Intelligent physical training system and method thereof

Abstract

The invention discloses an intelligent physical training system and method, and belongs to the technical field of physical training, and the system comprises a laser point cloud collection module which is used for transmitting a 1550 nm wavelength single echo and generating high-density point cloud data; the three-dimensional modeling module is used for performing multi-dimensional maximum pooling operation on the point cloud data based on PointNet to construct a three-dimensional human body model; the attitude estimation module is used for detecting key points of multiple persons in real time by adopting an RMPE technology and comparing the key points with an SMPL parameterized champion database; and the motion analysis module is used for segmenting a multi-person view based on Mask R-CNN and calculating joint angle deviation in combination with an OpenPose algorithm. According to the intelligent physical training system and the intelligent physical training method, by creating an intelligent physical training classroom space, comprehensive monitoring and analysis of personalized data of trainees can be achieved, accurate and dynamic training management and effect evaluation can be provided, the training effect and trainee experience are greatly improved, meanwhile, the training process is more scientific, efficient and safe, and the training experience is improved. And students are helped to better realize the fitness goal.

Description

Technical Field

[0001] This invention belongs to the field of sports training technology, specifically relating to an intelligent sports training system and method. Background Technology

[0002] The intelligent sports training system is a digital training platform based on sensor, artificial intelligence, and big data technologies. It collects real-time data on movement posture, physiological indicators, and training, and combines this with AI algorithms to analyze movement technique and physical condition, providing personalized training plans and immediate feedback. The system can simulate competitive scenarios, assess sports injury risks, help athletes optimize technical details and adjust training intensity, and assist coaches in scientifically developing periodic training plans. It is widely used in professional sports, school sports, and mass fitness, enabling visualization and precision in the training process, improving athletic performance, and reducing the probability of injury.

[0003] Existing sports training systems suffer from the following shortcomings: insufficient accuracy: traditional cameras or sensors cannot accurately construct three-dimensional human body models, resulting in large errors in motion analysis; poor real-time performance: data processing efficiency is low and feedback delay is significant in multi-person scenarios; lack of standardized references: there is a lack of databases based on the motion parameters of professional athletes, making it difficult to provide scientific training suggestions. To solve the above problems, an intelligent sports training system and its method are proposed. Summary of the Invention

[0004] In response to one or more of the above-mentioned defects or improvement needs of the existing technology, the present invention provides an intelligent sports training system and method, which utilizes laser point cloud technology, deep learning models, and multimodal data fusion.

[0005] To achieve the above objectives, the present invention provides an intelligent sports training system, including a laser point cloud acquisition module for emitting a 1550nm wavelength single echo and generating high-density point cloud data; The 3D modeling module uses PointNet to perform multi-dimensional max pooling operations on point cloud data to construct a 3D human body model. The attitude estimation module uses RMPE technology to detect key points of multiple people in real time and compares them with the SMPL parametric champion database. The motion analysis module uses Mask R-CNN to segment multi-person views and combines the OpenPose algorithm to calculate joint angle deviations. The real-time feedback module processes data through FreeRTOS multi-threading and outputs action correction instructions to the interactive terminal. The data monitoring module includes a laser interactive panel, floor pressure sensor, smart wearable device and cloud platform, used for multi-source data acquisition and transmission.

[0006] Furthermore, the SMPL parametric champion database contains parametric data of several world champions and national champions, covering body shape parameters, posture parameters, and shape parameters.

[0007] Furthermore, the point cloud data generated by the laser point cloud acquisition module has a density of 90 million points, and the wavelength is adjusted to 1550nm to meet human safety standards.

[0008] Furthermore, the motion analysis module segments the multi-person view into independent individuals using Mask R-CNN, and performs 3D pose analysis using the OpenPose algorithm to output motion deviation values.

[0009] Furthermore, the real-time feedback module sets a conditional function to dynamically adjust the false detection parameter by increasing the threshold from small to large until the erroneous feedback disappears.

[0010] On the other hand, the present invention also provides an intelligent sports training method, comprising the following steps: S1: The teacher operates the interactive panel and selects the corresponding sports scene mode. The student wears a smart bracelet and stands in front of the interactive panel. The LiDAR generates point cloud data, the smart camera captures the movement trajectory, the bracelet collects heart rate and step frequency, and PointNet builds a 3D model. S2: Use RMPE technology to extract key points and compare them with parametric actions in the champion database; S3: Multi-person view segmentation using Mask R-CNN, combined with OpenPose algorithm to analyze joint angle deviation; S4: Based on FreeRTOS, real-time feedback of motion correction commands is sent to the interactive panel, and personalized training reports are generated through the cloud platform.

[0011] Furthermore, the personalized training report includes exercise intensity assessment, posture deviation analysis, and health risk warning.

[0012] In summary, the beneficial effects of the above-described technical solutions conceived by this invention compared with the prior art include: The intelligent sports training system of the present invention significantly improves the accuracy, efficiency and practicality of training analysis by integrating laser point cloud technology, deep learning models (such as PointNet, RMPE, Mask R-CNN) and real-time operating system (FreeRTOS) and combining champion action database. The intelligent sports training method of this invention, by creating a smart sports training classroom space, can not only achieve comprehensive monitoring and analysis of students' personalized data, but also provide accurate and dynamic training management and effect evaluation, greatly improving training effectiveness and student experience. At the same time, it makes the training process more scientific, efficient and safe, helping students better achieve their fitness goals. Attached Figure Description

[0013] Figure 1 This is a flowchart of an embodiment of the present invention; Figure 2 This is a schematic diagram of PointNet point cloud data processing according to the present invention; Figure 3 This is a schematic diagram illustrating the RMPE attitude estimation principle of the present invention. Figure 4 This is a hardware deployment diagram of the present invention; Figure 5 This is a schematic diagram of the camera capture trajectory of the present invention. Detailed Implementation

[0014] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] See also Figure 1-5 The present invention provides an intelligent sports training system, including a laser point cloud acquisition module for emitting a 1550nm wavelength single echo and generating high-density point cloud data; The 3D modeling module uses PointNet to perform multi-dimensional max pooling operations on point cloud data to construct a 3D human body model. The attitude estimation module uses RMPE technology to detect key points of multiple people in real time and compares them with the SMPL parametric champion database. The motion analysis module uses Mask R-CNN to segment multi-person views and combines the OpenPose algorithm to calculate joint angle deviations. The real-time feedback module processes data through FreeRTOS multi-threading and outputs action correction instructions to the interactive terminal. The data monitoring module includes a laser interactive panel, floor pressure sensor, smart wearable device and cloud platform, used for multi-source data acquisition and transmission.

[0016] In this embodiment, PointNet point cloud data processing is a groundbreaking deep learning network in the prior art, designed specifically for processing 3D point cloud data. By performing max pooling operations on each dimension of the features, PointNet can acquire the global features of the point cloud. In addition, PointNet deep learning enables the system to achieve the optimal transformation matrix for point cloud alignment, solving the problem of point cloud rotation and translation. Based on PointNet point cloud data processing, max pooling operations on five dimensions are performed on the collected messy and disordered point cloud to extract all point cloud features. The messy point cloud is then connected in a reasonable way to generate a fixed-size feature vector for subsequent classification, detection or other tasks, providing a complete model for subsequent calculations. RMPE (Real-time Multi-Person Pose Estimation) is a technology for real-time detection and analysis of multiple person poses in images or videos. It can identify and track key points on multiple people's bodies, such as heads, wrists, shoulders, and knees. It is widely used in fields such as behavior recognition, health monitoring, augmented reality, and human-computer interaction. RMPE technology can quickly process image or video streams to achieve real-time pose estimation. Unlike methods that only estimate the pose of a single person, RMPE can simultaneously detect and estimate the poses of multiple people using images and videos captured by cameras. Based on RMPE pose estimation technology, the accuracy of motion recognition algorithms can be improved and the data can be entered into the system. This allows for the simultaneous identification of key points on multiple 3D human bodies and the handling of inaccurate human detection boxes and redundant detections, thereby improving the accuracy of pose estimation. It can accurately identify the movements and poses of primary and secondary school students in multi-person scenarios.

[0017] Specifically, the SMPL Parametric Champion Database contains parametric data of several world and national champions, covering body shape parameters, posture parameters, and shape parameters.

[0018] In this embodiment, in order to accurately compare the collected data with the data in the database, the SMPL model based on 3D human body modeling and motion capture will use data parameters to represent different shapes and postures of the human body. These parameters include body shape parameters, posture parameters, and shape parameters. In the construction of the champion model database, it is possible to simulate complex human movements, including the bulges and depressions of muscles during limb movement, avoiding surface distortion during movement. Based on the SMPL model, the optimal sports indicators of multiple world champions and national champions are constructed into a champion model database through parameters. The sports data in the champion model database has more standardized and safer movement references, and the collected data can be quickly identified.

[0019] Specifically, the laser point cloud acquisition module generates point cloud data with a density of 90 million points, and the wavelength is adjusted to 1550nm to meet human safety standards.

[0020] In this embodiment, a carbon dioxide laser is used to emit a single echo based on high-precision human body scanning technology. Relying on the characteristics of the wave, the system generates 90 million point cloud data distributed in various locations by receiving the different reflection times of the wave. These data are the contour curves of the human body, which meticulously depict the concave and convex parts of the human body. In order to minimize the harm of the laser detector to the eyes and other parts of primary and secondary school students, the wavelength of the continuous wave is adjusted to a uniform 1550nm in order to meet the standard of being harmless to primary and secondary school students.

[0021] Specifically, the motion analysis module uses Mask R-CNN to segment the multi-person view into individual entities, combines it with the OpenPose algorithm to perform 3D pose analysis, and outputs motion deviation values.

[0022] In this embodiment, Mask R-CNN is a deep learning architecture used for object detection and instance segmentation. It uses a feature pyramid network to construct high-resolution representations, facilitating the detection of targets of different sizes. It also uses a Representation Network (RPN) to quickly generate candidate target regions. Mask R-CNN not only identifies targets in images but also performs pixel-level segmentation of each instance, outputting a target mask. To provide real-time feedback and timely action improvement, training and testing code is implemented based on the Mask R-CNN benchmark provided by Facebook AI Research. This combines the Mask R-CNN learning architecture with AI, enabling the system to use Mask R-CNN for detection and instance segmentation of different limbs of the user. This allows the system to segment multi-person images in the same view, decomposing complex multi-person data into individual actions through segmentation, reducing the computational load of each action and significantly improving computational speed and recognition accuracy. R-CNN performs pixel-level segmentation of the human body model composed of a unified view, enabling separate calculation of each limb, reducing data complexity, significantly reducing computational load, and improving computational speed and feedback speed. It uses motion analysis algorithms to accurately analyze each joint point in the collected human images, and compares and overlays champion data from the champion model database with the movement postures of primary and secondary school students. At the same time, it combines OpenPose technology to implement a human posture analysis algorithm, which uses deep learning algorithms to detect and track human key points in images or videos, including key points of the body, face, hands, and even feet. It relies on motion recognition algorithms to identify key joint points of the human body and accurately analyze the human body model's movement posture. The motion recognition algorithm does not rely on specific detection instruments, but uses standard recognition devices as input devices to capture images or videos. It provides a computational benchmark for OpenPose based on the joint points provided by RMPE pose estimation technology. It compares and overlays the detected human model movements with parameters in the champion database, enabling it to perform 3D pose analysis based on 3D pose estimation. This solves the problem of accurately identifying the movement trajectory of each user's limbs, accurately analyzes each user's movement process and motion data, and provides scientific guidance for physical education teaching.

[0023] Specifically, the real-time feedback module sets a conditional function to dynamically adjust the false detection parameters by increasing the threshold from small to large until the erroneous feedback disappears; In this embodiment, the real-time operating system is a specially designed operating system that can guarantee task completion within strict time constraints. RTOS is typically used in control and data acquisition systems, which require high reliability and determinism. FreeRTOS supports multitasking, allowing multiple tasks (threads) to run concurrently. It can simultaneously calculate data from multiple limbs and multiple human bodies, achieving ideal computational conditions with low volume and high speed. It also has low latency. The system can perform multi-threaded task processing through its code and provide real-time feedback of the calculation results to the screen. By importing the video into the database, a virtual environment is created, and the FreeRTOS real-time operating system code is used to calculate and analyze the data obtained from the motion analysis algorithm and display the results. When false detections cause errors in the video's estimation of human posture, the threshold of the monitoring parameters is increased sequentially by setting the numbers in ascending order. This process is repeated until false detections cease, the display error disappears, and the erroneous feedback is reduced until completely correct feedback is achieved. Finally, the feedback will also be displayed on the screen as the display error disappears. Finally, a conditional function is added to provide corresponding teaching guidance based on the completeness of the motion fit.

[0024] Hardware deployment: Install lidar, high-definition cameras, floor pressure sensors, and laser interactive panels in the classroom; Sensor networks and smart cameras: High-precision sensor networks and smart cameras are installed in the smart physical training classroom, covering every corner. These sensors and cameras can capture trainees' movement postures, training movements and environmental parameters in real time, ensuring the comprehensiveness and accuracy of the data.

[0025] Floor pressure sensors: Pressure sensors installed on the classroom floor can accurately monitor students' foot pressure and center of gravity changes. This data allows for analysis of students' movement trajectories, gait, and force distribution, helping to optimize posture and prevent injuries. Utilizing piezoelectric materials or strain gauges, the floor pressure sensors are sensitive to pressure changes. Data is transmitted wirelessly in real-time to a central data processing system for rapid analysis and feedback. By combining this with biomechanical models, a detailed gait analysis report can be provided, identifying issues such as uneven strides and unstable center of gravity, helping students correct poor exercise habits. Motion capture cameras: Smart cameras installed around the classroom can capture students' movement trajectories and postures from all angles. These cameras are equipped with advanced image recognition technology and deep learning algorithms, which can analyze students' movement details in real time and provide accurate motion data. Smart cameras not only record students' movements but also generate 3D motion models through data processing, providing detailed motion breakdown and posture assessment. The cameras integrate with other sensor systems in the classroom to achieve synchronous data acquisition and comprehensive analysis, improving the comprehensiveness and accuracy of monitoring. Laser interactive panel: Installed in the classroom, it displays the trainees' training status and data in real time, facilitating on-site guidance and adjustments by coaches. The large-screen display system is seamlessly connected to the sensor network and cloud platform, updating the displayed content in real time. The display screen adopts a high-definition resolution and large size design to ensure clear visibility of data and images. Coaches can quickly retrieve trainees' training data through touch operation or voice control for detailed explanation and guidance. On the other hand, the present invention also provides an intelligent sports training method, comprising the following steps: S1: The teacher operates the interactive panel and selects the corresponding sports scene mode. The student wears a smart bracelet and stands in front of the interactive panel. The LiDAR generates point cloud data, the smart camera captures the movement trajectory, the bracelet collects heart rate and step frequency, and PointNet builds a 3D model. S2: Use RMPE technology to extract key points and compare them with parametric actions in the champion database; S3: Multi-person view segmentation using Mask R-CNN, combined with OpenPose algorithm to analyze joint angle deviation; S4: Based on FreeRTOS, real-time feedback of motion correction commands is sent to the interactive panel, and personalized training reports are generated through the cloud platform.

[0026] In this embodiment, each trainee can wear smart wearable devices, such as heart rate monitors and fitness trackers. These devices can collect trainees' physiological indicators in real time, such as heart rate, blood oxygen saturation, body temperature, and exercise data, such as steps, calories burned, and exercise intensity. This helps trainees understand their exercise status and energy expenditure, providing data support. Through intelligent algorithm analysis, personalized exercise suggestions can be provided, such as cadence adjustment and running posture optimization, to improve training effectiveness. The smart wearable devices transmit data to the data processing center in real time via wireless connection. Through the accumulation and analysis of trainees' long-term training data, the platform can generate personalized training reports and suggestions, forming a long-term exercise record.

[0027] Specifically, personalized training reports include exercise intensity assessment, postural deviation analysis, and health risk warnings.

[0028] In this embodiment, the aforementioned report can be fed back to parents, teachers, and coaches via a mobile application. Personalized training data and improvement suggestions can be viewed on mobile phones. Coaches can then develop more targeted training plans based on each student's situation. At the same time, the classroom display screen will also show the students' training status and data in real time, facilitating on-site guidance and adjustments by coaches. Through this multi-person data monitoring and cumulative evaluation, not only can comprehensive monitoring and analysis of students' personalized data be achieved, but also accurate and dynamic training management and effect evaluation can be provided, significantly improving training effectiveness and student experience. This makes the training process more scientific, efficient, and safe, helping students better achieve their fitness goals.

[0029] 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 alterations 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. An intelligent sports training system, characterized in that, Includes the following modules: The laser point cloud acquisition module is used to emit a 1550nm wavelength single echo and generate high-density point cloud data; The 3D modeling module uses PointNet to perform multi-dimensional max pooling operations on point cloud data to construct a 3D human body model. The attitude estimation module uses RMPE technology to detect key points of multiple people in real time and compares them with the SMPL parametric champion database. The motion analysis module uses Mask R-CNN to segment multi-person views and combines the OpenPose algorithm to calculate joint angle deviations. The real-time feedback module processes data through FreeRTOS multi-threading and outputs action correction instructions to the interactive terminal. The data monitoring module includes a laser interactive panel, floor pressure sensor, smart wearable device and cloud platform, used for multi-source data acquisition and transmission.

2. The intelligent sports training system according to claim 1, characterized in that, The SMPL Parametric Champion Database contains parametric data of several world and national champions, covering body shape parameters, posture parameters, and shape parameters.

3. The intelligent sports training system according to claim 1, characterized in that, The laser point cloud acquisition module generates point cloud data with a density of 90 million points, and the wavelength is adjusted to 1550nm to meet human safety standards.

4. The intelligent sports training system according to claim 1, characterized in that, The motion analysis module segments the multi-person view into independent individuals using MaskR-CNN, and performs 3D pose analysis using the OpenPose algorithm to output motion deviation values.

5. The intelligent sports training system according to claim 1, characterized in that, The real-time feedback module sets a conditional function to dynamically adjust the false detection parameters by increasing the threshold from small to large until the erroneous feedback disappears.

6. An intelligent sports training method, characterized in that, Includes the following steps: S1: The teacher operates the interactive panel and selects the corresponding sports scene mode. The student wears a smart bracelet and stands in front of the interactive panel. The LiDAR generates point cloud data, the smart camera captures the movement trajectory, the bracelet collects heart rate and step frequency, and PointNet builds a 3D model. S2: Use RMPE technology to extract key points and compare them with parametric actions in the champion database; S3: Multi-person view segmentation using Mask R-CNN, combined with OpenPose algorithm to analyze joint angle deviation; S4: Based on FreeRTOS, real-time feedback of motion correction commands is sent to the interactive panel, and personalized training reports are generated through the cloud platform.

7. The method according to claim 6, characterized in that, The personalized training report includes exercise intensity assessment, posture deviation analysis, and health risk warning.

Images