Personalized exercise rehabilitation risk early warning method based on IMU dynamic instability evaluation

Abstract

The invention relates to a personalized exercise rehabilitation risk early warning method and device based on IMU dynamic instability assessment, and the method comprises the steps: collecting multi-dimensional exercise data of a subject, carrying out the posture fusion, and obtaining the limb posture, joint angle change rate and angular momentum of the subject; performing time sequence mode decomposition to extract a single gait cycle and a specific jump landing attitude; calculating a biomechanical feature vector of each motion primitive; inputting the single gait cycle, the specific jump landing attitude and the biomechanical feature vector into a hybrid injury induced motion model, and outputting a multi-dimensional injury risk index vector; and comparing the multi-dimensional injury risk index vector with historical health data and a moving target dynamic adjustment threshold value of the subject, and if the threshold value is exceeded, providing real-time auditory, tactile or visual feedback to the subject. According to the method, the influence of movement speed and rhythm change is overcome, dynamic instability in movement can be more sensitively reflected, and personalized risk assessment in a real sense is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of sports medicine AI assistance, in particular to a personalized sports rehabilitation risk warning method and device based on IMU dynamic instability evaluation and computer equipment. BACKGROUND

[0002] Sports injuries not only affect the quality of life of individuals, but also increase the medical burden. Traditional sports injury risk assessment mainly relies on professional equipment in a laboratory environment (such as three-dimensional motion capture systems, force platforms, etc.), which has high precision, but is bulky, expensive, and needs to be performed in a specialized laboratory, and cannot be applied to daily training or competition scenes. Many subtle abnormalities that occur in actual sports and may lead to injuries cannot be discovered and intervened in a timely manner. In addition, although existing wearable devices (such as smart bracelets, smart watches) can provide basic sports data such as steps and heart rate, they have limited depth analysis capabilities for sports biomechanical characteristics, and cannot effectively capture key injury indicators such as joint stress and stability. Even some devices equipped with IMU, mostly stay in simple pose estimation or gait counting, lack the ability to quantitatively evaluate potential injury risks. For example, some IMU-based systems identify abnormal gait, but fail to further analyze biomechanical characteristics such as joint impact and moment change rate that are directly related to injury. Since existing risk assessment often uses static thresholds, it fails to fully consider individual differences (such as bone density, muscle strength, and history of injury) and dynamic changes in sports goals (such as competitive training, rehabilitation, and fall prevention), and cannot accurately guide the movement behavior of different groups of people, leading to overprotection or risk omission.

[0003] To solve the above problems, the present application provides a personalized sports rehabilitation risk warning method based on IMU dynamic instability evaluation, which decomposes and extracts single gait cycle and specific jump landing gesture motion primitives from continuous motion data to overcome the influence of movement speed and rhythm changes. The motion stability index is calculated by an enhanced phase space recursive quantitative analysis method to more sensitively reflect the dynamic instability in movement. According to the biomechanical load tolerance baseline of the subject's bone density DEXA data and serum collagen metabolism marker level, a truly personalized risk assessment is achieved. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a personalized sports rehabilitation risk warning method and device based on IMU dynamic instability evaluation and computer equipment to solve the above problems of the prior art.

[0005] According to one aspect of the present application, a personalized sports rehabilitation risk warning method based on IMU dynamic instability evaluation is provided, comprising:

[0006] The multi-dimensional motion data of the subject is collected in real time by a plurality of IMU sensor arrays, and the multi-dimensional motion data is fused according to a generalized Kalman filter and a nonlinear complementary filter to obtain the limb posture, joint angle rate of change and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data;

[0007] The limb posture, joint angle rate of change and angular momentum of the subject are decomposed into time series patterns according to a non-negative matrix factorization method and a dynamic time warping algorithm, and a single gait cycle and a specific jump landing posture are extracted;

[0008] The biomechanical feature vector of each motion primitive is calculated, and the biomechanical feature vector includes a joint impact peak value, a joint torque rate of change, a motion stability index and a specific frequency band energy distribution feature based on wavelet packet decomposition;

[0009] The single gait cycle, the specific jump landing posture and the biomechanical feature vector are input into a trained hybrid injury-induced motion model, and a multi-dimensional injury risk index vector is output, which includes overload risk, fatigue damage accumulation risk and acute fall or sprain risk of the knee joint and ankle joint;

[0010] The multi-dimensional injury risk index vector is compared with historical health data of the subject and a dynamic adjustment threshold of a motion target, and if the threshold is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through a wearable device or an associated mobile application.

[0011] In an optional manner, the formula for calculating the joint torque is:

[0012]

[0013] wherein, is a mass matrix; is a joint angle; is a joint angular acceleration; is a joint angular velocity; is an IMU torque estimation confidence factor; is an error variance between IMU torque estimation and laboratory reference data; is a force based on IMU estimation; is plantar pressure data.

[0014] In an optional manner, the motion stability index is enhanced by a phase space recursive quantitative analysis method, and the enhanced expression is:

[0015]

[0016] wherein, is the enhanced motion stability index; is the total amount of motion data samples; is the sampling time interval is the indicator function; is the time is the joint angle state of the motion trajectory at the time instant; is the time is the joint angle state of the motion trajectory at the time instant; is the recursive neighborhood threshold; is the maximum Lyapunov exponent; is the time is the local divergence rate of the motion trajectory at the time instant.

[0017] In an optional manner, the mixed injury-induced motion model comprises a biomechanics simulation layer and a clinical pathology knowledge graph;

[0018] The biomechanics simulation layer is a musculoskeletal multibody dynamics simulation layer of OpenSim, which is used to map IMU data into muscle activation and ligament strain energy.

[0019] The clinical pathology knowledge graph comprises an injury rule reasoning engine, which is used to mark a combination state of a knee anterior cruciate ligament strain energy greater than a first preset threshold and a tibial anterior shear force greater than a second preset threshold as ACL tear high risk, and associate cartilage wear positioning information in a historical health database, wherein, when the medial meniscus contact pressure peak value lasts more than a third preset threshold, an osteoarthritis progression warning is activated.

[0020] In an optional manner, the method further comprises:

[0021] Establishing a baseline of biomechanics load tolerance based on subject bone density DEXA data and serum collagen metabolism marker levels;

[0022] Adjusting the threshold in real time according to the motion target, wherein, in the competitive training mode, the threshold is 90% of the standard threshold, in the rehabilitation mode, the threshold is 60%, and in the elderly fall prevention mode, a gait symmetry weight coefficient is introduced; wherein, by the correlation between the muscle oxygen saturation decrease rate and the joint torque fluctuation entropy value, when the muscle oxygen recovery slope is lower than 0.25% / s, the motion intensity is forcibly locked by the wearable device.

[0023] In an optional manner, the method further comprises:

[0024] When the ankle varus angular velocity is detected to be greater than 500° / s and the center of gravity projection deviates from the support base by 60%, a micro-current functional electrical stimulation is triggered to enhance the contraction of the peroneal muscle group;

[0025] According to the risk index vector, a virtual center of gravity trajectory guide line is projected in the AR glasses, so that the foot bottom pressure center of the user is kept within the trajectory tolerance band when landing;

[0026] Among them, for patients after anterior cruciate ligament reconstruction, the knee flexion angle is prompted by bone conduction earphones at the initial stage of take-off phase.

[0027] In an optional manner, the posture fusion of the multi-dimensional motion data of the subject according to the generalized Kalman filter and the nonlinear complementary filter further comprises:

[0028] The three-axis acceleration and three-axis angular velocity data are preprocessed and state estimated by the generalized Kalman filter to obtain an initial limb posture;

[0029] The three-axis magnetic field data and the limb posture are fused according to the nonlinear complementary filter to correct the limb posture by the magnetic field and obtain a fused limb posture;

[0030] The relative posture between adjacent IMU sensors is calculated according to the fused limb posture to obtain the joint angle of the subject, and the time derivative operation is performed on the joint angle to obtain the joint angle change rate;

[0031] The angular momentum of the subject is calculated according to the fused limb posture, the limb mass distribution model and the joint angle change rate.

[0032] In an optional manner, the time sequence pattern decomposition of the limb posture, joint angle change rate and angular momentum of the subject according to the non-negative matrix factorization method and dynamic time warping algorithm further comprises:

[0033] The limb posture, joint angle change rate and angular momentum data of the subject are constructed into a multi-dimensional time sequence feature matrix;

[0034] The multi-dimensional time sequence feature matrix is decomposed by the non-negative matrix factorization method to obtain a base mode matrix and a corresponding weight coefficient matrix of the gait cycle and the jumping action;

[0035] The base mode matrix obtained by decomposition and the real-time collected motion data are matched and aligned by the dynamic time warping algorithm to correct the changes of motion execution speed and rhythm, and a single complete gait cycle and a specific jumping landing posture are extracted from continuous data;

[0036] The preset gait event detection algorithm and the jump event detection algorithm ensure the accuracy of the extracted motion primitives.

[0037] According to another aspect of the present application, a personalized motion rehabilitation risk warning device based on IMU dynamic instability evaluation is provided, comprising:

[0038] The data acquisition and attitude fusion module is configured to acquire multi-dimensional motion data of the subject in real time through a plurality of IMU sensor arrays, fuse the multi-dimensional motion data according to a generalized Kalman filter and a nonlinear complementary filter, and obtain the limb attitude, joint angle change rate and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data.

[0039] The time series pattern decomposition module is configured to perform time series pattern decomposition on the limb attitude, joint angle change rate and angular momentum of the subject according to a non-negative matrix decomposition method and a dynamic time warping algorithm, and extract a single gait cycle and a specific jump landing attitude.

[0040] The biomechanical feature calculation module is configured to calculate a biomechanical feature vector of each motion primitive, wherein the biomechanical feature vector includes a joint impact peak value, a joint torque change rate, a motion stability index and a specific frequency band energy distribution feature based on wavelet packet decomposition.

[0041] The injury risk assessment module is configured to input the single gait cycle, the specific jump landing attitude and the biomechanical feature vector into a trained hybrid injury-inducing motion model, and output a multi-dimensional injury risk index vector, wherein the multi-dimensional injury risk index vector includes overload risk, fatigue injury accumulation risk and acute fall or sprain risk of the knee joint and ankle joint.

[0042] The risk feedback and warning module is configured to compare the multi-dimensional injury risk index vector with historical health data of the subject and a dynamically adjusted threshold of a motion target, and if the threshold is exceeded, provide real-time auditory, tactile or visual feedback to the subject through a wearable device or an associated mobile application.

[0043] According to still another aspect of the present application, a computer device is provided, comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus.

[0044] The memory is configured to store at least one executable instruction, and the executable instruction causes the processor to perform the operations corresponding to the above-mentioned personalized motion rehabilitation risk warning method based on IMU dynamic instability evaluation.

[0045] According to the scheme provided by the application, multi-dimensional motion data of a subject is collected in real time by a plurality of IMU sensor arrays, and the multi-dimensional motion data is fused for posture by a generalized Kalman filter and a nonlinear complementary filter to obtain the limb posture, joint angle change rate and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data; the limb posture, joint angle change rate and angular momentum of the subject are decomposed for timing mode by a non-negative matrix factorization method and a dynamic time warping algorithm, and a single gait cycle and a specific jump landing posture are extracted; a biomechanical feature vector of each motion primitive is calculated, the biomechanical feature vector including a joint impact peak value, a joint torque change rate, a motion stability index and a specific frequency band energy distribution feature based on wavelet packet decomposition; the single gait cycle, the specific jump landing posture and the biomechanical feature vector are input into a trained hybrid injury-induced motion model to output a multi-dimensional injury risk index vector, the multi-dimensional injury risk index vector including overload risk, fatigue injury accumulation risk and acute fall or sprain risk of the knee joint and ankle joint; the multi-dimensional injury risk index vector is compared with historical health data of the subject and a motion target dynamic adjustment threshold value, and if the threshold value is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through a wearable device or an associated mobile application. The application extracts a single gait cycle and a specific jump landing posture motion primitive from continuous motion data, overcoming the influence of motion speed and rhythm changes. The motion stability index is calculated by an enhanced phase space recursive quantitative analysis method, which can more sensitively reflect the dynamic instability in motion and predict the risk of falling or spraining. The biomechanics simulation layer maps the IMU data to muscle activation and ligament strain energy, enhancing the injury interpretability. Based on the baseline of biomechanics load tolerance of the subject's bone density DEXA data and serum collagen metabolism marker level, a truly personalized risk assessment is realized.

[0046] The above description is only a summary of the technical scheme of the application. In order to enable one skilled in the art to better understand the technical means of the application, the contents of the specification can be implemented, and in order to enable the above and other purposes, features and advantages of the application to be more obvious and easy to understand, the specific embodiments of the application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0047] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not meant to limit the present application. Moreover, the same reference numerals in the attached drawings indicate the same or similar components. In the drawings:

[0048] Figure 1A flowchart of a personalized exercise rehabilitation risk early warning method based on IMU dynamic instability evaluation is shown.

[0049] Figure 2 A frame diagram of a personalized exercise rehabilitation risk early warning device based on IMU dynamic instability evaluation is shown.

[0050] Figure 3 A structural diagram of a computer device is shown. DETAILED DESCRIPTION

[0051] Exemplary embodiments of the present application will be described in detail with reference to the drawings. Although exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be accurately conveyed to those skilled in the art.

[0052] Figure 1 A flowchart of a personalized exercise rehabilitation risk early warning method based on IMU dynamic instability evaluation is shown. Specifically, as shown in Figure 1 the following steps are included:

[0053] In step S101, multi-dimensional motion data of a subject is collected in real time by a plurality of IMU sensor arrays, and the multi-dimensional motion data is fused for posture according to a generalized Kalman filter and a nonlinear complementary filter to obtain limb posture, joint angle rate of change and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data.

[0054] In this embodiment, three-axis acceleration is used for the estimation of the gravity direction (in static or low dynamic conditions) and to identify motion impacts and acceleration changes. Three-axis angular velocity directly measures angular motion, which is used to describe the speed of rotational motion. Three-axis magnetic field data corrects the heading drift problem caused by the error accumulation of accelerometers and gyroscopes in the process of pose fusion, especially in long-time motion or complex motion paths, and the magnetic field data provides an absolute attitude reference to enhance the robustness of the attitude estimation. IMU sensors are small in size, light in weight, low in power consumption, and easy to integrate into wearable devices, making data acquisition possible in natural and real motion environments without the need for cumbersome laboratory equipment. Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF) fuses noise data from different sensors through iterative updates of state estimation and error covariance matrix, thereby obtaining more accurate attitude estimation. The attitude obtained by integrating gyroscope data will accumulate drift errors over time, and Kalman filter uses accelerometers (which provide the direction of gravity in quasi-static conditions) and magnetometers (which provide the direction of the earth's magnetic field) as observations to correct the integration errors of gyroscopes, thereby significantly improving the long-term stability of attitude estimation. As a pose fusion method, the Nonlinear Complementary Filter (NCF) processes accelerometer and magnetometer data through a low-pass filter (to remove high-frequency noise and motion interference) and processes gyroscope data through a high-pass filter (to remove drift), and then fuses the two to make the filter balance dynamic response and long-term stability. In some application scenarios, especially in real-time systems with limited computing resources or requiring extremely low latency, NCF serves as an alternative or supplement to EKF, providing stable attitude estimation while the computational overhead is usually less than EKF / UKF. In this embodiment, both filters are used, and optimization strategies are adopted in different stages or for different IMU types, or the output of the Kalman filter is further smoothed and corrected using the fast response characteristics of the complementary filter to obtain a more stable and accurate attitude.

[0055] In an alternative way, the pose fusion of the multi-dimensional motion data according to the extended Kalman filter and the nonlinear complementary filter to obtain the limb pose, the joint angular rate of change, and the angular momentum of the subject further comprises:

[0056] The three-axis acceleration and three-axis angular velocity data are preprocessed and state-estimated by the extended Kalman filter to obtain an initial limb pose;

[0057] The limb pose is corrected by the magnetic field according to the fusion of the three-axis magnetic field data and the limb pose by the nonlinear complementary filter to obtain a fused limb pose;

[0058] According to the fused limb posture, the relative posture between adjacent IMU sensors is calculated, the joint angle of the subject is obtained, and the time derivative operation is performed on the joint angle to obtain the joint angle change rate;

[0059] According to the fused limb posture, the limb mass distribution model and the joint angle change rate, the angular momentum of the subject is calculated.

[0060] In this embodiment, the absolute posture of each limb (relative to the global coordinate system) is obtained through posture fusion, and then the joint angle is calculated through the relative relationship of adjacent limb postures, and the time derivative of the joint angle is obtained to obtain the joint angle change rate, which directly reflects the speed of joint movement. Combined with the limb mass distribution model and the joint angle change rate, the angular momentum is calculated, wherein the angular momentum measures the inertia of the rotating motion of the object, and is used to evaluate the motion stability, the fall risk and the transmission of impact force.

[0061] For example, when the athlete jumps up in the air, the accelerometer of the right shank IMU measures the gravitational acceleration, and the gyroscope measures the angular velocity of the shank rotation. The EKF estimates the initial 3D pose of the right shank relative to the global coordinate system (the ground) by constantly predicting and correcting, even if the athlete rotates in the air, the EKF can better track the changes in the pose of the shank. After the EKF obtains the initial pose, the nonlinear complementary filter fuses the magnetometer data considering the influence of the geomagnetic field. If there are magnetic objects such as iron stands near the training site, the local magnetic field will be distorted, and the magnetometer data may be temporarily inaccurate. The complementary filter gradually combines the magnetometer data to correct the yaw angle while believing in the short-term accuracy of the gyroscope, ensuring that the heading estimate of the shank does not drift significantly when there is no strong magnetic interference. Through the above pose fusion process, the accurate pose of the right thigh IMU and the right shank IMU in the global coordinate system is obtained, and the relative rotation between the right thigh and the shank is calculated. For example, if the thigh rotates forward by a certain angle, and the shank also rotates forward by a certain angle, the flexion angle of the knee joint is a component of the difference between the two limb poses. Convert the relative pose to Euler angles to get the flexion / extension angle, the inversion angle, and the rotation angle of the right knee joint. For example, during landing impact, the flexion angle of the knee joint may rapidly increase from 10 degrees to 90 degrees. Differentiate the change of the flexion angle of the knee joint with time to get the knee flexion angular velocity. The angular velocity will be very large in a short time of landing impact, indicating that the knee joint is rapidly absorbing energy. If the angular velocity is too high, the impact load is too large. The limb model estimates the mass, center of mass position, and moment of inertia of each limb (trunk, thigh, shank, foot) of the athlete according to the athlete's height and weight. Combine the pose, center of mass velocity, and angular velocity of each limb to calculate the linear momentum and angular momentum of each limb. Then add up the linear momentum and angular momentum of all limbs to get the total linear momentum and angular momentum of the athlete. During the jump landing process, the athlete needs to dissipate the vertical and horizontal angular momentum through body posture adjustment to maintain body balance. If there is still a large uncontrolled angular momentum in the body (such as rapid rotation of the trunk while the feet are already on the ground) during landing, it may increase the risk of ankle or knee sprain. By monitoring the total angular momentum and its rate of change in real time, the stability of the athlete's landing can be evaluated.

[0062] In step S102, the limb pose, joint angle rate of change, and angular momentum of the subject are decomposed into time series patterns according to the non-negative matrix factorization method and the dynamic time warping algorithm, and single gait cycles and specific jump landing poses are extracted.

[0063] In this embodiment, the non-negative matrix factorization (NMF) can automatically learn and extract base patterns such as gait and jump landing from complex and continuous motion data, and can effectively represent the motion characteristics of different individuals. Dynamic time warping (DTW) solves the problem that the motion duration, speed and rhythm of the same motion performed by different individuals or the same subject at different times may not be consistent, can find the best alignment path between two time series, can accurately compare and match even if there is stretching or compression on the time axis, and can accurately identify and extract complete gait cycles and jump landing postures regardless of the speed of execution.

[0064] For example, when the athlete starts running, continuously perform DTW matching of real-time posture, angular velocity and angular momentum data stream with NMF base patterns of standard running gait cycle, find the most matching segment in the data stream with the base pattern, even if the athlete's running speed is fast or slow. Combined with gait event detection (such as heel strike), the complete single gait cycle can be accurately segmented each time. When the athlete performs jump training (such as vertical jump) and lands, identify the peak value of vertical acceleration and immediately perform DTW matching of the landing phase posture, angular velocity and angular momentum data with the NMF base patterns such as knee flexion buffer pattern during jump landing and rigid knee posture during jump landing, which can accurately extract the specific jump landing posture of each jump.

[0065] In an optional manner, the time sequence pattern decomposition of the limb posture, joint angle rate of change and angular momentum of the subject according to the non-negative matrix factorization method and the dynamic time warping algorithm to extract single gait cycle and specific jump landing posture further comprises:

[0066] The limb posture, joint angle rate of change and angular momentum data of the subject are constructed into a multi-dimensional time sequence feature matrix;

[0067] The multi-dimensional time sequence feature matrix is decomposed by the non-negative matrix factorization method to obtain a base pattern matrix and a corresponding weight coefficient matrix of gait cycle and jump action;

[0068] The base pattern matrix obtained by decomposition is matched and aligned with the real-time collected motion data by the dynamic time warping algorithm to correct the changes of motion execution speed and rhythm, and to extract single complete gait cycle and specific jump landing posture from continuous data;

[0069] The preset gait event detection algorithm and jump event detection algorithm are used to ensure the accuracy of the extracted motion primitives.

[0070] In this embodiment, for example, a soccer player wears IMUs for daily training, and needs to evaluate his gait stability in each run and knee risk in each jump landing. The player wears 6 IMU sensors on the left and right shins, thighs, torso, arms. The IMUs transmit data to the processing unit in real time. After pose fusion, the pose of each limb is obtained. The flexion and extension angles, inversion and eversion angles, and their change rates of the left and right knee joints, the dorsiflexion, plantarflexion angles, inversion and eversion angles, and their change rates of the left and right ankle joints, the pitch, roll, yaw angles of the torso, and the angular momentum of the whole body and each major limb are calculated. The above data is constructed into a multi-dimensional time series feature matrix on the time axis (for example, every 10 milliseconds, containing about 50 dimensions). A large number of normal gait and typical jump landing (such as single or double leg jump, lateral jump) data of athletes are used for NMF training in advance. After NMF decomposition, a set of gait base patterns and a set of jump base patterns are obtained. For example, the gait base pattern contains the characteristics of the movement stages such as the initial support, the middle support, and the swing period. The jump base pattern contains the characteristics of the air period, the ground impact period, and the buffer stability. When the player runs on the field, his IMU data is received in real time. A sliding window (such as 5 seconds) of real-time motion data is taken, which is matched with the learned gait base pattern by DTW. DTW finds the best alignment path. Suppose at a certain time point, DTW finds that a segment of real-time data is highly similar and well aligned with the gait base pattern. According to the alignment result of DTW, this segment of single gait cycle is accurately extracted from the continuous data stream. For the just extracted single gait cycle, the preset gait event detection algorithm is called to verify it. Check whether there are obvious foot heel touch-down event and toe take-off event in this period, and confirm whether the logical order and time interval meet the definition. If the player performs a jump action, the jump landing event is preliminarily identified by detecting the vertical acceleration peak value of the foot IMU and the subsequent rapid change of the joint angle. Then, the motion data is matched and aligned with the jump landing base pattern, and it is confirmed whether it contains complete ground impact and buffer stability stage. For example, if the ankle or knee joint appears a specific angle change rate feature during landing and the DTW matching degree is high, it is confirmed to be a complete jump landing action and the landing pose data is accurately extracted from the touch-down instant.

[0071] In step S103, the biomechanical feature vector of each movement base element is calculated, which includes joint impact peak, joint torque change rate, movement stability index, and specific frequency band energy distribution feature based on wavelet packet decomposition.

[0072] In this embodiment, the joint impact peak reflects the instantaneous load size of the joint during the movement process, and evaluates the risk of acute injury (such as bone contusion, cartilage injury) and fatigue injury (such as stress fracture, arthritis progression). The joint torque change rate characterizes the dynamics and change speed of joint load, and too high torque change rate may mean insufficient muscle control, uncoordinated movement pattern, thereby increasing the risk of joint ligament, tendon injury, especially rotational injury. The motion stability index is enhanced by introducing a phase space recurrence quantitative analysis method to evaluate the repeatability and predictability of the movement trajectory. The decline of MSI usually indicates the decline of neuromuscular control, fatigue accumulation or injury precursor, and is an important predictor of the risk of falls and sprains. The energy distribution characteristics of specific frequency bands based on wavelet packet decomposition are difficult to capture by traditional time or frequency domain analysis. Wavelet packet decomposition can decompose the signal into different frequency subbands, thereby identifying weak signal characteristics related to specific movement patterns, tissue vibrations or abnormal movements (such as tremor, stiffness). For example, high-frequency vibrations may be related to muscle stiffness or joint friction, while specific low-frequency energy may reflect abnormalities in gait patterns, which can help to find early and hidden injury signals.

[0073] In an alternative way, the formula for calculating the joint torque is:

[0074]

[0075] wherein, is the mass matrix; is the joint angle; is the joint angular acceleration; is the joint angular velocity; is the IMU torque estimation confidence factor; is the error variance between the IMU torque estimation and the laboratory reference data; is the force estimated based on the IMU; is the plantar pressure data.

[0076] In this embodiment, the use of IMU alone to calculate torque is susceptible to cumulative errors and drift. By calibration and correction, the above shortcomings of IMU can be effectively compensated, making it more suitable for real-time and dynamic wearable applications.

[0077] In an alternative way, the motion stability index is enhanced by a phase space recurrence quantitative analysis method, wherein the enhanced expression is:

[0078]

[0079] wherein, is the enhanced motion stability index; is the total amount of motion data samples; is the sampling time interval is the indicator function is the time is the joint angle state of the motion trajectory at the time is the time is the joint angle state of the motion trajectory at the time is the recurrence neighborhood threshold is the maximum Lyapunov exponent is the time is the local divergence rate of the motion trajectory at the time

[0080] In this embodiment, the enhanced MSI reflects the predictability of the system state through the density and pattern of recurrence points. When the motion pattern is abnormal (such as gait instability caused by fatigue or compensatory action), the trajectory in the phase space changes, leading to changes in the recurrence pattern, so that the MSI can more sensitively detect potential signs of instability. A lower motion stability index is usually associated with a decline in motion control ability, a decrease in action pattern efficiency, and a decrease in adaptability to external disturbances, all of which are potential risk factors for motion injury (especially fatigue injury or falling). In the expression, not only the points within the recurrence neighborhood (reflecting repeatability) are considered, but also the maximum Lyapunov exponent (reflecting the overall chaotic divergence trend of the system) and the local divergence rate (reflecting the divergence degree of the local trajectory), so that the MSI can not only assess the overall stability of the motion, but also pay attention to local instability events, and be applied to various types of motion data (including gait, running, and jumping landing).

[0081] In step S104, the single gait cycle, the specific jump landing posture, and the biomechanical feature vector are input into the trained hybrid injury-inducing motion model, and a multi-dimensional injury risk index vector is output, which includes overload risk of the knee joint and ankle joint, fatigue injury accumulation risk, and acute falling or sprain risk.

[0082] In this embodiment, the hybrid injury-inducing motion model includes a biomechanics simulation layer and a clinical pathology knowledge graph.

[0083] The biomechanics simulation layer is a musculoskeletal multibody dynamics simulation layer of OpenSim, which is used to map IMU data into muscle activation force and ligament strain energy.

[0084] The clinical pathology knowledge graph comprises a damage rule reasoning engine, the damage rule reasoning engine is used for marking a combination state that a strain energy of an anterior cruciate ligament of a knee joint is greater than a first preset threshold and a tibial anterior shear force is greater than a second preset threshold as ACL tear high risk; and, the soft cartilage wear positioning information in the historical health database is associated, wherein, when the medial meniscus contact pressure peak value lasts more than a third preset threshold, the osteoarthritis progression early warning is activated.

[0085] In the embodiment, the IMU data is simulated through OpenSim, and is not only simple posture and speed, but is converted into deeper biomechanical indexes such as muscle force and ligament stress. The internal mechanism of sports injury is revealed, instead of only observing the appearance of the injury. For example, directly monitoring the ligament strain energy instead of only the joint angle can more accurately warn the ligament injury. By associating the soft cartilage wear positioning information in the historical health database and combining the condition that the medial meniscus contact pressure peak value lasts more than a third preset threshold, the osteoarthritis progression early warning can be activated. Especially for chronic overuse injuries, the injury can be effectively prevented from worsening.

[0086] In step S105, the multidimensional injury risk index vector is compared with the historical health data of the subject and the dynamic adjustment threshold of the movement target, and if the threshold is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through the wearable device or the associated mobile application.

[0087] In the embodiment, for example, in a certain landing, the strain energy of the anterior cruciate ligament of the knee joint of the athlete reaches 50J (the first preset threshold T1 is 40J) and the tibial anterior shear force is 1500N (the second preset threshold T2 is 1200N). The damage rule reasoning engine immediately triggers the rule to judge that the ACL tear risk of the landing action is extremely high. At the same time, the historical health database of the athlete is queried to find that the medial meniscus has a mild wear record. In multiple high-intensity jumping training, the medial meniscus contact pressure peak value lasts more than 1.5MPa (the third preset threshold T3) and the duration is long, and the osteoarthritis progression early warning is activated. The wearable device immediately warns the athlete through vibration (tactile feedback) and voice prompt (auditory feedback). If the athlete wears AR glasses, a virtual center of gravity trajectory guide line (visual feedback) is projected in the field of view to prompt the athlete to place the center of gravity in the middle of the foot next time.

[0088] In an optional mode, the method further comprises:

[0089] A baseline of biomechanical load tolerance is established based on DEXA data of bone density of the subject and serum collagen metabolism marker levels;

[0090] The threshold value is adjusted in real time according to the moving target, wherein the competitive training mode adopts FIFA11 and the standard threshold value is 90%, the rehabilitation mode adopts 60%, and the gait symmetry weight coefficient is introduced in the elderly fall prevention mode; wherein, through the correlation between the muscle oxygen saturation rate and the joint torque fluctuation entropy value, when the muscle oxygen recovery slope is lower than 0.25% / s, the exercise intensity is forcibly locked through the wearable device.

[0091] In this embodiment, for example, a football player conducts competitive training, the player conducts DEXA scanning before the season, and the bone density is shown to be normal and slightly high; the serum collagen metabolism marker detection result shows that the collagen synthesis ability is strong and the recovery speed is fast. According to this, a higher biomechanical load tolerance baseline is established. The exercise mode is set to competitive training, and the alarm threshold value of key indicators such as the risk of ACL (anterior cruciate ligament) tear of the knee joint and the risk of ankle sprain is set to the standard threshold value of 90%. The IMU data monitors that the joint impact peak value and the knee joint eversion torque approach the threshold value in multiple changes of direction and sprints. At the same time, the armband muscle oxygen sensor shows that the muscle oxygen saturation degree continues to decrease and the subsequent monitored muscle oxygen recovery slope reaches 0.2% / s (lower than 0.25% / s). The wearable earphone immediately issues an alarm: “fatigue overload, suggest pausing and resting”. If the player does not listen and the system judges that the fatigue degree has significantly affected the movement control ability, the smart football shoes connected by it remind the gait instability through the sole micro-vibration, or reduce the training difficulty through the smart goal / training equipment, and force it to rest.

[0092] In an optional mode, the method further comprises:

[0093] When the ankle inversion angular velocity is detected to be greater than 500° / s and the gravity center projection deviates from the support base by 60%, a micro-current functional electrical stimulation is triggered to enhance the contraction of the peroneal muscle group;

[0094] According to the risk index vector, a virtual gravity center trajectory guide line is projected in the AR glasses to make the foot bottom pressure center of the user keep within the trajectory tolerance band when landing;

[0095] Among them, for patients after anterior cruciate ligament reconstruction, the knee flexion angle is prompted to be insufficient at the initial stage of take-off phase through bone conduction earphones.

[0096] In this embodiment, by setting specific threshold values of ankle inversion angular velocity and gravity center deviation, the high-risk state of ankle sprain that is about to occur can be quickly identified. The micro-current functional electrical stimulation (FES) can enhance the contraction of the peroneal muscle group (such as the peroneus longus muscle and the peroneus brevis muscle) in a very short time, which is much better than the traditional passive protection. The AR glasses project a virtual gravity center trajectory guide line, and the user can intuitively adjust the landing posture to ensure that the foot bottom pressure center (COP) remains within the safe area, thereby effectively reducing the impact load of the knee joint and the ankle joint.

[0097] According to the scheme provided in the application, multi-dimensional motion data of a subject is collected in real time by a plurality of IMU sensor arrays, the multi-dimensional motion data is fused according to a generalized Kalman filter and a nonlinear complementary filter to obtain the limb posture, joint angle change rate and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data; the limb posture, joint angle change rate and angular momentum of the subject are decomposed in time sequence mode according to a non-negative matrix factorization method and a dynamic time warping algorithm to extract a single gait cycle and a specific jump landing posture; the biomechanical feature vector of each motion primitive is calculated, the biomechanical feature vector includes a joint impact peak value, a joint torque change rate, a motion stability index and a specific frequency band energy distribution feature based on wavelet packet decomposition; the single gait cycle, the specific jump landing posture and the biomechanical feature vector are input into a trained hybrid injury-induced motion model to output a multi-dimensional injury risk index vector, the multi-dimensional injury risk index vector includes overload risk, fatigue injury accumulation risk and acute fall or sprain risk of the knee joint and ankle joint; the multi-dimensional injury risk index vector is compared with historical health data and motion target dynamic adjustment threshold of the subject, and if the threshold is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through a wearable device or an associated mobile application. The application extracts a single gait cycle and a specific jump landing posture motion primitive from continuous motion data, overcoming the influence of motion speed and rhythm changes. The motion stability index is calculated by an enhanced phase space recursive quantitative analysis method, which can more sensitively reflect the dynamic instability in motion and predict the risk of falling or spraining. The biomechanics simulation layer maps the IMU data to muscle activation and ligament strain energy, enhancing the injury interpretation. Based on the baseline of biomechanics load tolerance of the subject's bone density DEXA data and serum collagen metabolism marker level, the personalized risk assessment in a true sense is realized.

[0098] Figure 2 A framework schematic diagram of the personalized motion rehabilitation risk early warning device based on IMU dynamic instability evaluation of the embodiment of the application is shown. The personalized motion rehabilitation risk early warning device based on IMU dynamic instability evaluation includes:

[0099] The data acquisition and posture fusion module 210 is used for collecting multi-dimensional motion data of a subject in real time by a plurality of IMU sensor arrays, fusing the multi-dimensional motion data according to a generalized Kalman filter and a nonlinear complementary filter to obtain the limb posture, joint angle change rate and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data;

[0100] The time series pattern decomposition module 220 is configured to perform time series pattern decomposition on the limb posture, joint angle rate of change, and angular momentum of the subject according to a non-negative matrix factorization method and a dynamic time warping algorithm, to extract a single gait cycle and a specific jump landing posture.

[0101] The biomechanical feature calculation module 230 is configured to calculate a biomechanical feature vector of each movement primitive, the biomechanical feature vector including a joint impact peak value, a joint torque rate of change, a movement stability index, and a specific frequency band energy distribution feature based on wavelet packet decomposition.

[0102] The injury risk assessment module 240 is configured to input the single gait cycle, the specific jump landing posture, and the biomechanical feature vector into a trained hybrid injury-inducing movement model, to output a multi-dimensional injury risk index vector, the multi-dimensional injury risk index vector including overload risk of the knee joint and the ankle joint, fatigue injury accumulation risk, and acute fall or sprain risk.

[0103] The risk feedback and early warning module 250 is configured to compare the multi-dimensional injury risk index vector with historical health data of the subject and a dynamically adjusted threshold of a movement target, and if the threshold is exceeded, to provide real-time auditory, tactile, or visual feedback to the subject through a wearable device or an associated mobile application.

[0104] Figure 3 A structural schematic diagram of an embodiment of a computer device of the present application is shown, and the embodiment of the present application does not limit the specific implementation of the computer device.

[0105] As shown in Figure 3 the computer device can include a processor 302, a communications interface 304, a memory 306, and a communications bus 308.

[0106] The processor 302, the communications interface 304, and the memory 306 can communicate with each other through the communications bus 308. The communications interface 304 is configured to communicate with network elements such as clients or other servers. The processor 302 is configured to execute the program 310, and specifically can execute the related steps in the above-mentioned IMU-based dynamic instability evaluation personalized movement rehabilitation risk early warning method embodiment.

[0107] Specifically, the program 310 can include program code including computer operation instructions.

[0108] The processor 302 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement one or more embodiments of the present application. The computer device can include one or more processors of the same type or different types, such as one or more CPUs and one or more ASICs.

[0109] The memory 306 stores a program 310. The memory 306 can include a high-speed RAM memory and can also include a non-volatile memory, such as at least one disk memory.

[0110] According to the scheme provided by the present application, multi-dimensional motion data of a subject is collected in real time by a plurality of IMU sensor arrays, and the multi-dimensional motion data is fused for posture by a generalized Kalman filter and a nonlinear complementary filter to obtain the limb posture, joint angle rate of change and angular momentum of the subject; wherein the multi-dimensional motion data includes three-axis acceleration, three-axis angular velocity and three-axis magnetic field data; the limb posture, joint angle rate of change and angular momentum of the subject are decomposed in time sequence mode according to a non-negative matrix decomposition method and a dynamic time warping algorithm, and a single gait cycle and a specific jump landing posture are extracted; the biomechanical feature vector of each motion primitive is calculated, the biomechanical feature vector includes joint impact peak, joint torque rate of change, motion stability index and specific frequency band energy distribution characteristics based on wavelet packet decomposition; the single gait cycle, the specific jump landing posture and the biomechanical feature vector are input into a trained hybrid injury-induced motion model, and a multi-dimensional injury risk index vector is output, the multi-dimensional injury risk index vector includes overload risk, fatigue injury accumulation risk and acute fall or sprain risk of the knee joint and ankle joint; the multi-dimensional injury risk index vector is compared with historical health data and motion target dynamic adjustment threshold of the subject, and if the threshold is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through a wearable device or an associated mobile application. The present application extracts single gait cycle and specific jump landing posture motion primitives from continuous motion data, overcoming the influence of motion speed and rhythm changes. The motion stability index is calculated by an enhanced phase space recursive quantitative analysis method, which can more sensitively reflect the dynamic instability in motion and predict the risk of falling or spraining. The biomechanics simulation layer maps the IMU data to muscle activation and ligament strain energy, enhancing the injury interpretability. Based on the baseline of biomechanics load tolerance of the subject's bone density DEXA data and serum collagen metabolism marker level, the truly personalized risk assessment is realized.

[0111] Those skilled in the art will appreciate that the modules in the apparatuses in the embodiments can be adapted and placed in one or more apparatuses other than the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and furthermore can be divided into multiple sub-modules or sub-units or sub-components. Any combination of all the features disclosed in the present specification (including the accompanying claims, abstract and drawings), and any method or apparatus so disclosed, can be taken in any combination, except that at least some of such features and / or processes or units are mutually exclusive, unless explicitly stated otherwise. Each feature disclosed in the present specification (including the accompanying claims, abstract and drawings) can be replaced by alternative features serving the same, equivalent or similar purpose, unless explicitly stated otherwise. Furthermore, the skilled person will appreciate that the combination of features of different embodiments implies that the features of the different embodiments are meant to be combined, unless explicitly stated otherwise. For example, in the claims below, any of the embodiments can be used in any combination. The application can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unitary claim, several of the devices mentioned in the embodiments can be implemented by means of one and the same hardware item. The steps of the above-described embodiments, unless explicitly stated otherwise, are not to be understood as having to be carried out in the order in which they are described.

Claims

1. A personalized exercise rehabilitation risk early warning method based on IMU dynamic instability assessment, characterized in that, include: Multidimensional motion data of the subject is collected in real time by multiple IMU sensor arrays. The multidimensional motion data is fused according to generalized Kalman filtering and nonlinear complementary filtering to obtain the subject's limb posture, joint angle change rate and angular momentum. The multidimensional motion data includes triaxial acceleration, triaxial angular velocity and triaxial magnetic field data. The subject's limb posture, joint angle change rate, and angular momentum are decomposed into temporal patterns using nonnegative matrix factorization and dynamic time warping algorithm to extract single gait cycles and specific jump landing postures. Calculate the biomechanical feature vector for each motion primitive, which includes the joint impact peak, joint torque change rate, motion stability index, and energy distribution characteristics of a specific frequency band based on wavelet packet decomposition. The single gait cycle, the specific jump landing posture, and the biomechanical feature vector are input into the trained hybrid injury-induced movement model, and a multi-dimensional injury risk index vector is output. The multi-dimensional injury risk index vector includes the overload risk of the knee and ankle joints, the risk of fatigue injury accumulation, and the risk of acute fall or sprain. The multidimensional injury risk index vector is compared with the subject's historical health data and a dynamic adjustment threshold for exercise goals. If the threshold is exceeded, real-time auditory, tactile, or visual feedback is provided to the subject through wearable devices or associated mobile applications.

2. The personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The formula for calculating the joint torque is: in, This is the quality matrix; For joint angle; Joint angular acceleration; Joint angular velocity; Confidence factor for IMU torque estimation; This represents the variance of the error between the IMU torque estimate and the laboratory reference data. Force estimated based on IMU; This is plantar pressure data.

3. The personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The motion stability index is enhanced using a phase space recursive quantitative analysis method, wherein the enhancement expression is: in, The enhanced motion stability index; This represents the total number of motion data samples. Sampling time interval For indicator functions; For time The joint angle state of the motion trajectory at any given moment; For time The joint angle state of the motion trajectory at any given moment; The threshold for the recursive neighborhood; The maximum Lyapunov index; For time The local divergence rate of the motion trajectory at any given moment.

4. The personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The hybrid injury-induced motion model includes a biomechanical simulation layer and a clinicopathological knowledge graph. The biomechanical simulation layer is the musculoskeletal multibody dynamics simulation layer of OpenSim, which is used to map IMU data into muscle activation force and ligament strain energy. The clinical pathology knowledge graph includes a damage rule reasoning engine, which marks the combined state of anterior cruciate ligament strain energy of the knee joint being greater than a first preset threshold and anterior tibial shear force being greater than a second preset threshold as a high risk of ACL tear; and associates cartilage wear location information in the historical health database, wherein an osteoarthritis progression warning is activated when the medial meniscus contact pressure peak continuously exceeds a third preset threshold.

5. The personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The method further includes: A baseline for biomechanical load tolerance was established based on the subjects' bone mineral density DEXA data and serum collagen metabolism markers. The thresholds are adjusted in real time according to the exercise goals. The competitive training mode uses FIFA 11 and the standard threshold of 90%, the rehabilitation mode uses 60%, and the elderly fall prevention mode introduces the gait symmetry weight coefficient. In particular, by using the correlation between the rate of decrease in muscle oxygen saturation and the entropy value of joint torque fluctuation, the exercise intensity is forcibly locked by the wearable device when the muscle oxygen recovery slope is less than 0.25% / s.

6. The personalized sports rehabilitation risk early warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The method further includes: When the ankle inversion velocity is detected to be >500° / s and the center of gravity projection deviates from the support base by 60%, microcurrent functional electrical stimulation is triggered to enhance the contraction of the peroneal muscle group. A virtual center of gravity trajectory guide line is projected into the AR glasses based on the risk index vector, so that the center of pressure on the user's feet remains within the trajectory tolerance zone when the user lands. Specifically, for patients who have undergone anterior cruciate ligament reconstruction, bone conduction headphones can be used to indicate insufficient knee flexion angle in the early stages of takeoff.

7. The personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The step of performing posture fusion on the multidimensional motion data based on generalized Kalman filtering and nonlinear complementary filtering to obtain the subject's limb posture, joint angle change rate, and angular momentum further includes: The triaxial acceleration and triaxial angular velocity data are preprocessed and state estimated by generalized Kalman filtering to obtain the initial limb posture; The three-axis magnetic field data and the limb posture are fused using a nonlinear complementary filter, and the limb posture is then corrected by magnetic field to obtain the fused limb posture. The relative posture between adjacent IMU sensors is calculated based on the fused limb posture to obtain the subject's joint angles. The time derivative of the joint angles is then calculated to obtain the joint angle change rate. The subject's angular momentum is calculated based on the fused limb posture, limb mass distribution model, and joint angle change rate.

8. The personalized sports rehabilitation risk early warning method based on IMU dynamic instability assessment according to claim 1, characterized in that, The step of performing temporal pattern decomposition of the subject's limb posture, joint angle change rate, and angular momentum based on the nonnegative matrix factorization method and dynamic time warping algorithm, and extracting single gait cycles and specific jump landing postures, further includes: The subjects' limb posture, joint angle change rate, and angular momentum data are used to construct a multidimensional temporal feature matrix; The multidimensional temporal feature matrix is ​​decomposed using a nonnegative matrix factorization method to obtain the base pattern matrix of gait cycle and jumping action and the corresponding weight coefficient matrix. The dynamic time warping algorithm is used to match and align the decomposed base pattern matrix with the real-time acquired motion data to correct changes in motion execution speed and rhythm, and to extract a single complete gait cycle and a specific jump landing posture from continuous data. The accuracy of the extracted motion primitives is ensured by using preset gait event detection algorithms and jump event detection algorithms.

9. A personalized sports rehabilitation risk early warning device based on IMU dynamic instability assessment, characterized in that, include: The data acquisition and posture fusion module is used to acquire multidimensional motion data of the subject in real time through multiple IMU sensor arrays, and perform posture fusion on the multidimensional motion data according to generalized Kalman filtering and nonlinear complementary filtering to obtain the subject's limb posture, joint angle change rate and angular momentum; wherein, the multidimensional motion data includes triaxial acceleration, triaxial angular velocity and triaxial magnetic field data; The temporal pattern decomposition module is used to perform temporal pattern decomposition on the subject's limb posture, joint angle change rate, and angular momentum according to the non-negative matrix factorization method and dynamic time warping algorithm, and extract single gait cycles and specific jump landing postures. The biomechanical feature calculation module is used to calculate the biomechanical feature vector of each motion primitive. The biomechanical feature vector includes the joint impact peak value, the joint torque change rate, the motion stability index, and the energy distribution characteristics of a specific frequency band based on wavelet packet decomposition. The injury risk assessment module is used to input the single gait cycle, the specific jump landing posture and the biomechanical feature vector into the trained hybrid injury-induced movement model, and output a multi-dimensional injury risk index vector, which includes the overload risk of the knee joint and ankle joint, the risk of fatigue injury accumulation and the risk of acute fall or sprain. The risk feedback and early warning module is used to compare the multi-dimensional injury risk index vector with the subject's historical health data and the dynamic adjustment threshold of the exercise target. If the threshold is exceeded, real-time auditory, tactile or visual feedback is provided to the subject through wearable devices or associated mobile applications.

10. A computer device, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-mentioned personalized sports rehabilitation risk warning method based on IMU dynamic instability assessment.

Images