A real-time wearable running posture monitoring system

Abstract

A real-time wearable running posture monitoring system belongs to the field of sports biomechanics and intelligent wearable devices. The application comprises an inertial measurement unit module, a data preprocessing module, a running cycle real-time segmentation module, a running posture parameter real-time calculation module and a user interface real-time monitoring module. The running cycle real-time segmentation module adopts a streaming data processing architecture, constructs a real-time data buffer, identifies key event points frame by frame based on state machine logic of timing constraints, and uses a preset time window to correct the identification results in real time, thereby realizing real-time segmentation of the running cycle. The application combines the zero-speed updating method to correct drift error and real-time calculate the space-time parameters and kinematic parameters of running posture. The application can solve the problems of high processing delay, difficult identification in the starting stage and weak anti-interference ability of the prior art, and realize full-process, low-delay and high-precision monitoring from static start to continuous running. The application is suitable for the fields of sports training, injury prevention and sports rehabilitation.

Description

Technical Field

[0001] This invention relates to the fields of sports biomechanics and smart wearable devices, specifically to a real-time wearable running posture monitoring method and system. Background Technology

[0002] In the fields of sports biomechanics and smart wearable devices, running posture monitoring has become an important reference tool for optimizing athletic performance, preventing sports injuries, and personalizing training. With the growth of the sports population and the development of smart sports, the public's demand for scientific running has increased significantly, and the application scenarios of running posture monitoring technology are becoming increasingly widespread, covering not only professional competitive sports but also extending to multiple fields such as mass fitness, sports rehabilitation, medical rehabilitation, smart wearables, and sports education. However, existing running posture monitoring methods face many challenges, mainly in terms of insufficient real-time performance and poor algorithm robustness.

[0003] First, while traditional laboratory-level measurement methods, such as optical motion capture systems and force tables, provide high-precision data, these devices are typically expensive, require specific venues and environments, and are difficult to apply in daily training or outdoor sports. Furthermore, their bulky size and fixed installation make them inconvenient for athletes to use in various sports environments. Additionally, these systems lack real-time capabilities, failing to provide immediate feedback and hindering effective adjustments by athletes and coaches during training.

[0004] Secondly, while wearable and portable monitoring devices, which have emerged in recent years, have solved the portability problem, they have significant shortcomings in data processing architecture. Existing wearable monitoring technologies generally adopt an offline or batch processing mode of "collection-storage-processing," that is, collecting a complete set of data first and then performing overall analysis. This architecture results in significant delays in monitoring results and cannot meet the real-time feedback needs of "running and watching" during running.

[0005] Furthermore, existing algorithms lack robustness in identifying key event points. Due to drift errors, noise interference, and impact vibrations during running, traditional single-threshold judgment algorithms are highly susceptible to interference, leading to errors in identifying key event points such as ground contact and takeoff, which in turn affects the accuracy of parameters such as stride frequency and ground contact time. At the same time, existing algorithms are mostly designed for steady-state running processes, often neglecting the unique characteristics of the first step, resulting in data loss or identification errors at system startup, and failing to achieve accurate monitoring throughout the entire process. Summary of the Invention

[0006] To address the issues of poor real-time performance and insufficient robustness in existing running posture monitoring technologies, this invention aims to provide a real-time wearable running posture monitoring system. This system employs a streaming data processing architecture combined with time-constrained state machine logic to achieve real-time and accurate monitoring of the entire running process, from a standstill to continuous running. The system described in this invention can provide athletes, running enthusiasts, coaches, and the rehabilitation medicine field with a scientific, accurate, and real-time running posture monitoring tool. It is applicable to multiple scenarios such as sports training, sports rehabilitation, scientific research, and health management, contributing to the intelligentization and widespread application of running posture monitoring technology.

[0007] The objective of this invention is achieved through the following technical solution.

[0008] The present invention discloses a real-time wearable running posture monitoring system, including an inertial measurement unit (IMU) module, a data preprocessing module, a real-time running cycle segmentation module, a real-time running posture parameter calculation module, and a real-time user interface monitoring module.

[0009] An inertial measurement unit module is fixedly installed on the upper of the runner's left and right shoes, respectively. It is used to collect triaxial acceleration data and triaxial angular velocity data of the foot at a preset sampling frequency, and transmit the data to a computing terminal via wireless communication.

[0010] The data preprocessing module is used to perform coordinate system alignment on the data collected by the left and right foot inertial measurement units, and to filter the aligned acceleration and angular velocity data to obtain standardized real-time motion data.

[0011] The real-time segmentation module for running cycles is configured to adopt a streaming data processing architecture. By constructing a real-time data buffer and using state machine logic based on time constraints, it identifies key event points in the running process frame by frame and uses a preset time window to make immediate corrections to the identification results, thereby achieving real-time segmentation of the running cycle.

[0012] The real-time running posture parameter calculation module is used to update the attitude of the inertial measurement unit on the segmented running cycle data and calculate the running posture parameters of the corresponding running cycle in real time.

[0013] The user interface real-time monitoring module is used to display and update the running cycle and running posture parameters in real time.

[0014] Furthermore, the inertial measurement unit module is configured such that the x-axis points towards the toe of the shoe, the z-axis points vertically upward, and the triaxial acceleration data and triaxial angular velocity data are collected at a sampling frequency of 60 Hz.

[0015] Furthermore, the data preprocessing module inverts some axial data from the left and right foot inertial measurement units to unify the left and right foot data into the same human body reference coordinate system, where the X-axis points forward, the Y-axis points outward, and the Z-axis points towards the ground. A second-order Butterworth low-pass filter with a cutoff frequency of 6 Hz is then used to filter the data.

[0016] Furthermore, the real-time running cycle segmentation module employs streaming data processing. By constructing a real-time data buffer, it identifies key event points in the running process frame by frame and instantly segments the complete running cycle based on these key event points. These key event points include the contact point (IC), departure point (TC), maximum swing point (MSW), and minimum speed point (MV).

[0017] Furthermore, the real-time running cycle segmentation module is implemented based on the following steps:

[0018] Step 1: Real-time Data Buffering: The acquired radial angular velocity signals are added to the data buffer in real time, sequentially over time. Real-time calculation begins immediately when the buffer size meets the preset minimum frame threshold; otherwise, the system waits for data to be filled.

[0019] Step 2: Real-time gradient calculation: If the buffer size is greater than 3 frames, then calculate the radial angular velocity values ​​from the three adjacent frames. and Real-time calculation of angular velocity gradient value and The calculation formula is: , If the buffer size is less than 3 frames, continue waiting for data to be filled.

[0020] Step 3: Real-time state machine identification of key event points: Using state machine logic, the key event points are judged frame by frame according to the timing constraints. If the previous event is not successfully identified, the subsequent event will not be judged.

[0021] Step 4: Real-time correction of key event points: After marking key event points, the marking results are dynamically corrected using real-time data within a preset time window.

[0022] Step 5: Real-time segmentation of running cycle: Using the currently identified MV point and the previous MV point as the boundary, the real-time collected triaxial acceleration and triaxial angular velocity data are instantly divided into single stride data corresponding to the running cycle, thus completing the real-time segmentation of the running cycle.

[0023] Furthermore, the implementation method for step three is as follows:

[0024] Step 1: Abrupt Change Identification Mode: Identify the minimum velocity point MV to anchor the initial period. For the judgment frame n, if the angular velocity value satisfies the abrupt change condition from rest to motion... and and If the minimum speed point MV is found, then the process proceeds to determine the departure point TC: if the conditions are met... and and If the result is not found, it is determined to be a TC (Tracking Point) at the departure point. After a successful TC determination, the system automatically switches to a loop determination mode that is not the first step.

[0025] Non-first step loop judgment mode: When the system is in the continuous running phase, it performs frame-by-frame scanning judgment according to the timing sequence of MSW, IC, TC, and MV.

[0026] Maximum Swing Point (MSW) Determination: If the following conditions are met... and and If so, the current frame will be immediately marked as MSW.

[0027] Contact point IC determination: After successful MSW marking, determine whether subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as IC.

[0028] Location-based TC determination: After successful IC marking, a 150ms waiting period is followed by a check to determine if subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as TC.

[0029] Minimum velocity point MV determination: After both IC and TC are successfully marked, a sliding window calculation is performed on the data between the two, and the midpoint of the window containing the minimum sum of squares is marked as MV.

[0030] Furthermore, step four is implemented as follows:

[0031] Maximum Swing Point (MSW) Correction: If a larger angular velocity value is detected within 200ms after the MSW is marked, the MSW position is updated.

[0032] Contact point IC correction: If an angular velocity value is detected within 50ms after IC marking. Less than the original IC value Subtract 10° / s (i.e.) If so, then update the IC position.

[0033] Furthermore, the real-time running posture parameter calculation module determines the initial attitude of the inertial measurement unit by calculating the acceleration data and gravitational acceleration vector in a stationary state before the start of running, and updates the attitude of the inertial measurement unit based on real-time angular velocity data using the quaternion differential method.

[0034] Furthermore, the running posture parameters include time parameters, spatial parameters, and kinematic parameters. The time parameters include at least stride time, ground contact time, swing time, two-foot support time, air time, duty cycle, and air ratio. The spatial parameters include at least stride length, stride frequency, and pace. The kinematic parameters include at least landing angle, landing pattern, and foot pronation / supination angle.

[0035] This invention also discloses a real-time wearable running posture monitoring method, implemented based on the aforementioned real-time wearable running posture monitoring system. The implementation steps of the real-time wearable running posture monitoring method are as follows:

[0036] Inertial measurement units (IMUs) fixed to the uppers of the runner's left and right shoes respectively collect real-time triaxial acceleration and triaxial angular velocity data of the feet. The collected data undergoes coordinate system alignment and filtering to obtain standardized real-time motion data. A real-time data buffer is constructed, and a time-constrained state machine logic is used to identify key event points in the running process frame by frame. Preset time windows are used to instantly correct the identification results to segment the running cycle. The segmented running cycle data undergoes initial orientation calibration and attitude update by the IMUs. Based on the updated attitude data, the running posture parameters for the corresponding running cycle are calculated in real-time and displayed.

[0037] Beneficial effects:

[0038] 1. The present invention discloses a real-time wearable running posture monitoring system and method, which adopts a streaming data processing architecture that constructs a real-time data buffer and calculates the angular velocity gradient frame by frame. This enables real-time reception and calculation of running data, breaks the latency limitation brought about by overall data analysis, and thus meets the low-latency real-time feedback requirement of "running and watching" during running, significantly improving the timeliness and user experience of running posture monitoring.

[0039] 2. The present invention discloses a real-time wearable running posture monitoring system and method, which adopts a dual-mode state machine logic that distinguishes between "first step mutation identification" and "non-first step cycle judgment", and performs frame-by-frame judgment based on strict timing constraints of MSW, IC, TC, and MV, thereby realizing rapid anchoring of the initial cycle and anti-logic conflict judgment in the continuous running stage, and realizing complete cycle monitoring of the entire process from static standing to continuous running.

[0040] 3. The present invention discloses a real-time wearable running posture monitoring system and method, which adopts a dynamic correction strategy of secondary optimization using real-time data within a preset time window after marking key event points. This achieves effective filtering and correction of inertial sensor drift error, high-frequency noise, and real-time data transmission jitter, thereby ensuring high robustness and high accuracy in identifying key event points such as ground contact and ground lift, and providing a reliable guarantee for the calculation of running posture time parameters.

[0041] 4. The present invention discloses a real-time wearable running posture monitoring system and method, which adopts a zero-speed update (ZUPT) method based on the running cycle segmented in real time and combines the characteristic that the minimum speed point (MV point) is approximately stationary to correct the linear error of the integral speed. This effectively eliminates the accelerometer bias error and integral drift, thereby significantly improving the calculation accuracy of spatial parameters such as stride, stride length, and pace, and ensuring that the monitoring system achieves high accuracy while having high real-time performance. Attached Figure Description

[0042] Figure 1 This is a system block diagram of a real-time wearable running posture monitoring system.

[0043] Figure 2 A schematic diagram showing the alignment of the inertial measurement unit with the sensor coordinate system;

[0044] Figure 3 This is a flowchart for identifying key event points based on state machine logic.

[0045] Figure 4 This is a diagram illustrating the segmentation of a running cycle. Detailed Implementation

[0046] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0047] like Figure 1 As shown in the figure, this embodiment discloses a real-time wearable running posture monitoring system, including an inertial measurement unit (IMU) module, a data preprocessing module, a real-time running cycle segmentation module, a real-time running posture parameter calculation module, and a real-time user interface monitoring module.

[0048] The inertial measurement unit module is fixed to the upper part of the left and right shoes by the user. The front end of the module's x-axis points to the toe of the shoe, and the z-axis points vertically upward. For details on aligning the inertial measurement unit with the sensor coordinate system, please refer to [link / reference]. Figure 2 The system maintains a 60Hz sampling frequency during operation, collecting real-time triaxial acceleration data of the runner's feet. , , ) and triaxial angular velocity ( It transmits data to the computing terminal in real time via Bluetooth Low Energy (BLE) wireless communication using a streaming protocol.

[0049] The data preprocessing module, due to the difference in coordinate system orientation between the left and right feet of the IMU data reflecting the symmetry of the human body, performs coordinate transformation on the collected acceleration and angular velocity data to ensure unified algorithmic calculations and directional representation for both feet. Combined with... Figure 2 The sensor coordinate system alignment is illustrated below. The specific operation is as follows: Invert the values ​​of the z-axis of the left foot accelerometer, the x, y, and z-axis of the left foot gyroscope, the y and z-axis of the right foot accelerometer, and the y-axis of the right foot gyroscope. This aligns the original sensor coordinate systems of the left and right feet with the human body reference coordinate system (the positive X-axis points forward, the positive Y-axis points outward, and the positive Z-axis points towards the ground). Subsequently, a second-order Butterworth low-pass filter with a cutoff frequency of 6Hz is used to filter the data. This 6Hz cutoff frequency is determined based on the basic stride frequency of human running (usually 1.5Hz~3.5Hz) and its main harmonic distribution. It effectively filters out high-frequency vibration noise (usually greater than 10Hz) generated by the foot impacting the ground while preserving the characteristic shape of the original angular velocity waveform to the greatest extent possible without distortion, providing high signal-to-noise ratio standard real-time data for subsequent state machine gradient calculations.

[0050] The real-time running cycle segmentation module is the core module of this invention. For example... Figure 3 The flowchart shown illustrates the key event point identification process. This module is configured to use streaming data processing. By constructing a real-time data buffer, it identifies key event points (maximum swing point MSW, contact point IC, departure point TC, minimum speed point MV) frame by frame during the running process, and segments the complete running cycle based on these key event points. The specific implementation method includes the following steps:

[0051] Step 1: Real-time data buffering: The acquired radial angular velocity signals are added to the data buffer in real time according to the time sequence; when the buffer size meets the preset minimum frame threshold (e.g., 3 frames), real-time calculation is started immediately; otherwise, continue to wait for data to be filled.

[0052] Step 2: Real-time gradient calculation: If the buffer size is greater than 3 frames, then calculate the radial angular velocity values ​​from the three adjacent frames. , and Real-time calculation of angular velocity gradient value and The calculation formula is: , .

[0053] Step 3: Real-time state machine identification of key event points:

[0054] State machine logic is used to distinguish between the running start phase and the continuous running phase. Decisions are made frame-by-frame based on the timing constraints of key event points. For details on the key event point identification process, please refer to [link / reference]. Figure 3 :

[0055] The first step is the mutation identification mode: The system monitors whether it is in the running start phase. If so, it prioritizes identifying the minimum velocity point MV to anchor the initial cycle. Specifically, for the judgment frame n, if the angular velocity value satisfies the mutation condition from rest to motion... and and If the minimum speed point MV is determined, then the process proceeds to determine the departure point TC: if the conditions are met... and and If the result is not found, it is determined to be a TC (Tracking Point) at the departure point; after a successful TC determination, the system automatically switches to a loop determination mode that is not the first step.

[0056] Non-first-step loop determination mode: When the system is in the continuous running phase, frame-by-frame scanning and determination are performed according to the timing sequence of MSW, IC, TC, and MV.

[0057] Maximum Swing Point (MSW) Determination: If the following conditions are met... and and If so, the current frame will be immediately marked as MSW;

[0058] Contact point IC determination: After successful MSW marking, determine whether subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as IC.

[0059] Location-based TC determination: After successful IC marking, a 150ms waiting period is followed by a check to determine if subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as TC;

[0060] Minimum velocity point MV determination: After both IC and TC are successfully marked, a sliding window calculation is performed on the data between the two, and the midpoint of the window containing the minimum sum of squares is marked as MV.

[0061] Step 4: Real-time correction of key event points:

[0062] After marking key event points, the marking results are dynamically corrected using real-time data within a preset time window, specifically including:

[0063] Maximum Swing Point (MSW) Correction: If a point with a larger angular velocity value is detected within 200ms after the MSW is marked, the MSW position is updated.

[0064] Contact point IC correction: If an angular velocity value is detected within 50ms after IC marking. Less than the original IC value Subtract 10° / s (i.e.) If so, then update the IC position.

[0065] Step 5: Instantly segment the running cycle:

[0066] Using the currently identified MV point and the previous MV point as boundaries, the real-time collected triaxial acceleration and triaxial angular velocity data are instantly divided into single-step data corresponding to the running cycle, completing the real-time segmentation of the running cycle. See details of the running cycle segmentation results. Figure 4 The division of a complete running cycle will provide a basis for subsequent calculations of running posture parameters.

[0067] The real-time running posture parameter calculation module, upon receiving the segmented accelerometer and gyroscope data from the calculation terminal, immediately performs IMU initial orientation calibration, IMU orientation update, and running posture parameter calculation, calculating the running posture parameters in real time for the current input running cycle. The specific method is as follows:

[0068] IMU Initial Orientation Calibration: Before system measurement, to determine the orientation of the IMU in its initial state, the initial orientation needs to be calculated using the static acceleration vector and the gravitational acceleration vector for initial attitude calibration. The specific method is as follows: After the system starts data acquisition, the runner is required to remain stationary for at least 0.25 seconds. At this time, the orientation can be initialized using the IMU's accelerometer data: The three-axis acceleration vectors of the first four frames of real-time running data are taken. , and the gravitational acceleration vector The axis of rotation is obtained through cross product. and standardize. Rotation angle The following relationship exists between the sensor's static acceleration and gravitational acceleration: The rotation angle can be calculated using the inverse cosine function. Take the midpoint between the rotation axis and the rotation angle of these four frames. , This is related to the axis of rotation formed by gravitational acceleration. With rotation angle This refers to the initial orientation of the IMU. The initial orientation calibration can be represented as a quaternion of a rotating object. The unit vector of the equivalent rotation axis direction The rotation angle is The transformation relationship has , .

[0069] IMU Orientation Update: During running, the angular velocity collected by the IMU can be used to update the attitude orientation in real time. The quaternion differential method can be used to describe the change in the IMU's rotational direction over a short period of time. The calculation method is as follows: using the real-time angular velocity collected by the IMU... The rate of change of quaternions over time It can be calculated that: The current angular velocity Multiplying the quaternion by 2 and dividing by 2 gives the rate of change of the quaternion. Initial direction With frame-by-frame quaternion change rate By summing the integrals, we can obtain the updated IMU direction. Use this direction to process the triaxial accelerometer data frame by frame. Rotate to obtain the updated acceleration. ,in Representing quaternions . conjugate.

[0070] Running posture parameter calculation: Using running cycle, direction, speed, and displacement data, the spatiotemporal and kinematic parameters of running posture are calculated in real time. The specific calculation method is as follows:

[0071] The detailed calculation process for the time parameter is as follows:

[0072] Stride time: The difference in time between two consecutive foot strikes the ground. .

[0073] Ground contact time: The time from when the foot touches the ground to when the toes leave the ground. Ground contact time can be determined by the time difference between the ground contact and takeoff events within a stride. .

[0074] Swing time: The time interval between when one foot leaves the ground and when the same foot touches the ground again. .

[0075] Double support time: The time that both feet are in contact with the ground at the same time. , .

[0076] Flight time: The time interval between one foot leaving the ground and the other foot touching the ground. , .

[0077] Duty factor: The ratio of the support time of one foot during the gait cycle to the total gait cycle time. .

[0078] Flight ratio: The proportion of time spent in the air during the entire gait cycle. .

[0079] The detailed calculation process for spatial parameters is as follows:

[0080] Using the trapezoidal integral method to analyze the acceleration sequence By integrating, the velocity sequence during the running process can be obtained. At this point, the running speed exhibits significant integral drift and bias errors, necessitating the use of the zero-speed update method to correct the velocity sequence. The zero-speed update technique leverages the principle that the IMU is almost stationary at the MV point of the running cycle. Assuming the actual velocity at MV point is zero, the difference between this velocity and the velocity calculated through integration within the running cycle is used to correct the errors caused by the IMU's integral drift. Specifically, it is expressed as follows: ,in This indicates the acceleration measured by the accelerometer. Let ε represent the actual acceleration and ε represent the bias error. Assuming the bias error is constant during the short period of the swing phase, and the foot velocity is zero at the beginning of the swing phase, then we have... ,in The velocity, obtained by integral calculation, consists of two parts: The true velocity is derived from the accelerometer's integral; the other component is the product of the error ε and t, which arises from the integral error, i.e., drift. At the MV point, i.e., t=T, the IMU is approximately stationary, and the true velocity is assumed to be 0. Therefore, the error can be calculated. Adjust the speed accordingly during the running cycle: , obtained This represents the corrected velocity sequence. The displacement can be calculated by performing a trapezoidal integral on the corrected velocity sequence. .

[0081] Stride length: The Euclidean distance of the change in displacement in the forward and backward and offset directions within a single stride. .

[0082] Step length: The distance from when one foot leaves the ground to when the other foot touches the ground, which is half of the stride. .

[0083] Step frequency: the number of steps taken per minute. .

[0084] Stride pace: the number of minutes required to travel one kilometer. .

[0085] The detailed calculation process for kinematic parameters is as follows:

[0086] Dorsiflexion angle: The angle between the sole of the foot and the ground at the point of contact. During plantar flexion of the ankle, the toes point downwards, increasing the distance between the instep and the lower leg. Conversely, dorsiflexion involves the toes pointing upwards, bringing the instep closer to the lower leg. The dorsiflexion angle is calculated as follows: Let... Let be the set of quaternions of the direction sequence after trajectory reconstruction, where This represents the direction quaternion at time i. The direction sequence quaternion after trajectory reconstruction is converted into Euler angles. Take the value of the landing time k on the sagittal axis, and the landing angle. .

[0087] Foot strike type refers to the part of the foot that contacts the ground at the point of impact. It is categorized by different ranges of landing angle, calculated as follows: the plantarflexion / dorsiflexion angle of the ankle joint at landing is divided into three different ranges, corresponding to heel strike, midfoot strike, and forefoot strike. Rearfoot strike: Dorsiflexion angle > 8.0°; Midfoot strike: −1.6 < Dorsiflexion angle ≤ 8.0°; Forefoot strike: Dorsiflexion angle ≤ −1.6°.

[0088] Foot pronation / supination angles: Pronation refers to the foot tilting inwards towards the midline of the body, with the inner side of the foot raised and the outer side lowered; supination refers to the foot tilting outwards towards the midline of the body, with the outer side of the foot raised and the inner side lowered. Pronation is considered a potential indicator of running injury, but it is also a way to absorb shock upon landing. The pronation / supination angles are calculated by converting the quaternion of the reconstructed trajectory direction sequence into Euler angles. Take the value at the moment of impact on the frontal axis, i.e., the inversion / valgus angle of the foot.

[0089] The user interface real-time monitoring module displays and updates the running cycle segmentation and running posture parameters in real time.

[0090] Through actual testing, the system described in this invention can output various running posture indicators in real time during tests conducted on a treadmill. Compared with the treadmill benchmark data and the OptoJump photoelectric testing system, the consistency correlation coefficient of key parameters (such as ground contact time and cadence) reaches more than 0.9, proving that the system can ensure high accuracy of monitoring results while possessing high real-time performance.

[0091] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A real-time wearable running posture monitoring system, characterized in that: It includes an inertial measurement unit (IMU) module, a data preprocessing module, a real-time running cycle segmentation module, a real-time running posture parameter calculation module, and a real-time user interface monitoring module; An inertial measurement unit module is fixedly installed on the upper of the runner's left and right shoes, respectively. It is used to collect triaxial acceleration data and triaxial angular velocity data of the foot at a preset sampling frequency, and transmit the data to the computing terminal via wireless communication. The data preprocessing module is used to perform coordinate system alignment on the data collected by the left and right foot inertial measurement units, and to filter the aligned acceleration and angular velocity data to obtain standardized real-time motion data. The real-time segmentation module for running cycles is configured to adopt a streaming data processing architecture. By constructing a real-time data buffer and using state machine logic based on time constraints, it identifies key event points in the running process frame by frame and uses a preset time window to correct the identification results in real time, thereby realizing the real-time segmentation of the running cycle. The real-time running posture parameter calculation module is used to update the attitude of the inertial measurement unit on the segmented running cycle data and calculate the running posture parameters of the corresponding running cycle in real time. The user interface real-time monitoring module is used to display and update the running cycle and running posture parameters in real time.

2. The real-time wearable running posture monitoring system according to claim 1, characterized in that: The inertial measurement unit module is configured such that the x-axis points towards the toe of the shoe, the z-axis points vertically upward, and the triaxial acceleration data and triaxial angular velocity data are collected at a sampling frequency of 60 Hz.

3. The real-time wearable running posture monitoring system according to claim 1, characterized in that: The data preprocessing module inverts some axial data from the left and right foot inertial measurement units to unify the left and right foot data into the same human reference coordinate system, where the X-axis points forward, the Y-axis points outward, and the Z-axis points towards the ground; a second-order Butterworth low-pass filter with a cutoff frequency of 6 Hz is used to filter the data.

4. The real-time wearable running posture monitoring system according to claim 1, characterized in that: The real-time running cycle segmentation module adopts a streaming data processing method. By constructing a real-time data buffer, it identifies key event points in the running process frame by frame and segments the complete running cycle in real time based on the key event points. The key event points include the contact point IC, the departure point TC, the maximum swing point MSW, and the minimum speed point MV.

5. The real-time wearable running posture monitoring system according to claim 4, characterized in that: The real-time running cycle segmentation module is implemented based on the following steps. Step 1: Real-time data buffering: The acquired radial angular velocity signals are added to the data buffer in real time according to the time sequence; when the buffer size meets the preset minimum frame threshold, real-time calculation is started immediately; otherwise, it continues to wait for data to be filled. Step 2: Real-time gradient calculation: If the buffer size is greater than 3 frames, then calculate the radial angular velocity values ​​from the three adjacent frames. and Real-time calculation of angular velocity gradient value and The calculation formula is: , If the buffer size is less than 3 frames, continue waiting for data to be filled; Step 3: Real-time state machine identification of key event points: Using state machine logic, the key event points are judged frame by frame according to the timing constraints. If the previous event is not successfully identified, the subsequent event will not be judged. Step 4: Real-time correction of key event points: After marking key event points, the marking results are dynamically corrected using real-time data within a preset time window; Step 5: Real-time segmentation of running cycle: Using the currently identified MV point and the previous MV point as the boundary, the real-time collected triaxial acceleration and triaxial angular velocity data are instantly divided into single stride data corresponding to the running cycle, thus completing the real-time segmentation of the running cycle.

6. The real-time wearable running posture monitoring system according to claim 5, characterized in that: The implementation method for step three is as follows: Step 1: Abrupt Change Identification Mode: Identify the minimum velocity point MV to anchor the initial period. For the judgment frame n, if the angular velocity value satisfies the abrupt change condition from rest to motion... and and If the minimum speed point MV is determined, then the minimum speed point TC is determined. and and If so, it is determined to be away from location TC; After the TC determination is successful, the system automatically switches to the loop determination mode that is not the first step. Non-first step loop judgment mode: When the system is in the continuous running phase, it performs frame-by-frame scanning and judgment according to the timing sequence of MSW, IC, TC, and MV. Maximum Swing Point (MSW) Determination: If the following conditions are met... and and If so, the current frame will be immediately marked as MSW; Contact point IC determination: After successful MSW marking, determine whether subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as IC. Location-based TC determination: After successful IC marking, a 150ms waiting period is followed by a check to determine if subsequent frames meet the requirements. and and If the condition is met, the current frame will be immediately marked as TC; Minimum velocity point MV determination: After both IC and TC are successfully marked, a sliding window calculation is performed on the data between the two, and the midpoint of the window containing the minimum sum of squares is marked as MV.

7. The real-time wearable running posture monitoring system according to claim 6, characterized in that: Step four is implemented as follows: Maximum Swing Point (MSW) Correction: If a point with a larger angular velocity value is detected within 200ms after the MSW is marked, the MSW position is updated. Contact point IC correction: If an angular velocity value is detected within 50ms after IC marking. Less than the original IC value Subtract 10° / s (i.e.) If so, then update the IC position.

8. The real-time wearable running posture monitoring system according to claim 1, characterized in that: The real-time running posture parameter calculation module determines the initial attitude of the inertial measurement unit by calculating the acceleration data and gravitational acceleration vector in a stationary state before the start of running, and updates the attitude of the inertial measurement unit based on real-time angular velocity data using the quaternion differential method.

9. The real-time wearable running posture monitoring system according to claim 8, characterized in that: The running posture parameters include time parameters, spatial parameters, and kinematic parameters. The time parameters include at least stride time, ground contact time, swing time, two-foot support time, air time, duty cycle, and air ratio. The spatial parameters include at least stride length, stride frequency, and pace. The kinematic parameters include at least landing angle, landing pattern, and pronation / supination angle.

10. A real-time wearable running posture monitoring method, implemented based on the real-time wearable running posture monitoring system as described in claims 1, 2, 3, 4, 5, 6, 7, 8, or 9; characterized in that, The implementation steps are as follows: Inertial measurement units (IMUs) fixed to the uppers of the runner's left and right shoes respectively are used to collect triaxial acceleration and triaxial angular velocity data of the feet in real time. The collected data is then aligned to a coordinate system and filtered to obtain standardized real-time motion data. A real-time data buffer is constructed, and a state machine logic based on time constraints is used to identify key event points in the running process frame by frame. The identification results are then corrected in real time using a preset time window to segment the running cycle. The segmented running cycle data is then subjected to initial orientation calibration and attitude update of the IMUs. Based on the updated attitude data, the running posture parameters for the corresponding running cycle are calculated in real time and displayed in real time.

Images