specified time, such as bow stance punch, horse stance block punch, and resting stance downward punch. There are also strict limitations on body posture, limb angles, center of gravity transfer, and movement rhythm.

[0003] However, in traditional teaching models, the instruction and evaluation of martial arts movements heavily rely on teachers' experience and judgment, resulting in problems such as strong subjectivity, delayed feedback, and difficulty in quantification. Especially in large-class teaching environments, teachers cannot accurately monitor and correct the details of each student's movements in real time, leading to low training efficiency, serious accumulation of movement deviations, and affecting the final score. In addition, although existing video-based motion capture technology has been applied in the field of sports research, it generally suffers from problems such as difficulty in occlusion recognition, insufficient accuracy of 3D reconstruction, and high dynamic response latency; while inertial sensing devices can achieve motion acquisition without visual occlusion, they are easily affected by drift errors and initial calibration, and cannot meet the high-fidelity reproduction requirements of martial arts movements in the middle school entrance examination when used alone.

[0004] Therefore, it is necessary to propose a method, system, equipment, and medium for real-time correction of martial arts movements in the middle school entrance examination, so as to at least partially solve the problem that teachers cannot accurately check and correct the details of each student's movements in real time in the existing technology.

[0005] The information disclosed above in this background section is only for enhancing the understanding of the background section of this invention, and therefore may include prior art that is not known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method, system, device, and medium for real-time correction of martial arts movements in the middle school entrance examination, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for real-time correction of martial arts movements in the middle school entrance examination includes:

[0009] Obtain a standard movement library for martial arts in the high school entrance examination. The standard movement library contains martial arts routine movement sequences that conform to the national high school entrance examination martial arts scoring rules. Each movement sequence consists of calibrated standard three-dimensional human joint point trajectory data.

[0010] During the training process, a camera array consisting of multiple synchronous high-speed cameras deployed on the training ground is used to collect multi-view two-dimensional image sequences of multiple students performing the same martial arts routine in real time. At the same time, the angular velocity, acceleration and orientation information of body segments are obtained through multiple IMU sensors worn on key parts of the human body.

[0011] The coordinates of two-dimensional human joints are extracted from images from various perspectives using a deep learning pose estimation algorithm. Based on epipolar geometric constraints and triangulation principles, and combined with time synchronization signals, the two-dimensional joints from multiple perspectives are mapped to a unified three-dimensional spatial coordinate system to obtain a preliminary three-dimensional human pose.

[0012] Polar geometric constraint formula:

[0013]

[0014] In the formula, x1 and x2 are the corresponding homogeneous coordinates of two-dimensional joints, and F is the basic matrix;

[0015] Trigonometric measurement formula:

[0016]

[0017] In the formula, P1 and P2 are the projection matrices of each camera, and X is the coordinate of the three-dimensional joint point;

[0018] A sensor spatiotemporal alignment strategy is adopted to register and fuse the attitude data of the IMU sensor with the preliminary three-dimensional human posture in a preset common reference coordinate system, thereby constructing a complete, continuous, and occlusion-resistant full-body motion trajectory model.

[0019] The whole-body motion trajectory model is time-aligned and similarity-matched with the corresponding standard three-dimensional human joint trajectory data in the standard motion library. The motion difference is calculated by dynamic time warping algorithm, and multi-dimensional refined error analysis is performed on key bony landmarks.

[0020] The formula for weighted Euclidean distance at key points:

[0021]

[0022] In the formula, N is the number of joints, ω k The weighting is based on the national middle school entrance examination martial arts scoring rules, X k It is the three-dimensional coordinate of the k-th joint of the standard movement, Y k These are the three-dimensional coordinates of the k-th joint of the student's movement;

[0023] Angle error formula:

[0024]

[0025] In the formula, v1 and v2 are vectors between joints, used to represent the direction between two joints, and θ k It is the angle between these two vectors, used to represent the joint rotation error;

[0026] Dynamic Time Warping Algorithm Formula:

[0027] D(a,b)=||X a -Y b ||+min{D(a-1,b),D(a,b-1),D(a-1,b-1)}

[0028] In the formula, X a and Y b These are the 3D joint vectors of the standard action and the student action in the same frame, respectively, and D(a,b) is the cumulative minimum distance;

[0029] Based on the preset national middle school entrance examination martial arts scoring rules engine, it automatically generates real-time correction instructions that include the type of movement error, the time of occurrence, the degree of deviation, and improvement suggestions. It also provides students with multi-channel interactive feedback through AR augmented reality interface, voice broadcast, or haptic feedback device.

[0030] Preferably, the process of constructing the standard three-dimensional human joint trajectory data includes:

[0031] When professional demonstrators perform standard martial arts movements, an inertial navigation sensor (AHRS) is used in conjunction with an attitude reference system and an inertial measurement unit (IMU) to jointly collect martial arts movement information.

[0032] A pose estimation algorithm is used to identify and extract the joint information of the human body from martial arts action videos, and then fuse it with the collected martial arts action information to form a complete motion trajectory.

[0033] The fused data is modeled using 3D modeling software to generate a human body model and 3D trajectory that conform to the national middle school entrance examination martial arts scoring rules, namely the standard 3D human joint point trajectory data.

[0034] Preferably, the IMU sensors are distributed across nine key segments of the human body, including the upper limbs, torso, and lower limbs: left / right upper arm, left / right forearm, chest, waist, left / right thigh, and left / right calf. Each sensor has a built-in inertial navigation sensor (AHRS) module with a sampling frequency of not less than 200Hz.

[0035] Preferably, the physical location, shooting angle, and coverage of the camera array are automatically calculated by an AI-driven visualization deployment platform based on the training space geometry, which recommends the optimal configuration scheme and generates a virtual installation simulation diagram for user confirmation.

[0036] Preferably, the deep learning pose estimation algorithm employs a two-stream graph convolutional network architecture, and the graph convolution formula is as follows:

[0037]

[0038] In the formula, N(i) represents the nodes' neighbors, and H... j (l) Let c be the l-th layer feature of the j-th key node's neighbors. ij W is the normalization coefficient. (l) Let σ(·) be the learnable weight matrix, and let σ(·) be the activation function.

[0039] The first stream processes single-view 2D joint point sequences, while the second stream fuses multi-view cross-frame spatiotemporal context relationships to output highly robust 3D human pose prediction results. The dual-stream graph convolutional network architecture uses the standard 3D human joint point trajectory dataset as training samples and employs a self-supervised contrastive learning strategy to achieve accurate localization of human joint points and key bony markers in 3D space.

[0040] Preferably, the method for establishing the common reference coordinate system includes:

[0041] Using a known spatially distributed IMU node group fixed on a calibration rod, a spatial rigid body transformation is performed to determine the rotation matrix and translation vector from the IMU local coordinate system to the world coordinate system.

[0042] The camera's intrinsic and extrinsic parameters are calibrated using a checkerboard calibration board or a multi-plane target, thereby unifying the vision system's coordinate system to the same world coordinate system.

[0043] Based on the timestamp synchronization protocol, the time consistency error between the IMU and the vision system data is ensured to be less than 5ms.

[0044] A real-time correction system for martial arts movements in the middle school entrance examination includes:

[0045] The standard motion modeling unit is used to collect and generate a three-dimensional human motion trajectory database of standard martial arts movements for the middle school entrance examination.

[0046] The multimodal data acquisition unit is used to synchronously acquire the trainee's motion data through a wearable IMU sensor network and a multi-view high-speed camera group;

[0047] The intelligent deployment planning unit is used to automatically recommend the optimal camera layout scheme and generate a visual deployment map based on the 3D modeling and line-of-sight coverage analysis of the training site;

[0048] The data fusion and pose reconstruction unit is used to integrate sensor fusion algorithms and deep learning models to achieve spatiotemporal alignment of IMU and visual data and 3D pose reconstruction.

[0049] The real-time comparison and correction feedback unit is used to dynamically compare the reconstruction action with the standard template, output deviation reports, and provide real-time feedback through various media.

[0050] The data management and teaching support platform is used to perform user identification, data archiving, growth curve plotting, and remote supervision by teachers.

[0051] The units operate collaboratively through an edge computing server to achieve low-latency closed-loop control in a local area network environment.

[0052] A real-time correction device for martial arts movements in the middle school entrance examination, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement a real-time correction method for martial arts movements in the middle school entrance examination as described in any one of the above descriptions.

[0053] A real-time correction medium for martial arts movements in the middle school entrance examination is provided. The medium is a computer-readable storage medium, and a computer program is stored on the medium. When the computer program is executed by a processor, it implements a real-time correction method for martial arts movements in the middle school entrance examination as described in any one of the above descriptions.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] This invention breaks through the limitations of single sensors by creatively integrating AHRS / IMU inertial navigation systems with multi-view visual attitude estimation. This synergistic effect significantly improves the integrity and stability of motion capture, making it particularly suitable for complex posture changes common in martial arts movements, such as aerial maneuvers, turns, and low-posture movements. Through real-time data acquisition and analysis, deviations are promptly identified and corrective suggestions are provided during students' martial arts practice. The real-time feedback mechanism allows students to quickly see their progress, boosting their confidence and motivation, and thus stimulating their drive for self-directed learning and improvement. By using dense, label-free distributed acquisition from cameras and deep learning-based keypoint recognition, personalized corrective feedback can be provided based on each student's movement characteristics and error type, helping them to accurately improve their movements and avoiding the inability of a uniform standard to meet the diverse needs of each student. Attached Figure Description

[0056] Figure 1 This is a flowchart of a real-time correction method for martial arts movements in the middle school entrance examination according to the present invention;

[0057] Figure 2 This is a framework diagram of a real-time correction system for martial arts movements in the middle school entrance examination according to the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0059] Example 1:

[0060] Please see Figure 1 As shown, a real-time correction method for martial arts movements in the middle school entrance examination includes:

[0061] Obtain the standard movement library for martial arts in the high school entrance examination. The standard movement library contains martial arts routine movement sequences that conform to the national high school entrance examination martial arts scoring rules. Each movement sequence consists of calibrated standard three-dimensional human joint point trajectory data.

[0062] The process of constructing standard 3D human joint trajectory data includes:

[0063] When professional demonstrators perform standard martial arts movements, an inertial navigation sensor (AHRS) is used in conjunction with an attitude reference system and an inertial measurement unit (IMU) to jointly collect martial arts movement information.

[0064] IMU sensors are distributed across nine key segments of the human body, including the upper limbs, trunk, and lower limbs: left / right upper arm, left / right forearm, chest, waist, left / right thigh, and left / right lower leg. Each sensor has a built-in inertial navigation sensor (AHRS) module with a sampling frequency of no less than 200Hz.

[0065] A pose estimation algorithm is used to identify and extract the joint information of the human body from martial arts action videos, and then fuse it with the collected martial arts action information to form a complete motion trajectory.

[0066] The fused data is modeled using 3D modeling software to generate a human body model and 3D trajectory that conform to the national middle school entrance examination martial arts scoring rules, namely standard 3D human joint point trajectory data.

[0067] This invention is the first to systematically establish a standardized three-dimensional movement template database that conforms to the national middle school entrance examination martial arts scoring rules. This database not only records the standard movement trajectories of professional athletes, but also annotates key scoring points for each step (such as "the knee of the front leg in a bow stance does not go past the toes" and "the force of the punch reaches the fist surface") through a joint annotation method using the national middle school entrance examination martial arts scoring rules and an intelligent human body model. This database becomes the core basis for subsequent movement comparison and scoring reasoning, realizing a fundamental shift from "experience-based judgment" to "data-driven" approaches.

[0068] During training, a camera array consisting of multiple synchronous high-speed cameras deployed on the training ground is used to collect multi-view two-dimensional image sequences of multiple students performing the same martial arts routine in real time. The physical position, shooting angle and coverage of the camera array are automatically calculated by an AI-driven visualization deployment platform based on the geometric structure of the training space, and the optimal configuration scheme is recommended. A virtual installation simulation diagram is also generated for user confirmation.

[0069] To address the issues of traditional multi-camera systems relying on manual experience and prone to blind spots during installation, this invention introduces an AI-assisted deployment mechanism. The system can automatically deduce the optimal number of cameras, installation height, tilt angle, and overlapping field of view based on the actual dimensions of the training ground, lighting conditions, and student activity areas, and generate a 3D visualization simulation. This function significantly lowers the system deployment threshold, ensures maximum visual acquisition quality, and is particularly suitable for application and promotion in non-professional sports spaces such as school gymnasiums.

[0070] The coordinates of two-dimensional human joints are extracted from images from various perspectives using a deep learning pose estimation algorithm. Based on epipolar geometric constraints and triangulation principles, and combined with time synchronization signals, the two-dimensional joints from multiple perspectives are mapped to a unified three-dimensional spatial coordinate system to obtain a preliminary three-dimensional human pose.

[0071] The deep learning pose estimation algorithm adopts a dual-flow graph convolutional network architecture. The first flow processes the single-view two-dimensional joint sequence, and the second flow fuses the spatiotemporal context relationship across frames from multiple views, outputting a highly robust three-dimensional human pose prediction result. The dual-flow graph convolutional network architecture uses a standard three-dimensional human joint trajectory dataset as training samples and adopts a self-supervised contrastive learning strategy to achieve accurate localization of human joints and key bony markers in three-dimensional space.

[0072] This invention utilizes dense, label-free, distributed acquisition with cameras and deep learning-based keypoint recognition to provide personalized error correction feedback based on each student's movement characteristics and error type. This helps students accurately improve their movements and avoids the inability of a uniform standard to meet the diverse needs of each student.

[0073] By acquiring the angular velocity, acceleration, and orientation information of body segments through multiple IMU sensors worn on key parts of the human body, and employing a sensor spatiotemporal alignment strategy, the attitude data of the IMU sensors and the preliminary three-dimensional human posture are registered and fused in a preset common reference coordinate system to construct a complete, continuous, and occlusion-resistant full-body motion trajectory model.

[0074] Methods for establishing a common reference coordinate system include:

[0075] Using a known spatially distributed IMU node group fixed on a calibration rod, a spatial rigid body transformation is performed to determine the rotation matrix and translation vector from the IMU local coordinate system to the world coordinate system.

[0076] The camera's intrinsic and extrinsic parameters are calibrated using a checkerboard calibration board or a multi-plane target, thereby unifying the vision system's coordinate system to the same world coordinate system.

[0077] Based on the timestamp synchronization protocol, the time consistency error between the IMU and the vision system data is ensured to be less than 5ms;

[0078] The whole-body motion trajectory model is temporally aligned and similarity matched with the corresponding standard three-dimensional human joint trajectory data in the standard motion library. The motion difference is calculated by dynamic time warping algorithm, and multi-dimensional refined error analysis is performed on key bony landmarks.

[0079] Based on the preset national middle school entrance examination martial arts scoring rules engine, it automatically generates real-time correction instructions that include the type of movement error, the time of occurrence, the degree of deviation, and improvement suggestions. It also provides students with multi-channel interactive feedback through AR augmented reality interface, voice broadcast, or haptic feedback device.

[0080] This method not only serves for immediate correction but can also be extended to the management of the entire teaching process. Each student's movement data is tracked long-term, forming a personal movement evolution map. By automatically identifying common error patterns (such as "unstable starting posture" and "disordered ending rhythm"), customized training task packages are generated and pushed to a mobile app for after-class practice. Teachers can view the overall class mastery, a heatmap of weak areas, and a performance prediction model through the backend.

[0081] Example 2:

[0082] A real-time correction system for martial arts movements in the middle school entrance examination includes:

[0083] The standard motion modeling unit is used to collect and generate a three-dimensional human motion trajectory database of standard martial arts movements for the middle school entrance examination.

[0084] The multimodal data acquisition unit is used to synchronously acquire the trainee's motion data through a wearable IMU sensor network and a multi-view high-speed camera group;

[0085] The intelligent deployment planning unit is used to automatically recommend the optimal camera layout scheme and generate a visual deployment map based on the 3D modeling and line-of-sight coverage analysis of the training site;

[0086] The data fusion and pose reconstruction unit is used to integrate sensor fusion algorithms and deep learning models to achieve spatiotemporal alignment of IMU and visual data and 3D pose reconstruction.

[0087] The real-time comparison and correction feedback unit is used to dynamically compare the reconstruction action with the standard template, output deviation reports, and provide real-time feedback through various media.

[0088] The data management and teaching support platform is used to perform user identification, data archiving, growth curve plotting, and remote supervision by teachers.

[0089] Each unit operates collaboratively through an edge computing server, enabling low-latency closed-loop control in a local area network environment.

[0090] Application example: Martial arts teachers instructing students on martial arts movements for the high school entrance examination.

[0091] I. Scene Background

[0092] In a middle school physical education class, Mr. Zhang, a martial arts instructor, is using a real-time correction system for martial arts movements required for the high school entrance examination to guide students in their training. This system integrates multi-view high-speed cameras, an IMU sensor network, deep learning algorithms, and AR augmented reality feedback technology, aiming to correct students' martial arts movements accurately and in real time, thereby improving training effectiveness.

[0093] II. Application Steps

[0094] 1. System Preparation and Configuration

[0095] Equipment setup: Teacher Zhang set up multiple synchronous high-speed cameras around the training area, forming a camera array to cover all student activity areas.

[0096] Meanwhile, the students put on sports equipment equipped with IMU sensors, which are distributed in nine key segments of the students' upper limbs, torso, and lower limbs.

[0097] Intelligent deployment: The system uses an AI-driven visualization deployment platform to automatically calculate and recommend the optimal camera layout scheme based on the geometry of the training site, and generates a virtual installation simulation diagram for Teacher Zhang to confirm.

[0098] 2. Establishment of a standard motion library

[0099] In the preliminary preparation stage, the system has collected standard martial arts movements by professional demonstrators and standard three-dimensional human joint point trajectory data, constructing a standard movement library that conforms to the national junior high school entrance examination martial arts scoring rules. This data serves as the benchmark for subsequent movement comparisons.

[0100] 3. Real-time data acquisition

[0101] Multi-view image acquisition: Students begin to perform the same martial arts routine, and multiple high-speed cameras acquire multi-view two-dimensional image sequences in real time.

[0102] IMU Data Acquisition: Simultaneously, the IMU sensor network acquires real-time information on the student's body segment angular velocity, acceleration, and orientation.

[0103] 4. Data Processing and Pose Reconstruction

[0104] Two-dimensional joint point extraction: The system uses a deep learning pose estimation algorithm to extract the coordinates of two-dimensional joint points of the human body from images from various perspectives.

[0105] 3D pose reconstruction: Based on epipolar geometric constraints and triangulation principles, combined with time synchronization signals, multi-view 2D joint points are mapped to a unified 3D spatial coordinate system to obtain preliminary 3D human pose.

[0106] Subsequently, a sensor spatiotemporal alignment strategy was adopted to register and fuse the attitude data of the IMU sensor with the preliminary three-dimensional human posture in a preset common reference coordinate system, thereby constructing a complete, continuous, and occlusion-resistant full-body motion trajectory model.

[0107] 5. Motion comparison and correction feedback

[0108] Motion difference calculation: The system performs temporal alignment and similarity matching between the whole body motion trajectory model and the corresponding data in the standard motion library, calculates the motion difference through dynamic time warping algorithm, and performs multi-dimensional refined error analysis for key bony landmarks.

[0109] Real-time correction instruction generation: Based on the preset national middle school entrance examination martial arts scoring rules engine, the system automatically generates real-time correction instructions that include the type of movement error, the time of occurrence, the degree of deviation, and improvement suggestions.

[0110] Multi-channel feedback: Correction instructions are conveyed to students in real time through AR augmented reality interface, voice broadcast or haptic feedback device.

[0111] For example, when a student performs a "bow stance punch" movement, if the front knee goes past the toes, the AR interface will immediately mark the incorrect position in the student's field of vision and provide a voice prompt: "The front knee has crossed the line, please adjust."

[0112] 6. Personalized guidance and continuous improvement

[0113] Personalized correction: The system provides personalized correction feedback based on each student's movement characteristics and error type, helping students to accurately improve their movements.

[0114] Growth curve mapping: Long-term tracking of each student's movement data to form an individual movement evolution map, automatically identify common error patterns, and generate customized training task packages.

[0115] Teacher Zhang can adjust her teaching strategies based on this data and implement more precise, personalized guidance.

[0116] Remote supervision by teachers: Teacher Zhang uses the data management and teaching support platform to view the overall mastery of the class, the heat map of weak points, and the performance prediction model, realizing remote supervision and precise physical education teaching reform based on "learning-oriented teaching".

[0117] III. Application Effects

[0118] Through the application of this system, Teacher Zhang can provide real-time and precise corrections to students' martial arts movements, significantly improving training effectiveness. Students also expressed that the real-time feedback mechanism allows them to quickly see progress, enhancing their confidence and enthusiasm, and further stimulating their motivation for self-directed learning and improvement.

[0119] Example 3:

[0120] A real-time correction device for martial arts movements in the middle school entrance examination, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement a real-time correction method for martial arts movements in the middle school entrance examination as described above.

[0121] A real-time correction medium for martial arts movements in the middle school entrance examination is disclosed. This medium is a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of any of the above-described embodiments of the real-time correction method for martial arts movements in the middle school entrance examination, achieving the same technical effect. To avoid repetition, further details are omitted here. The computer-readable storage medium may include, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0122] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0123] The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to general designs. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0124] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0125] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for real-time correction of martial arts movements in the middle school entrance examination, characterized in that, include: Obtain a standard movement library for martial arts in the high school entrance examination. The standard movement library contains martial arts routine movement sequences that conform to the national high school entrance examination martial arts scoring rules. Each movement sequence consists of calibrated standard three-dimensional human joint point trajectory data. During the training process, a camera array consisting of multiple synchronous high-speed cameras deployed on the training ground is used to collect multi-view two-dimensional image sequences of multiple students performing the same martial arts routine in real time. At the same time, the angular velocity, acceleration and orientation information of body segments are obtained through multiple IMU sensors worn on key parts of the human body. The coordinates of two-dimensional human joints are extracted from images from various perspectives using a deep learning pose estimation algorithm. Based on epipolar geometric constraints and triangulation principles, and combined with time synchronization signals, the two-dimensional joints from multiple perspectives are mapped to a unified three-dimensional spatial coordinate system to obtain a preliminary three-dimensional human pose. Polar geometric constraint formula: In the formula, x1 and x2 are the corresponding homogeneous coordinates of two-dimensional joints, and F is the basic matrix; Trigonometric measurement formula: In the formula, P1 and P2 are the projection matrices of each camera, and X is the coordinate of the three-dimensional joint point; A sensor spatiotemporal alignment strategy is adopted to register and fuse the attitude data of the IMU sensor with the preliminary three-dimensional human posture in a preset common reference coordinate system, thereby constructing a complete, continuous, and occlusion-resistant full-body motion trajectory model. The whole-body motion trajectory model is time-aligned and similarity-matched with the corresponding standard three-dimensional human joint trajectory data in the standard motion library. The motion difference is calculated by dynamic time warping algorithm, and multi-dimensional refined error analysis is performed on key bony landmarks. The formula for weighted Euclidean distance at key points: In the formula, N is the number of joints, ω k The weighting is based on the national middle school entrance examination martial arts scoring rules, X k It is the three-dimensional coordinate of the k-th joint of the standard movement, Y k These are the three-dimensional coordinates of the k-th joint of the student's movement; Angle error formula: In the formula, v1 and v2 are vectors between joints, used to represent the direction between two joints, and θ k It is the angle between these two vectors, used to represent the joint rotation error; Dynamic Time Warping Algorithm Formula: D(a,b)=||X a -Y b ||+min{D(a-1,b),D(a,b-1),D(a-1,b-1)} In the formula, X a and Y b These are the 3D joint vectors of the standard action and the student action in the same frame, respectively, and D(a,b) is the cumulative minimum distance; Based on the preset national middle school entrance examination martial arts scoring rules engine, it automatically generates real-time correction instructions that include the type of movement error, the time of occurrence, the degree of deviation, and improvement suggestions. It also provides students with multi-channel interactive feedback through AR augmented reality interface, voice broadcast, or haptic feedback device.

2. The method for real-time correction of martial arts movements in the middle school entrance examination according to claim 1, characterized in that: The process of constructing the standard three-dimensional human joint trajectory data includes: When professional demonstrators perform standard martial arts movements, an inertial navigation sensor (AHRS) is used in conjunction with an attitude reference system and an inertial measurement unit (IMU) to jointly collect martial arts movement information. A pose estimation algorithm is used to identify and extract the joint information of the human body from martial arts action videos, and then fuse it with the collected martial arts action information to form a complete motion trajectory. The fused data is modeled using 3D modeling software to generate a human body model and 3D trajectory that conform to the national middle school entrance examination martial arts scoring rules, namely the standard 3D human joint point trajectory data.

3. The method for real-time correction of martial arts movements in the middle school entrance examination according to claim 2, characterized in that: The IMU sensors are distributed across nine key segments of the human body, including the upper limbs, trunk, and lower limbs: left / right upper arm, left / right forearm, chest, waist, left / right thigh, and left / right calf. Each sensor has a built-in inertial navigation sensor (AHRS) module with a sampling frequency of no less than 200Hz.

4. The method for real-time correction of martial arts movements in the middle school entrance examination according to claim 3, characterized in that: The physical location, shooting angle, and coverage of the camera array are automatically calculated by the AI-driven visualization deployment platform based on the training space geometry, which recommends the optimal configuration scheme and generates a virtual installation simulation diagram for user confirmation.

5. The method for real-time correction of martial arts movements in the middle school entrance examination according to claim 4, characterized in that: The deep learning pose estimation algorithm employs a two-stream graph convolutional network architecture, and the graph convolution formula is as follows: In the formula, N(i) represents the nodes' neighbors, and H... j (l) Let c be the l-th layer feature of the j-th key node's neighbors. ij W is the normalization coefficient. (l) Let σ(·) be the learnable weight matrix, and let σ(·) be the activation function. The first stream processes single-view 2D joint point sequences, while the second stream fuses multi-view cross-frame spatiotemporal context relationships to output highly robust 3D human pose prediction results. The dual-stream graph convolutional network architecture uses the standard 3D human joint point trajectory dataset as training samples and employs a self-supervised contrastive learning strategy to achieve accurate localization of human joint points and key bony markers in 3D space.

6. The method for real-time correction of martial arts movements in the middle school entrance examination according to claim 1, characterized in that: The method for establishing the common reference coordinate system includes: Using a known spatially distributed IMU node group fixed on a calibration rod, a spatial rigid body transformation is performed to determine the rotation matrix and translation vector from the IMU local coordinate system to the world coordinate system. The camera's intrinsic and extrinsic parameters are calibrated using a checkerboard calibration board or a multi-plane target, thereby unifying the vision system's coordinate system to the same world coordinate system. Based on the timestamp synchronization protocol, the time consistency error between the IMU and the vision system data is ensured to be less than 5ms.

7. A real-time correction system for martial arts movements in the middle school entrance examination, characterized in that, include: The standard motion modeling unit is used to collect and generate a three-dimensional human motion trajectory database of standard martial arts movements for the middle school entrance examination. The multimodal data acquisition unit is used to synchronously acquire the trainee's motion data through a wearable IMU sensor network and a multi-view high-speed camera group; The intelligent deployment planning unit is used to automatically recommend the optimal camera layout scheme and generate a visual deployment map based on the 3D modeling and line-of-sight coverage analysis of the training site; The data fusion and pose reconstruction unit is used to integrate sensor fusion algorithms and deep learning models to achieve spatiotemporal alignment of IMU and visual data and 3D pose reconstruction. The real-time comparison and correction feedback unit is used to dynamically compare the reconstruction action with the standard template, output deviation reports, and provide real-time feedback through various media. The data management and teaching support platform is used to perform user identification, data archiving, growth curve plotting, and remote supervision by teachers. The units operate collaboratively through an edge computing server to achieve low-latency closed-loop control in a local area network environment.

8. A real-time correction device for martial arts movements in the middle school entrance examination, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement a real-time correction method for martial arts movements in the middle school entrance examination as described in any one of claims 1 to 6.

9. A real-time correction medium for martial arts movements in the middle school entrance examination, characterized in that, The medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements a real-time correction method for martial arts movements in the middle school entrance examination as described in any one of claims 1 to 6.

Images