Treadmill running posture analysis method and system based on motion capture

Abstract

The invention discloses a running machine running posture analysis method and system based on motion capture, and the method comprises the following steps: S1, obtaining multi-mode synchronous perception data of a user performing full-dynamic range motion on a running machine, the full-dynamic range motion comprising steady jogging, unstable sprint running and turning running; and S2, performing high-fidelity fusion and space-time calibration processing on the multi-modal synchronous sensing data, and constructing an anti-distortion user attitude space-time sequence. According to the invention, by deploying a surrounding type synchronous capture array formed by a plurality of high-frame-rate visual sensors and combining with a running board pressure sensor, multi-mode, non-blind-area and high-temporal-spatial-resolution data acquisition of unsteady-state high-speed motion such as sprint running and turning running is realized; the hardware architecture fundamentally solves the problems of motion blur, key attitude frame loss and the like caused by insufficient sampling of a traditional single-camera or low-frame-rate system, and provides a high-fidelity original data basis for subsequent analysis.

Description

Technical Field

[0001] This invention relates to the field of treadmill technology, specifically to a method and system for analyzing treadmill running posture based on motion capture. Background Technology

[0002] With the increasing health awareness of the public and the growing demand for fitness, treadmills, as convenient and efficient indoor fitness equipment, have been widely used in homes, gyms and other settings. Running posture, as a core factor affecting running efficiency, exercise effect and injury risk, has become an important direction for the intelligent development of treadmills through scientific analysis and real-time guidance. At present, motion capture-based running posture analysis systems collect posture data during the running process and combine algorithms to extract key parameters such as stride frequency, stride length, ground contact angle and center of gravity trajectory to provide users with personalized exercise suggestions. It has already made some progress in the application of running posture monitoring in basic jogging scenarios.

[0003] However, existing running posture analysis systems struggle to accurately capture rapid limb swings during sprints and instantaneous posture transitions during changes of direction when dealing with high-speed and complex movements due to low video sampling frequencies. This results in the loss or distortion of key posture data. Furthermore, traditional analysis algorithms are mostly designed for smooth, slow jogging scenarios and are not well-suited to adapting to nonlinear posture changes and complex motion coupling relationships under high-speed conditions. They cannot accurately analyze relevant motion parameters, directly impacting the accuracy of running posture analysis. This makes it difficult for the system to effectively identify non-standard running postures in high-speed and complex movements, thereby reducing the reliability of real-time feedback and failing to meet users' scientific fitness needs in diverse sports scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for analyzing running posture on a treadmill based on motion capture. By constructing a complete technical closed loop from high-speed data acquisition and intelligent fusion analysis to scenario-based evaluation and real-time feedback, it effectively overcomes the bottlenecks faced by existing technologies in highly dynamic and complex sports scenarios.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a treadmill running posture analysis method based on motion capture, comprising the following steps: Step S1: Acquire multimodal synchronous perception data of the user performing full dynamic range motion on the treadmill, wherein the full dynamic range motion includes steady-state jogging and non-steady-state sprinting and changing direction running; Step S2: Perform high-fidelity fusion and spatiotemporal calibration on the multimodal synchronous sensing data to construct a distortion-resistant user attitude spatiotemporal sequence; Step S3: The user's spatiotemporal posture sequence is analyzed and predicted in real time using an intelligent posture parsing engine based on dynamic prior constraints, in order to compensate for data loss under high-speed motion and extract high-dimensional posture features. Step S4: Based on the high-dimensional posture features, dynamically calculate a set of refined running posture evaluation parameters covering spatial, temporal, and dynamic dimensions; Step S5: Match and assess the risks of the refined running posture evaluation parameters with the preset multi-level running posture standard models adapted to different sports scenarios; Step S6: Based on the results of the risk assessment, generate and drive the multi-channel real-time feedback interface to output adaptation guidance and early warning information.

[0006] As a preferred embodiment, step S1, specifically including the acquisition of multimodal synchronous sensing data, includes: Deploy a surround motion capture array consisting of four high frame rate visual sensors, with a sampling frequency of no less than 240fps; The motion capture array is controlled to acquire multi-view raw video streams of the user's motion in full dynamic range using a synchronous triggering method. The multimodal synchronous sensing data also includes dynamic signals from the treadmill running board pressure sensor that are aligned with the video stream timestamp.

[0007] As a preferred embodiment, in step S2, the high-fidelity fusion and spatiotemporal calibration processing includes: Denoising and feature-point-based frame-level synchronization alignment are performed on the original video streams from multiple perspectives to eliminate timing errors caused by differences in device response. Based on synchronized multi-view data, a three-dimensional reconstruction algorithm is used to generate real-time temporal three-dimensional coordinates of the user's skeletal joints in the treadmill coordinate system. The time-series three-dimensional coordinates are fused with the synchronously acquired running board pressure signal to perform kinematic data correction and enhancement based on the ground reaction force phase, forming the user attitude spatiotemporal sequence.

[0008] As a preferred embodiment, in step S3, the intelligent posture parsing engine based on dynamic prior constraints is a hybrid architecture processor that integrates human motion biomechanical model and data-driven prediction model. The human movement biomechanical model is used to provide reasonable physiological range constraints for joint angles and center of mass trajectory; The data-driven prediction model is used to predict key pose frames that are blurred or lost during high-speed motion within the data sampling interval. The intelligent posture analysis engine outputs smooth, continuous, and physiologically consistent high-dimensional posture features through iterative optimization.

[0009] As a preferred embodiment, in step S4, the refined running posture evaluation parameters include: Spatial dimension parameters: stride length, ground contact angle, torso forward lean angle, and lateral sway of center of gravity; Time-related parameters: cadence, ground contact time, and airtime; Dynamic parameters: impact loading rate and left-right foot balance index estimated based on running board pressure signals and kinematic data.

[0010] As a preferred embodiment, step S5, which involves matching and risk assessment of the multi-level running posture standard model, specifically includes: The pre-built running posture standard model library contains a set of benchmark parameters and allowable fluctuation thresholds for different scenarios such as jogging, sprinting, and changing direction running. Based on the real-time identified motion scene, the corresponding benchmark parameter set is called, and the real-time deviation of each evaluation parameter is calculated; A dynamic risk assessment function based on multi-parameter weighted fusion is established. When the function value exceeds the preset warning threshold related to the motion scenario, a graded early warning signal containing the risk level and the identifier of the main abnormal parameters is generated.

[0011] As a preferred embodiment, in step S6, the information output by the multi-channel real-time feedback interface includes: Graphical posture comparison and parameter curves output through the treadmill's embedded display screen or a linked mobile device APP; Real-time voice guidance and rhythm cues output through bone conduction headphones or a speaker system; Vibration alerts synchronized with abnormal postures are output by haptic actuators integrated into smart running shoes or wearable devices.

[0012] A treadmill running posture analysis system based on motion capture, the system comprising: A high-speed multimodal data acquisition unit is used to acquire multimodal synchronous sensing data of the user's full dynamic range movement on the treadmill; The data fusion and spatiotemporal calibration unit is communicatively connected to the high-speed multimodal data acquisition unit and is used to process the multimodal synchronous sensing data and output a distortion-resistant user attitude spatiotemporal sequence. The intelligent hybrid attitude parsing unit is communicatively connected to the data fusion and spatiotemporal calibration unit. It has a built-in intelligent attitude parsing engine based on dynamic prior constraints, which is used to parse the user's attitude spatiotemporal sequence and output high-dimensional attitude features. The running posture assessment and risk analysis unit is communicatively connected to the intelligent hybrid posture analysis unit. It is used to calculate refined running posture assessment parameters based on the high-dimensional posture features and to perform risk assessment in combination with a multi-level running posture standard model. A multi-channel real-time feedback and guidance unit is connected in communication with the running posture assessment and risk analysis unit, and is used to generate and drive the output of adaptive guidance and early warning information based on the risk assessment results.

[0013] As a preferred embodiment, the high-speed multimodal data acquisition unit includes a distributed surround array consisting of four high frame rate visual sensors. The visual sensors operate synchronously at a sampling frequency of no less than 240fps and are deployed in front, behind, and on the left and right sides of the treadmill running platform, with their main optical axes converging in the typical exercise area of ​​the user in the center of the running platform.

[0014] As a preferred embodiment, the data fusion and spatiotemporal calibration unit, the intelligent hybrid posture analysis unit, and the running posture assessment and risk analysis unit are integrated and deployed in an edge computing device; The edge computing device is connected to the high-speed multimodal data acquisition unit and the multi-channel real-time feedback and guidance unit via a high-speed bus to realize real-time processing of the entire process from data acquisition to feedback output, wherein the end-to-end system latency is no more than 100 milliseconds.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention utilizes a surround synchronous capture array composed of multiple high frame rate visual sensors, combined with a running board pressure sensor, to achieve multimodal, blind-spot-free, and high spatiotemporal resolution data acquisition for non-steady-state high-speed movements such as sprinting and changing direction. This hardware architecture fundamentally solves the problems of motion blur and loss of key posture frames caused by insufficient sampling in traditional single-camera or low frame rate systems, providing a high-fidelity raw data foundation for subsequent analysis and ensuring that millisecond-level rapid limb swings and instantaneous posture changes can be clearly captured.

[0016] 2. This invention utilizes an intelligent posture analysis engine based on dynamic prior constraints and employs a hybrid architecture that integrates a human motion biomechanical model and a data-driven prediction model. The biomechanical model imposes constraints on joint range of motion and center of mass trajectory that conform to physiological laws, ensuring the rationality of the analysis results. The data-driven prediction model can learn the continuous patterns of high-speed motion and intelligently predict and complete frames that are lost or blurred within the sampling interval. The two work together through iterative optimization to effectively compensate for data distortion under high-speed motion, outputting smooth, continuous, and physiologically reliable high-dimensional posture features, greatly enhancing the ability to analyze and adapt to complex and nonlinear motion postures.

[0017] 3. This invention overcomes the limitations of analyzing only basic parameters such as stride frequency and stride length, and can dynamically calculate a refined set of parameters covering three dimensions: space, time, and dynamics. By pre-setting a multi-level standard model library of running postures that matches different scenarios such as jogging, sprinting, and changing direction running, the system can identify the movement state in real time, call the corresponding model for comparison, and establish a dynamic risk assessment function based on multi-parameter weighted fusion. This achieves a leap from simple threshold alarms to intelligent and scenario-based hierarchical risk assessment, making the running posture analysis results more scientific, accurate, and with targeted guidance.

[0018] 4. By integrating the core processing unit into the edge computing device, this invention achieves end-to-end ultra-low latency processing from data acquisition to feedback output, ensuring the real-time nature of guidance. Through real-time feedback interfaces with multiple channels of vision, hearing, and touch, it transforms abstract risk assessment results into specific information that users can easily perceive and understand instantly. This immersive interaction method can provide accurate correction guidance while the user is exercising, significantly improving the efficiency of running posture correction and training experience. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system architecture block diagram of the present invention. Detailed Implementation

[0020] 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.

[0021] Please see Figure 1 and Figure 2 As shown, this invention provides a treadmill running posture analysis method based on motion capture, including the following steps: Step S1: Acquire multimodal synchronous perception data of the user performing full dynamic range motion on the treadmill. Full dynamic range motion includes steady-state jogging and non-steady-state sprinting and change-of-direction running. Step S2: Perform high-fidelity fusion and spatiotemporal calibration on the multimodal synchronous sensing data to construct a distortion-resistant user attitude spatiotemporal sequence; Step S3: Using an intelligent attitude parsing engine based on dynamic prior constraints, the user's attitude spatiotemporal sequence is parsed and predicted in real time to compensate for data loss under high-speed motion and extract high-dimensional attitude features. Step S4: Based on high-dimensional posture features, dynamically calculate a set of refined running posture evaluation parameters covering spatial, temporal, and dynamic dimensions; Step S5: Match and assess the risks of refined running posture evaluation parameters with pre-set multi-level running posture standard models adapted to different sports scenarios; Step S6: Based on the risk assessment results, generate and drive the multi-channel real-time feedback interface to output adaptation guidance and early warning information.

[0022] In step S1, acquiring multimodal synchronous sensing data specifically includes: Deploy a surround motion capture array consisting of four high frame rate visual sensors with a sampling frequency of no less than 240fps. The motion capture array is controlled to capture raw video streams from multiple perspectives of the user's motion in a synchronous triggering manner; The multimodal synchronous sensing data also includes kinetic signals from the treadmill running board pressure sensor, which are aligned with the video stream timestamp.

[0023] In this technical solution, by deploying a surround synchronous capture array composed of multiple high frame rate visual sensors and combining it with a running board pressure sensor, multimodal, blind-spot-free, and high spatiotemporal resolution data acquisition of non-steady-state high-speed movements such as sprinting and changing direction is achieved. This hardware architecture fundamentally solves the problems of motion blur and loss of key posture frames caused by insufficient sampling in traditional single-camera or low frame rate systems, providing a high-fidelity raw data foundation for subsequent analysis and ensuring that millisecond-level rapid limb swings and instantaneous posture changes can be clearly captured.

[0024] In step S2, the high-fidelity fusion and spatiotemporal calibration processing includes: Denoising and feature-point-based frame-level synchronization alignment are performed on the original video streams from multiple perspectives to eliminate timing errors caused by differences in device response. Based on synchronized multi-view data, a three-dimensional reconstruction algorithm is used to generate real-time temporal three-dimensional coordinates of the user's skeletal joints in the treadmill coordinate system. By fusing the temporal three-dimensional coordinates with the synchronously acquired running board pressure signal, the kinematic data is corrected and enhanced based on the phase of the ground reaction force, forming a spatiotemporal sequence of user posture.

[0025] In step S3, the intelligent posture parsing engine based on dynamic prior constraints is a hybrid architecture processor that integrates human motion biomechanical model and data-driven prediction model. Human biomechanical models are used to provide reasonable physiological constraints for joint angles and center-of-mass trajectories. Data-driven prediction models are used to predict key pose frames that are blurred or lost during high-speed motion within a data sampling interval. The intelligent posture analysis engine outputs smooth, continuous, and physiologically consistent high-dimensional posture features through iterative optimization.

[0026] This technical solution employs an intelligent posture analysis engine based on dynamic prior constraints, and utilizes a hybrid architecture that integrates a human motion biomechanical model and a data-driven prediction model. The biomechanical model imposes physiologically consistent constraints on joint range of motion and center of mass trajectory, ensuring the rationality of the analysis results. The data-driven prediction model learns continuous patterns of high-speed motion and intelligently predicts and completes lost or blurred frames within the sampling interval. Through iterative optimization and collaborative work, the two effectively compensate for data distortion under high-speed motion, outputting smooth, continuous, and physiologically reliable high-dimensional posture features, greatly enhancing the analytical adaptability to complex and nonlinear motion postures. In step S4, the refined running posture evaluation parameters include: Spatial dimension parameters: stride length, ground contact angle, torso forward lean angle, and lateral sway of center of gravity; Time-related parameters: cadence, ground contact time, and airtime; Dynamic parameters: impact loading rate and left-right foot balance index estimated based on running board pressure signals and kinematic data.

[0027] In step S5, the matching and risk assessment of the multi-level running posture standard model specifically includes: The pre-built running posture standard model library contains a set of benchmark parameters and allowable fluctuation thresholds for different scenarios such as jogging, sprinting, and changing direction running. Based on the real-time identified motion scene, the corresponding benchmark parameter set is called, and the real-time deviation of each evaluation parameter is calculated; A dynamic risk assessment function based on multi-parameter weighted fusion is established. When the function value exceeds the preset warning threshold related to the motion scenario, a graded early warning signal containing the risk level and the identifier of the main abnormal parameters is generated.

[0028] This technical solution overcomes the limitations of analyzing only basic parameters such as stride frequency and stride length, and can dynamically calculate a refined set of parameters covering three dimensions: space, time, and dynamics. It has a pre-built multi-level running posture standard model library that matches different scenarios such as jogging, sprinting, and changing direction running, and can identify the movement state in real time, calling the corresponding model for comparison. By establishing a dynamic risk assessment function based on multi-parameter weighted fusion, it achieves a leap from simple threshold alarms to intelligent, scenario-based hierarchical risk assessment, making the running posture analysis results more scientific, accurate, and with targeted guidance.

[0029] In step S6, the information output by the multi-channel real-time feedback interface includes: Graphical posture comparison and parameter curves output through the treadmill's embedded display screen or a linked mobile device APP; Real-time voice guidance and rhythm cues output through bone conduction headphones or a speaker system; Vibration alerts synchronized with abnormal postures are output by haptic actuators integrated into smart running shoes or wearable devices.

[0030] A treadmill running posture analysis system based on motion capture, the system includes: A high-speed multimodal data acquisition unit is used to acquire multimodal synchronous sensing data of the user's full dynamic range movement on the treadmill; The data fusion and spatiotemporal calibration unit is connected to the high-speed multimodal data acquisition unit to process multimodal synchronous sensing data and output a distortion-resistant user attitude spatiotemporal sequence. The intelligent hybrid attitude parsing unit communicates with the data fusion and spatiotemporal calibration unit. Its built-in intelligent attitude parsing engine based on dynamic prior constraints is used to parse the user's attitude spatiotemporal sequence and output high-dimensional attitude features. The running posture assessment and risk analysis unit communicates with the intelligent hybrid posture analysis unit to calculate refined running posture assessment parameters based on high-dimensional posture features and to conduct risk assessment in conjunction with a multi-level running posture standard model. The multi-channel real-time feedback and guidance unit communicates with the running posture assessment and risk analysis unit to generate and drive the output of adaptive guidance and early warning information based on the risk assessment results.

[0031] The high-speed multimodal data acquisition unit includes a distributed surround array consisting of four high frame rate visual sensors; The visual sensors operate synchronously at a sampling frequency of no less than 240fps and are deployed at the front, rear, and left and right sides of the treadmill running platform, with their main optical axes converging at the typical exercise area of ​​the user in the center of the running platform.

[0032] The data fusion and spatiotemporal calibration unit, the intelligent hybrid attitude analysis unit, and the running posture assessment and risk analysis unit are integrated and deployed in a single edge computing device; Edge computing devices communicate with high-speed multimodal data acquisition units and multi-channel real-time feedback and guidance units via high-speed buses to achieve real-time processing of the entire process from data acquisition to feedback output, with an end-to-end system latency of no more than 100 milliseconds.

[0033] In this technical solution, by integrating the core processing unit into the edge computing device, end-to-end ultra-low latency processing from data acquisition to feedback output is achieved, ensuring the real-time nature of the guidance. Through real-time feedback interfaces with multiple channels of vision, hearing, and touch, the abstract risk assessment results are transformed into specific information that users can easily perceive and understand in real time. This immersive interaction method can provide accurate correction guidance while the user is exercising, significantly improving the efficiency of running posture correction and training experience.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for analyzing running posture on a treadmill based on motion capture, characterized in that, Includes the following steps: Step S1: Acquire multimodal synchronous perception data of the user performing full dynamic range motion on the treadmill, wherein the full dynamic range motion includes steady-state jogging and non-steady-state sprinting and changing direction running; Step S2: Perform high-fidelity fusion and spatiotemporal calibration on the multimodal synchronous sensing data to construct a distortion-resistant user attitude spatiotemporal sequence; Step S3: The user's spatiotemporal posture sequence is analyzed and predicted in real time using an intelligent posture parsing engine based on dynamic prior constraints, in order to compensate for data loss under high-speed motion and extract high-dimensional posture features. Step S4: Based on the high-dimensional posture features, dynamically calculate a set of refined running posture evaluation parameters covering spatial, temporal, and dynamic dimensions; Step S5: Match and assess the risks of the refined running posture evaluation parameters with the preset multi-level running posture standard model adapted to different sports scenarios; Step S6: Based on the results of the risk assessment, generate and drive the multi-channel real-time feedback interface to output adaptation guidance and early warning information.

2. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S1, acquiring multimodal synchronous sensing data specifically includes: Deploy a surround motion capture array consisting of four high frame rate visual sensors, with a sampling frequency of no less than 240fps; The motion capture array is controlled to acquire multi-view raw video streams of the user's motion in full dynamic range using a synchronous triggering method. The multimodal synchronous sensing data also includes dynamic signals from the treadmill running board pressure sensor that are aligned with the video stream timestamp.

3. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S2, the high-fidelity fusion and spatiotemporal calibration processing includes: Denoising and feature-point-based frame-level synchronization alignment are performed on the original video streams from multiple perspectives to eliminate timing errors caused by differences in device response. Based on synchronized multi-view data, a three-dimensional reconstruction algorithm is used to generate real-time temporal three-dimensional coordinates of the user's skeletal joints in the treadmill coordinate system. The time-series three-dimensional coordinates are fused with the synchronously acquired running board pressure signal to perform kinematic data correction and enhancement based on the ground reaction force phase, forming the user attitude spatiotemporal sequence.

4. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S3, the intelligent posture parsing engine based on dynamic prior constraints is a hybrid architecture processor that integrates human motion biomechanical model and data-driven prediction model. The human movement biomechanical model is used to provide reasonable physiological range constraints for joint angles and center of mass trajectory; The data-driven prediction model is used to predict key pose frames that are blurred or lost during high-speed motion within the data sampling interval. The intelligent posture analysis engine outputs smooth, continuous, and physiologically consistent high-dimensional posture features through iterative optimization.

5. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S4, the refined running posture evaluation parameters include: Spatial dimension parameters: stride length, ground contact angle, torso forward lean angle, and lateral sway of center of gravity; Time-related parameters: cadence, ground contact time, and airtime; Dynamic parameters: impact loading rate and left-right foot balance index estimated based on running board pressure signals and kinematic data.

6. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S5, the matching and risk assessment of the multi-level running posture standard model specifically includes: The pre-built running posture standard model library contains a set of benchmark parameters and allowable fluctuation thresholds for different scenarios such as jogging, sprinting, and changing direction running. Based on the real-time identified motion scene, the corresponding benchmark parameter set is called, and the real-time deviation of each evaluation parameter is calculated; A dynamic risk assessment function based on multi-parameter weighted fusion is established. When the function value exceeds the preset warning threshold related to the motion scenario, a graded warning signal containing the risk level and the identifier of the main abnormal parameters is generated.

7. The treadmill running posture analysis method based on motion capture according to claim 1, characterized in that: In step S6, the information output by the multi-channel real-time feedback interface includes: Graphical posture comparison and parameter curves output through the treadmill's embedded display screen or a linked mobile device APP; Real-time voice guidance and rhythm cues output through bone conduction headphones or a speaker system; Vibration alerts synchronized with abnormal postures are output by haptic actuators integrated into smart running shoes or wearable devices.

8. A treadmill running posture analysis system based on motion capture, the system being used to execute the treadmill running posture analysis method based on motion capture as described in any one of claims 1-7, characterized in that, The system includes: A high-speed multimodal data acquisition unit is used to acquire multimodal synchronous sensing data of the user's full dynamic range movement on the treadmill; The data fusion and spatiotemporal calibration unit is communicatively connected to the high-speed multimodal data acquisition unit and is used to process the multimodal synchronous sensing data and output a distortion-resistant user attitude spatiotemporal sequence. The intelligent hybrid attitude parsing unit is communicatively connected to the data fusion and spatiotemporal calibration unit. It has a built-in intelligent attitude parsing engine based on dynamic prior constraints, which is used to parse the user's attitude spatiotemporal sequence and output high-dimensional attitude features. The running posture assessment and risk analysis unit is communicatively connected to the intelligent hybrid posture analysis unit. It is used to calculate refined running posture assessment parameters based on the high-dimensional posture features and to perform risk assessment in combination with a multi-level running posture standard model. A multi-channel real-time feedback and guidance unit is connected in communication with the running posture assessment and risk analysis unit, and is used to generate and drive the output of adaptive guidance and early warning information based on the risk assessment results.

9. The treadmill running posture analysis system based on motion capture according to claim 8, characterized in that: The high-speed multimodal data acquisition unit includes a distributed surround array consisting of four high frame rate visual sensors. The visual sensors operate synchronously at a sampling frequency of no less than 240fps and are deployed in front, behind, and on the left and right sides of the treadmill running platform, with their main optical axes converging in the typical exercise area of ​​the user in the center of the running platform.

10. The treadmill running posture analysis system based on motion capture according to claim 8, characterized in that: The data fusion and spatiotemporal calibration unit, the intelligent hybrid posture analysis unit, and the running posture assessment and risk analysis unit are integrated and deployed in an edge computing device; The edge computing device is connected to the high-speed multimodal data acquisition unit and the multi-channel real-time feedback and guidance unit via a high-speed bus to realize real-time processing of the entire process from data acquisition to feedback output, wherein the end-to-end system latency is no more than 100 milliseconds.

Images