Human body posture stability intelligent early warning system for running

Abstract

The invention discloses a human body posture stability intelligent early warning system for running. The system comprises a sensor module used for collecting a motion signal and a pressure sensing signal of a runner; the information preprocessing module is used for preprocessing the motion signals and the pressure sensing signals through calibration and filtering; the edge processing module is used for mapping the preprocessed pressure sensing signal into runner gravity center position information and mapping the preprocessed motion signal into runner posture angle information to form a state vector measurement value of the runner; the Kalman filtering module is used for predicting the state vector of the runner according to the state vector measurement value of the runner; and the posture stability evaluation module is used for constructing a posture stability evaluation model according to the predicted state vector of the runner, calculating the posture stability of the runner and carrying out early warning if the posture stability of the runner is smaller than a posture stability threshold value. The posture stability of the runner can be accurately monitored, and early warning can be carried out in advance.

Description

Technical Field

[0001] This invention relates to the field of sports equipment safety technology, specifically to an intelligent early warning system for human posture stability during running. Background Technology

[0002] Treadmills are among the most commonly used fitness equipment in gyms and rehabilitation centers, and their safety has always been a major concern. In recent years, with the continuous development of sensor technology and signal processing technology, treadmill safety monitoring devices have also made some progress.

[0003] Traditional treadmill safety monitoring devices primarily use a single sensor, such as a pressure sensor or accelerometer, installed on the treadmill to monitor the runner's postural stability. When the sensor detects an abnormal signal, it triggers an alarm to alert the runner.

[0004] Monitoring data from a single sensor is easily affected by external interference, leading to inaccurate monitoring results. Furthermore, traditional Kalman filtering algorithms are not precise enough in estimating noise when processing sensor data and cannot dynamically adapt to changes in noise, thus affecting the accuracy of attitude stability assessment. In addition, existing devices have relatively simple early warning mechanisms, providing only basic alarm functions and lacking effective preventative measures. They cannot predict fall trends in advance, and after an early warning, they rely heavily on manual intervention, making it difficult to proactively adjust equipment parameters to reduce risk.

[0005] Meanwhile, existing technologies lack visualization of runners' postures, making it difficult for managers and users to intuitively understand posture changes. Furthermore, they do not fully utilize historical data for personalized analysis, failing to provide customized exercise advice and equipment maintenance solutions for different users. The level of intelligence and personalization needs to be improved. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides an intelligent early warning system for human posture stability during running, which can accurately monitor the posture stability of runners and provide early warnings, thereby improving the safety and intelligence of treadmill use.

[0007] To achieve the above technical objectives, the present invention adopts the following technical solution: A smart early warning system for human posture stability during running includes: a sensor module, an information preprocessing module, an edge processing module, a Kalman filter module, and a posture stability evaluation module; The sensor module is used to collect the runner's motion signals and pressure sensing signals; The information preprocessing module is used to perform calibration and filtering preprocessing on the collected motion signals and pressure sensing signals; The edge processing module is used to map the pre-processed pressure sensing signal into the runner's center of gravity position and the rate of change of the runner's center of gravity position, and to map the pre-processed motion signal into the runner's posture angle and the rate of change of the runner's posture angle, forming the runner's state vector measurement value. The Kalman filter module is used to predict the runner's state vector based on the runner's state vector measurement value; The posture stability assessment module is used to construct a posture stability assessment model based on the predicted state vector of the runner, calculate the posture stability of the runner, and issue an early warning if the posture stability of the runner is less than the posture stability threshold.

[0008] Furthermore, the runner's state vector measurement value ,in, express The x-axis measurement of the runner's center of gravity at that moment. express The vertical coordinate measurement of the runner's center of gravity at that moment. express Measurements of the runner's posture angles at any given time. express Measurement of the rate of change of the runner's center of gravity at any given moment. express Measurement of the vertical rate of change of the runner's center of gravity at any given time. express The measured values ​​of the runner's posture angle and velocity at any given moment. This indicates the transpose operation.

[0009] Furthermore:

[0010]

[0011]

[0012]

[0013]

[0014]

[0015] in, Indicates the first in the pressure array The coordinates of each unit, Indicates the first in the pressure array Unit The corresponding pressure value at that moment. express The triaxial acceleration of the runner's torso in the motion signal at the given moment. express The runner's torso in the motion signal at that moment axial angular velocity, Indicates the sampling time. This indicates the weight associated with angular velocity.

[0016] Furthermore, the specific process by which the Kalman filter module predicts the runner's state vector based on the runner's state vector measurement is as follows: i: Based on the motion signals and pressure sensing signals of the runner standing still on the treadmill, set the initial state vector measurement values. And set the initial state vector estimation error covariance matrix. State vector process noise covariance matrix and state vector measurement noise covariance matrix ; ii: Based on the runner's state vector predicted from the previous moment and state transition matrix Estimate the state vector at the current time step. And combine the state vector from the previous time step to estimate the error covariance matrix. The state vector and process noise covariance matrix of the previous time step The covariance matrix of the prediction error for the current state vector is calculated. ; iii: Estimate the error covariance matrix based on the predicted state vector at the current moment. Combine the state vector from the previous time step with the noise covariance matrix. Calculate Kalman gain ,in, The measurement matrix represents the state vector at the current moment; iv: Based on the estimated state vector at the current time The current state vector measurement value and Kalman gain Predict the runner's state vector at the current moment. ; v: Measurement matrix based on the current state vector. and the estimated state vector at the current time. Calculate the state vector to estimate the measured value Based on the measured value of the state vector at the current moment With state vector estimation measurement Residual vector update , Repeat steps ii-iv to continue using the state vector of the runner at the next moment.

[0017] Furthermore, based on the current state vector measurement value With state vector estimation measurement Residual vector update , The specific process is as follows: If the mean square value of the residual vector is greater than 0.0003 for several consecutive iterations, the state vector measurement noise covariance matrix will be... Increase the scaling factor; if the mean square value of the residual vector is less than 0.0001 for several consecutive iterations, decrease the scaling factor of the state vector measurement noise covariance matrix; otherwise, ; Calculate the rate of change of the runner's center of gravity in the residual vector at the current moment. ,like The state vector process noise covariance matrix The standard deviation of process noise related to the rate of change of the runner's center of gravity is increased; if The state vector process noise covariance matrix Reduce the standard deviation of process noise related to the rate of change of the runner's center of gravity; if , ;in, Indicates the current time Measurement of the rate of change of the runner's center of gravity horizontally. Indicates the current time Estimates of the rate of change of the center of gravity of runners at different levels. Indicates the current time Measurement of the vertical rate of change of the runner's center of gravity. Indicates the current time Estimated rate of vertical change of the center of gravity of a runner.

[0018] Furthermore, the construction process of the attitude stability evaluation model is as follows:

[0019] in, Indicates the current time Postural stability of runners Indicates the current time The stability of the runner's center of gravity position , Indicates the current time The predicted runner's center of gravity coordinates in the runner's state vector. This represents the coordinates of the runner's center of gravity in the initial measurement. Indicated Weight; Indicates the current time Stability of runner's posture angle , Indicates the current time The predicted runner's posture angle values ​​in the runner's state vector. express The weights; Indicates the current time treadmill speed stability , Indicates the current time The speed value when getting off the treadmill This represents the average speed of the treadmill over the past several seconds. express The weights; Indicates the current time Treadmill incline stability , Indicates the current time The incline value of the treadmill. This indicates the average incline value of the treadmill over the past several seconds. express The weights; Indicates the current time Stability of the rate of change of the center of gravity of the runner , Indicates the current time The predicted rate of change of the runner's center of gravity in the runner's state vector. express The weights; , Indicates the current time The predicted rate of change of the runner's posture angle in the runner's state vector. express The weight.

[0020] Furthermore, it also includes a digital twin, which constructs a dynamic virtual representation of the runner based on the predicted state vector of the runner, visualizes it, and renders the runner's posture stability in the dynamic virtual representation.

[0021] Furthermore, it also includes a prediction module, which uses an LSTM model to predict the runner's posture stability at the next moment based on the runner's state vector predicted by the Kalman filter module.

[0022] Furthermore, the state vector of the runner predicted in the history of the Kalman filter module is used as the input of the prediction module, and the posture stability of the corresponding runner's state vector calculated by the posture stability evaluation module is used as the label of the prediction module. The prediction module is trained until the mean squared error loss function of the predicted posture stability and the corresponding label converges, thus completing the training of the prediction module.

[0023] Furthermore, the mean square error loss function Represented as:

[0024] in, This represents the number of samples used to train the prediction module. express index, This indicates the first prediction made by the prediction module. The pose stability of each sample Indicates the first The attitude stability of each sample was calculated using the attitude stability assessment module.

[0025] Compared with the prior art, the present invention has the following beneficial effects: (1) The intelligent early warning system for human posture stability for running of the present invention can adaptively adjust the state vector process noise covariance matrix and the state vector measurement noise covariance matrix by setting a Kalman filter module. It can dynamically adapt to the changes in the runner's motion state and the fluctuations in sensor measurement error, thereby accurately predicting the runner's state vector based on the runner's state vector measurement value. This adaptive adjustment makes the Kalman filter module more robust when dealing with complex motion scenarios, significantly improves the accuracy of runner state vector prediction, and provides high-quality data support for subsequent stability assessment. (2) The intelligent early warning system for human posture stability of running in this invention sets up a posture stability assessment module, constructs a posture stability assessment model based on the predicted state vector of the runner, and calculates the posture stability of the runner by comprehensively considering the runner's center of gravity position, center of gravity change rate, posture angle and posture angle change rate. It can comprehensively reflect the runner's dynamic balance ability and posture adjustment ability, realize accurate monitoring of the runner's posture stability, and promptly remind the runner and management personnel when the posture stability is lower than the posture stability threshold. (3) The intelligent early warning system for human posture stability for running in this invention is also equipped with a prediction module to predict the posture stability of the runner at the next moment, thereby effectively preventing the runner from falling due to unstable posture and ensuring his safety on the treadmill. Attached Figure Description

[0026] Figure 1This is a schematic diagram of the intelligent early warning system for human posture stability during running according to the present invention; Figure 2 This is a flowchart of the Kalman filter module predicting the runner's state vector based on the runner's state vector measurement values ​​in this invention; Figure 3 This is a schematic diagram illustrating the prediction of a runner's posture stability by the prediction module in this invention. Detailed Implementation

[0027] The technical solution of the present invention will be further explained and described below with reference to the accompanying drawings.

[0028] like Figure 1 This is a schematic diagram of the intelligent early warning system for human posture stability during running according to the present invention. The system includes: a sensor module, an information preprocessing module, an edge processing module, a Kalman filter module, a posture stability assessment module, a digital twin, a prediction module, and a data storage module. It can accurately monitor the posture stability of runners and provide early warnings to improve the safety and intelligence of treadmill use.

[0029] In this invention, the sensor module mounts an accelerometer and a gyroscope sensor on the runner's waist to collect the runner's motion signals in real time at a sampling frequency of 100Hz, including the runner's three-axis acceleration. and triaxial angular velocity The pressure sensor is placed on the treadmill belt and the pressure sensing signal is collected in real time at a sampling frequency of 100Hz. The collected motion signal and pressure sensing signal of the runner are sent to the information preprocessing module via Bluetooth serial port.

[0030] In this invention, the information preprocessing module is used to perform calibration and filtering preprocessing on the acquired motion signals and pressure sensing signals; specifically: Since the pressure sensor uses a piezoresistive thin-film sensor, its resistance decreases as the pressure increases. First, the voltage divider circuit converts the resistance change into a corresponding voltage signal. This signal amplitude is usually small and may contain interference. Therefore, an RC filter circuit is needed to filter out high-frequency noise such as power supply noise and electromagnetic interference to make the signal smoother and more stable. Then, the voltage amplifier circuit boosts it to the optimal input range of the ADC to ensure measurement resolution. Finally, the ADC completes the conversion from analog signal to digital signal to obtain the pre-processed pressure sensing signal. Because accelerometers and gyroscopes are susceptible to inherent component errors such as zero bias and scaling factor mismatch, as well as environmental factors such as temperature drift and external magnetic field distortion during production, assembly, and use, their direct output data exhibits deviations. These errors require calibration to eliminate them. Specifically, zero bias calibration is performed on gyroscopes, fitting the zero drift curve and random noise; six-sided static calibration is implemented on accelerometers to correct zero bias and installation errors, ensuring measurement accuracy. Furthermore, since accelerometers and gyroscopes are easily affected by noise interference when acquiring human motion signals, the received data needs to be filtered to obtain pre-processed motion signals.

[0031] In this invention, the edge processing module is used to map the preprocessed pressure sensing signal into the runner's center of gravity position and the rate of change of the runner's center of gravity position, and to map the preprocessed motion signal into the runner's posture angle and the rate of change of the runner's posture angle, thus forming the runner's state vector measurement value: , in, express The x-axis measurement of the runner's center of gravity at that moment. , express The vertical coordinate measurement of the runner's center of gravity at that moment. , Indicates the first in the pressure array The coordinates of each unit, Indicates the first in the pressure array Unit The pressure value at that moment; express Measurements of the runner's posture angles at any given time. , express The triaxial acceleration of the runner's torso in the motion signal at the given moment. express The runner's torso in the motion signal at that moment axial angular velocity, Indicates the sampling time. Indicates the weights related to angular velocity; express Measurement of the rate of change of the runner's center of gravity at any given moment. , express Measurement of the vertical rate of change of the runner's center of gravity at any given time. , express The measured values ​​of a runner's posture, angle, and speed at any given moment can sensitively reflect trends of instability. , This indicates the transpose operation.

[0032] In this invention, the Kalman filter module is used to predict the runner's state vector based on the runner's state vector measurement values. It can adaptively adjust the state vector process noise covariance matrix and the state vector measurement noise covariance matrix, dynamically adapting to changes in the runner's motion state and fluctuations in sensor measurement errors. This allows for accurate prediction of the runner's state vector based on the measured values. This adaptive adjustment makes the Kalman filter module more robust when handling complex motion scenarios, significantly improving the accuracy of runner state vector prediction and providing high-quality data support for subsequent stability assessments. Figure 2 The specific process is as follows: i: Based on the motion signals and pressure sensing signals of the runner standing still on the treadmill, set the initial state vector measurement values. And set the initial state vector estimation error covariance matrix. State vector process noise covariance matrix and state vector measurement noise covariance matrix ,in, Let each of these represent the standard deviation of the estimate for each component in the initial state vector. These represent the standard deviations of process noise, This represents the standard deviation of the noise measured by the accelerometer. This represents the standard deviation of the noise measured by the gyroscope sensor. ii: Based on the runner's state vector predicted from the previous moment and state transition matrix Estimate the state vector at the current time step. And combine the state vector from the previous time step to estimate the error covariance matrix. The state vector and process noise covariance matrix of the previous time step The covariance matrix of the prediction error for the current state vector is calculated. ; iii: Estimate the error covariance matrix based on the predicted state vector at the current moment. Combine the state vector from the previous time step with the noise covariance matrix. Calculate Kalman gain ,in, The measurement matrix represents the state vector at the current moment. ; iv: Based on the estimated state vector at the current time The current state vector measurement value and Kalman gain Predict the runner's state vector at the current moment. ; v: Measurement matrix based on the current state vector. and the estimated state vector at the current time. Calculate the state vector to estimate the measured value Based on the measured value of the state vector at the current moment With state vector estimation measurement Residual vector update , Repeat steps ii-iv to continue using the state vector of the runner at the next moment.

[0033] In one technical solution of the present invention, the measurement value of the state vector at the current moment is used. With state vector estimation measurement Residual vector update , The specific process is as follows: Since the runner's motion state and the sensor measurement environment are dynamically changing, if the mean square value of the residual vector is greater than 0.0003 for several consecutive measurements, the state vector measurement noise covariance matrix will be affected. Increase the scaling factor; if the mean square value of the residual vector is less than 0.0001 for several consecutive iterations, decrease the scaling factor of the state vector measurement noise covariance matrix; otherwise, This can more accurately reflect changes in state vector measurement noise; Calculate the rate of change of the runner's center of gravity in the residual vector at the current moment. ,like The state vector process noise covariance matrix The standard deviation of process noise related to the rate of change of the runner's center of gravity is increased; if The state vector process noise covariance matrix Reduce the standard deviation of process noise related to the rate of change of the runner's center of gravity; if , Dynamically adjust according to the rate of change of the center of gravity It can better adapt to the fluctuations in a runner's exercise state; among them, Indicates the current time Measurement of the rate of change of the runner's center of gravity horizontally. Indicates the current time Estimates of the rate of change of the center of gravity of runners at different levels. Indicates the current time Measurement of the vertical rate of change of the runner's center of gravity. Indicates the current time Estimated rate of vertical change of the center of gravity of a runner.

[0034] In one technical solution of the present invention, ,in, This indicates that the estimated standard deviation of the initial center of gravity position in the horizontal plane is 0.1 meters. In the initial state, due to individual differences and measurement errors, the center of gravity position may have a certain degree of deviation. For example, the center of gravity position of different runners will vary due to their body type and posture. Even with the same height and weight, the center of gravity position may vary by ±0.1 meters. The standard deviation of the initial attitude angle estimate is . This is because even if a runner believes they are maintaining an upright posture, the actual posture angle may deviate slightly from the ideal value due to factors such as muscle tension and shoe sole thickness. For example, when standing still, a runner's body may lean slightly forward or backward due to fatigue, resulting in a ±1 / 2 difference between the measured and actual posture angle values. The error; These represent the estimated standard deviations of the rate of change of the body's center of gravity position and posture angle at the initial moment. These values ​​are determined based on the physiological characteristics of the runner in a static state before starting to run. Although the runner is in a static state at the initial moment, there may still be slight fluctuations in the center of gravity position and posture angle due to physiological factors such as breathing and slight muscle tremors.

[0035] In one technical solution of the present invention, , This indicates that the standard deviation of the process noise in the horizontal plane during exercise is 0.05 meters. It reflects the degree of random change in the center of gravity position caused by factors such as fluctuations in the runner's muscle strength and the irregularity of the running movements. For example, during running, the runner's stride length may vary slightly due to fatigue or changes in rhythm, resulting in a random fluctuation of about ±0.05 meters in the center of gravity position in the front-back and left-right directions. This indicates that the standard deviation of the process noise affecting the attitude angle during motion is... The random changes in posture angle caused by factors such as the runner's body swing and foot landing method; for example, during running, the runner's upper body naturally swings with each stride, resulting in a posture angle ranging from ± Fluctuations within a range; These represent the process noise standard deviations of the rate of change of the body's center of gravity position and posture angle, respectively. These are the random fluctuations in the rate of change of the center of gravity position and posture angle caused by factors such as changes in muscle strength and uneven ground reaction force during running. For example, when a runner accelerates or decelerates, due to the gradual change in muscle strength, the rate of change of the center of gravity position may fluctuate within ±0.1 m / s, while the rate of change of the posture angle may vary within ±3 degrees / s.

[0036] In one technical solution of the present invention, , This reflects the expected noise level of the sensor measurements. The accelerometer has lower noise, while the angular velocity sensor has higher noise. This configuration ensures that the Kalman filter module can accurately estimate the runner's posture stability when processing sensor data, thereby improving the system's safety and reliability.

[0037] In this invention, the posture stability assessment module is used to construct a posture stability assessment model based on the predicted state vector of the runner, calculate the runner's posture stability, and issue an early warning if the runner's posture stability is less than the posture stability threshold. In this invention, the posture stability threshold can be set to 0.7.

[0038] This invention's posture stability assessment model comprehensively evaluates stability using six indicators: center of gravity position stability, center of gravity change rate stability, posture angle stability, posture angle change rate stability, treadmill speed stability, and treadmill incline stability. This multi-dimensional approach reflects the runner's motion state and the influence of external conditions, improving the accuracy of runner stability assessment, identifying potential risks in advance, providing a scientific basis for early warning and proactive intervention, and significantly enhancing the system's reliability and practicality. The construction process of this posture stability assessment model is as follows:

[0039] in, Indicates the current time Postural stability of runners Indicates the current time The stability of the runner's center of gravity position , Indicates the current time The predicted runner's center of gravity coordinates in the runner's state vector. This represents the coordinates of the runner's center of gravity in the initial measurement. Indicated Weight; Indicates the current time Stability of runner's posture angle , Indicates the current time The predicted runner's posture angle values ​​in the runner's state vector. express The weights; Indicates the current time treadmill speed stability , Indicates the current time The speed value when getting off the treadmill This represents the average speed of the treadmill over the past several seconds. express The weights; Indicates the current time Treadmill incline stability , Indicates the current time The incline value of the treadmill. This indicates the average incline value of the treadmill over the past several seconds. express The weights; Indicates the current time Stability of the rate of change of the center of gravity of the runner , Indicates the current time The predicted rate of change of the runner's center of gravity in the runner's state vector. express The weights; , Indicates the current time The predicted rate of change of the runner's posture angle in the runner's state vector. express The weight.

[0040] In this invention, the data storage module stores the raw data collected by the sensors, the attitude vector predicted by the Kalman filter module, the attitude stability evaluation results calculated by the attitude stability evaluation module, and early warning records. The stored data adopts a relational database table structure. For example, the sensor data table contains fields such as timestamp, sensor type, and sensor value, and the record is "2025-06-13 10:30:25.000, Accelerometer, 0.01g"; the attitude information table contains fields such as timestamp, body center of gravity position, and attitude angle, for example, "2025-06-13 10:30:25.000, 0.55m, 0.35m, 5°"; the early warning record table contains fields such as early warning time, runner ID, treadmill number, attitude stability index value, and early warning level, for example, "2025-06-13 10:30:25, 123456789, R001, 0.607, Level 1 Early Warning".

[0041] In this invention, the digital twin constructs a dynamic virtual body of the runner based on the predicted state vector of the runner, visualizes it, and renders the runner's posture stability within the dynamic virtual body. Specifically: Based on the runner's basic body parameters entered during registration—height, weight, and limb proportions—a personalized initial virtual body model is generated in a digital twin. This model is then used to measure the runner's initial center of gravity position. and initial attitude angle measurement value Calibrate the virtual body's initial posture to ensure it matches the actual initial state vector measurement of the runner; Every 0.01 seconds, the digital twin receives the runner's state vector predicted by the Kalman filter module, driving the virtual body's torso center of gravity to update its position and posture angle in three-dimensional space. The rate of change of center of gravity and the rate of change of posture angle are used to smoothly transition the virtual body's movements and avoid visual distortion. Through a 3D rendering engine, the virtual body's posture evolution process is displayed in real time on the treadmill screen, and the movement trajectory of the center of gravity within the past second is marked with a dashed line. At the same time, the digital twin receives the posture stability calculated from the posture stability assessment model. If the posture stability is less than the stability safety threshold, it is marked as high risk, and the visualization environment is rendered in red. Based on the current tilt angle, the possible direction of fall is indicated to remind the user to pay attention to safety.

[0042] In this invention, the prediction module uses an LSTM model. Based on the runner's state vector predicted by the Kalman filter module, it predicts the runner's posture stability at the next moment, achieving early prediction of the runner's posture stability and improving running safety. Figure 3 The prediction module uses the runner's state vector predicted by the Kalman filter module in the past as the input of the prediction module, and uses the posture stability of the corresponding runner's state vector calculated by the posture stability evaluation module as the label of the prediction module. The prediction module is trained until the mean squared error loss function of the predicted posture stability and the corresponding label converges, thus completing the training of the prediction module.

[0043] The mean square error loss function in this invention Represented as:

[0044] in, This represents the number of samples used to train the prediction module. express index, This indicates the first prediction made by the prediction module. The pose stability of each sample Indicates the first The attitude stability of each sample was calculated using the attitude stability assessment module.

[0045] In one technical solution of the present invention, an early warning module is also included. As the core execution unit for system risk response, it needs to be deeply linked with the attitude stability assessment module and the prediction module. Based on the real-time risk and future trend judgment results, it performs graded early warning and collaborative intervention. The specific process is as follows: (1) Current risk trigger: When the real-time comprehensive risk level output by the attitude stability assessment module is "high risk", that is, when the current attitude stability is < attitude stability threshold, an early warning is immediately activated; (2) Future trend trigger: When the prediction module determines that it is "high risk", that is, when the predicted attitude stability at the next moment is less than the attitude stability threshold, an early warning is immediately activated.

[0046] This invention introduces Kalman filtering algorithm and sensor fusion technology to achieve accurate monitoring of runner's posture stability, and promptly alerts runners and administrators when posture stability falls below a safe threshold, thereby effectively preventing runners from falling due to posture instability and ensuring their safety on the treadmill.

[0047] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A smart early warning system for human posture stability during running, characterized in that, include: The module includes a sensor module, an information preprocessing module, an edge processing module, a Kalman filter module, and an attitude stability evaluation module. The sensor module is used to collect the runner's motion signals and pressure sensing signals; The information preprocessing module is used to perform calibration and filtering preprocessing on the collected motion signals and pressure sensing signals; The edge processing module is used to map the pre-processed pressure sensing signal into the runner's center of gravity position and the rate of change of the runner's center of gravity position, and to map the pre-processed motion signal into the runner's posture angle and the rate of change of the runner's posture angle, forming the runner's state vector measurement value. The Kalman filter module is used to predict the runner's state vector based on the runner's state vector measurement value; The posture stability assessment module is used to construct a posture stability assessment model based on the predicted state vector of the runner, calculate the posture stability of the runner, and issue an early warning if the posture stability of the runner is less than the posture stability threshold.

2. The intelligent early warning system for human posture stability during running according to claim 1, characterized in that, The runner's state vector measurement value ,in, express The x-axis measurement of the runner's center of gravity at that moment. express The vertical coordinate measurement of the runner's center of gravity at that moment. express Measurements of the runner's posture angles at any given time. express Measurement of the rate of change of the runner's center of gravity at any given moment. express Measurement of the vertical rate of change of the runner's center of gravity at any given time. express The measured values ​​of the runner's posture angle and velocity at any given moment. This indicates the transpose operation.

3. The intelligent early warning system for human posture stability during running according to claim 2, characterized in that: in, Indicates the first in the pressure array The coordinates of each unit, Indicates the first in the pressure array Unit The corresponding pressure value at that moment. express The triaxial acceleration of the runner's torso in the motion signal at the given moment. express The runner's torso in the motion signal at that moment axial angular velocity, Indicates the sampling time. This indicates the weight associated with angular velocity.

4. The intelligent early warning system for human posture stability during running according to claim 1, characterized in that, The specific process by which the Kalman filter module predicts the runner's state vector based on the runner's state vector measurement is as follows: i: Based on the motion signals and pressure sensing signals of the runner standing still on the treadmill, set the initial state vector measurement values. And set the initial state vector estimation error covariance matrix. State vector process noise covariance matrix and state vector measurement noise covariance matrix ; ii: Based on the runner's state vector predicted from the previous moment and state transition matrix Estimate the state vector at the current time step. And combine the state vector from the previous time step to estimate the error covariance matrix. The state vector and process noise covariance matrix of the previous time step The covariance matrix of the prediction error for the current state vector is calculated. ; iii: Estimate the error covariance matrix based on the predicted state vector at the current moment. Combine the state vector from the previous time step with the noise covariance matrix. Calculate Kalman gain ,in, The measurement matrix represents the state vector at the current moment; iv: Based on the estimated state vector at the current time The current state vector measurement value and Kalman gain Predict the runner's state vector at the current moment. ; v: Measurement matrix based on the current state vector. and the estimated state vector at the current time. Calculate the state vector to estimate the measured value Based on the measured value of the state vector at the current moment With state vector estimation measurement Residual vector update , Repeat steps ii-iv to continue using the state vector of the runner at the next moment.

5. The intelligent early warning system for human posture stability during running according to claim 4, characterized in that, Measured value based on the current state vector With state vector estimation measurement Residual vector update , The specific process is as follows: If the mean square value of the residual vector is greater than 0.0003 for several consecutive iterations, the state vector measurement noise covariance matrix will be... Increase the scaling factor; if the mean square value of the residual vector is less than 0.0001 for several consecutive iterations, decrease the scaling factor of the state vector measurement noise covariance matrix; otherwise, ; Calculate the rate of change of the runner's center of gravity in the residual vector at the current moment. ,like The state vector process noise covariance matrix The standard deviation of process noise related to the rate of change of the runner's center of gravity is increased; if The state vector process noise covariance matrix Reduce the standard deviation of process noise related to the rate of change of the runner's center of gravity; if , ;in, Indicates the current time Measurement of the rate of change of the runner's center of gravity horizontally. Indicates the current time Estimates of the rate of change of the center of gravity of runners at different levels. Indicates the current time Measurement of the vertical rate of change of the runner's center of gravity. Indicates the current time Estimated rate of vertical change of the center of gravity of a runner.

6. The intelligent early warning system for human posture stability during running according to claim 1, characterized in that, The process of constructing the attitude stability evaluation model is as follows: in, Indicates the current time Postural stability of runners Indicates the current time The stability of the runner's center of gravity position , Indicates the current time The predicted runner's center of gravity coordinates in the runner's state vector. This represents the coordinates of the runner's center of gravity in the initial measurement. Indicated Weight; Indicates the current time Stability of runner's posture angle , Indicates the current time The predicted runner's posture angle values ​​in the runner's state vector. express The weights; Indicates the current time treadmill speed stability , Indicates the current time The speed value when getting off the treadmill This represents the average speed of the treadmill over the past several seconds. express The weights; Indicates the current time Treadmill incline stability , Indicates the current time The incline value of the treadmill. This indicates the average incline value of the treadmill over the past several seconds. express The weights; Indicates the current time Stability of the rate of change of the center of gravity of the runner , Indicates the current time The predicted rate of change of the runner's center of gravity in the runner's state vector. express The weights; , Indicates the current time The predicted rate of change of the runner's posture angle in the runner's state vector. express The weight.

7. The intelligent early warning system for human posture stability during running according to claim 1, characterized in that, It also includes a digital twin, which constructs a dynamic virtual representation of the runner based on the predicted state vector of the runner, visualizes it, and renders the runner's posture stability in the dynamic virtual representation.

8. The intelligent early warning system for human posture stability during running according to claim 1, characterized in that, It also includes a prediction module, which uses an LSTM model to predict the runner's posture stability at the next moment based on the runner's state vector predicted by the Kalman filter module.

9. The intelligent early warning system for human posture stability during running according to claim 8, characterized in that, The prediction module is trained by using the runner's state vector predicted by the Kalman filter module in the past as the input of the prediction module, and the posture stability of the corresponding runner's state vector calculated by the posture stability evaluation module as the label of the prediction module. The training of the prediction module is completed until the mean squared error loss function of the predicted posture stability and the corresponding label converges.

10. The intelligent early warning system for human posture stability during running according to claim 9, characterized in that, The mean square error loss function Represented as: in, This represents the number of samples used for training the prediction module. express index, This indicates the first prediction made by the prediction module. The pose stability of each sample Indicates the first The attitude stability of each sample was calculated using the attitude stability assessment module.

Images