Real-time motion error correction method and device and computer readable storage medium

Abstract

The invention discloses a real-time motion error correction method and device and a computer readable storage medium, and the method comprises the steps: obtaining an initial skeleton key point coordinate of a user in a preset calibration posture, and determining the initial length of each virtual elastic connection unit as a personalized zero load reference based on a preset virtual biomechanical constraint model; calculating an elastic deformation load index of each unit relative to the reference in real time, and performing weighted aggregation to generate regional aggregation pressure of the joint region; calculating an instantaneous fatigue generation rate by using a nonlinear mapping function representing the tolerance characteristic of the biological tissue, and updating an accumulated injury risk value based on a difference value between the rate and a preset physiological recovery rate through discrete integral operation of a time dimension; and finally generating hierarchical causal feedback containing a load source identifier or a risk warning. According to the method, the hidden compensation risk of shape-to-force error can be accurately identified, and the nonlinear cumulative effect of exercise fatigue can be quantified.

Description

Technical Field

[0001] This invention relates to the field of sports rehabilitation technology, and in particular to a real-time motion error correction method, device, and computer-readable storage medium. Background Technology

[0002] With the growing awareness of health among the public, home fitness guidance applications based on computer vision are becoming increasingly popular. Existing mainstream technologies typically use ordinary cameras to capture users' movement videos, extract key points of the human skeleton through pose estimation algorithms, and geometrically compare the user's real-time movement geometric parameters (such as joint angles and limb trajectories) with preset "standard movement templates".

[0003] However, existing technologies suffer from the following insurmountable technical contradictions in practical applications:

[0004] First, traditional visual methods can only capture the external geometric shape of the human body, i.e., "form"; while the root cause of sports injuries often lies in the internal soft tissues (muscles, tendons, ligaments) being subjected to stress beyond their limits, i.e., "force". Due to the complex "compensation mechanisms" of human movement, users may maintain a geometrically perfect "standard posture" by using incorrect muscle exertion patterns (for example, using the lumbar spine to compensate for the hip exertion during squats). This hidden risk of "correct form but incorrect force" is a blind spot that existing technologies based on pure geometric comparison cannot identify.

[0005] Secondly, most existing error correction logics are based on a "single-frame" slice-style judgment. Once the motion angle of the current frame returns to the standard range, the error reporting stops. However, physiological studies show that sports injuries are often the result of "non-linear accumulation" (i.e., fatigue effects). Maintaining a small, sub-healthy posture for a long time, without triggering the instantaneous error threshold, can cause cumulative damage that far exceeds that of a large instantaneous deviation. Current technology lacks the ability to handle this time integration effect, leading to underreporting of the risk of chronic strain injuries.

[0006] Therefore, how to overcome the limitations of pure geometric shape comparison under the condition of relying solely on monocular visual input, and establish a motion risk assessment mechanism that can reflect both the instantaneous internal load distribution and the cumulative effect of fatigue, so as to achieve accurate identification of implicit compensation and chronic injury, has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] This application provides a real-time motion correction method that aims to achieve real-time assessment and causal feedback of latent biomechanical loads and time-cumulative damage risks, relying solely on a vision-based approach.

[0008] To achieve the above objectives, embodiments of this application provide a real-time motion error correction method, including:

[0009] The initial skeletal key point coordinates of the user in a preset calibration posture are obtained. Based on the preset virtual biomechanical constraint model, the initial length of each virtual elastic connection unit in the virtual biomechanical constraint model is determined as a personalized zero-load reference.

[0010] Obtain the current real-time coordinates of the skeletal key points, calculate the real-time length of all virtual elastic connection units in the virtual biomechanical constraint model, and calculate the elastic deformation load index characterizing the tension of each virtual elastic connection unit based on the deformation degree of the real-time length relative to the personalized zero-load reference.

[0011] For a preset joint region, the elastic deformation load index of all virtual elastic connection units associated with the joint region is weighted and aggregated to generate a regional aggregated pressure value that characterizes the current stress state of the joint region.

[0012] Based on the regional aggregated pressure value, the instantaneous fatigue generation rate is calculated using a nonlinear mapping function that characterizes the tolerance properties of biological tissues, wherein the nonlinear mapping function defines an exponential growth relationship between the fatigue generation rate and the pressure intensity.

[0013] Discrete integral operations are performed in the time dimension. When the aggregated pressure value of the region is higher than the preset safety tolerance threshold, the cumulative damage risk value at the current moment is updated based on the difference between the instantaneous fatigue generation rate and the preset physiological recovery rate.

[0014] The elastic deformation load index and the cumulative damage risk value are compared with the corresponding preset thresholds to generate hierarchical causal feedback information containing the source identifier of elastic deformation load or the warning of cumulative damage risk, and then rendered and displayed.

[0015] In one embodiment, the preset calibration posture is defined as the user being in a naturally upright state without active muscle contraction;

[0016] The preset virtual biomechanical constraint model is constructed in the following way:

[0017] The human skeleton vector is defined as a rigid connection unit with constant length;

[0018] Based on human anatomy, virtual elastic connection units are established between adjacent joints and along cross-joint pathways, wherein the virtual elastic connection units are used to simulate the flexible connection relationships of human muscles, tendons or ligaments.

[0019] In one embodiment, calculating the elastic deformation load index, which characterizes the tension of each virtual elastic connection unit, includes:

[0020] For each virtual elastic connection unit j in the virtual biomechanical constraint model, the elastic deformation load index is calculated using the following formula:

[0021]

[0022] In the formula, Index load,j Let L be the elastic deformation load index of the j-th virtual elastic connection unit. j L is the real-time length. base,j K is the corresponding initial length in the personalized zero-load reference. j This is the preset tissue stiffness coefficient.

[0023] In one embodiment, generating a regional aggregate pressure characterizing the current stress state of the joint region includes:

[0024] The elastic deformation load index of all virtual elastic connection units associated with the joint region is obtained, and the basic pressure value is obtained by weighted summation using preset weighting coefficients.

[0025] The geometric angle between rigid connection units within the joint area is detected in real time. When the geometric angle exceeds the preset safety angle range, a shear force penalty term is calculated based on the magnitude of the excess angle.

[0026] The shear force penalty term is superimposed on the base pressure value to obtain the regional cohesive pressure.

[0027] In one embodiment, the regional polymerization pressure is calculated using the following formula:

[0028] P region =∑ j∈Ω (w j Index load,j )+Penalty shear (θ),

[0029] In the formula, P region The aggregated pressure of the region, Ω is the set of virtual elastic connection units associated with the corresponding joint region, w j For preset weight coefficients, Index load,j Penalty represents the elastic deformation load index of each virtual elastic connection unit. shear (θ) is the shear force penalty function based on the detected joint geometry angle θ.

[0030] In one embodiment, the instantaneous fatigue generation rate is calculated using the following formula:

[0031] Rate fatigue=A·exp(β·(P) region -P threshold )),

[0032] In the formula, Rate fatigue P represents the instantaneous fatigue generation rate; region P represents the polymerization pressure in the region. threshold The preset safety tolerance threshold is represented by A; A is the basic frequency factor; and β is the risk sensitivity coefficient.

[0033] In one embodiment, discrete integration is performed in the time dimension, using the following update rule:

[0034] Risk cum (t)=Risk cum (t-1)+(Rate fatigue (t)-Rate recovery )·Δt,

[0035] In the formula, Risk cum (t) represents the cumulative damage risk value at the current moment; Risk cum (t-1) represents the cumulative damage risk value at the previous time step; Rate fatigue (t) represents the instantaneous fatigue generation rate at the current moment; Rate recovery The preset physiological recovery rate is used to simulate the metabolic recovery process of the human body under low load conditions; Δt is the sampling time interval.

[0036] The cumulative damage risk value is limited to a preset range [0,1].

[0037] In one embodiment, generating hierarchical causal feedback information that includes an identifier of the source of elastic deformation load or a warning of cumulative damage risk includes:

[0038] When any of the elastic deformation load indices exceeds the first load threshold, a visual connection marker line is rendered between the two skeletal key points connected by the corresponding virtual elastic connection unit in the user motion video stream.

[0039] When the cumulative damage risk value exceeds the second load threshold, a warning rendering effect covering the corresponding joint area is generated in the user's motion video stream.

[0040] To achieve the above objectives, this application also proposes a real-time motion correction device, including a memory, a processor, and a real-time motion correction program stored in the memory and executable on the processor. When the processor executes the real-time motion correction program, it implements the real-time motion correction method as described in any of the above claims.

[0041] To achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a real-time motion correction program, wherein the real-time motion correction program, when executed by a processor, implements the real-time motion correction method as described in any of the preceding claims.

[0042] The real-time motion error correction method of this application has the following beneficial effects:

[0043] 1. Achieved accurate identification of implicit compensatory risks of "correct form but incorrect force": By employing a "virtual biomechanical constraint model" and establishing a "personalized zero-load benchmark," this application can calculate the "elastic deformation load index," reflecting the tension of internal soft tissues, in real time based on sparse skeletal coordinates obtained from monocular vision. Thus, in the absence of contact sensors (such as electromyography patches), this application can penetrate the appearance of external geometric posture to infer the relative tension state of the human body's internal structures, effectively identifying implicit biomechanical risks caused by users who, although conforming to standards in geometric appearance, actually rely on incorrect muscle group exertion (such as compensatory mechanisms), overcoming the limitations of simply comparing geometric angles.

[0044] 2. Achieved fatigue accumulation and chronic injury prediction consistent with human physiological characteristics: By introducing a "damage dynamics evolution model," utilizing a "nonlinear mapping function" characterizing the tolerance properties of biological tissues, and "discrete integral operations" including "physiological recovery rate," this application can simulate the exponential growth of nonlinear damage in biological soft tissues under overload conditions, as well as the metabolic recovery mechanism under rest conditions, thereby establishing a risk assessment system with "time memory" capabilities. This enables the system not only to alert to instantaneous large-amplitude errors but also to keenly detect the "fatigue accumulation effect" caused by maintaining small deviations in posture for extended periods, effectively preventing chronic strain.

[0045] 3. Real-time biomechanical simulation with extremely low computational power consumption was achieved: By constructing a topological network composed of "rigid connection units" and "virtual elastic connection units," the traditional finite element analysis (FEM) mesh model was replaced. This application transforms the solution of complex differential equations into lightweight linear algebra and matrix operations, successfully solving the technical challenge of real-time internal force analysis under engineering constraints of "sparse input data (only coordinates)" and "limited computational resources (mobile terminal)," enabling high-precision biomechanical assessments to run smoothly at high frame rates on ordinary consumer electronic devices (such as mobile phones and smart TVs).

[0046] 4. Provides hierarchical visualization guidance with causal explanation: By establishing a dual evaluation system of "elastic deformation load index (space / cause)" and "cumulative damage risk value (time / effect)," and generating "connection marker lines" and "regional warning effects" respectively. This application breaks away from the black-box mode of traditional technologies that only provide binary "right / wrong" feedback, and can intuitively show users the specific anatomical source of the risk (i.e., "which muscle is tight") and the current overall fatigue level, realizing a leap from "simple error reporting" to "root cause error correction," significantly improving the user's training efficiency and safety. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0048] Figure 1 This is a block diagram of a module structure of an embodiment of the real-time motion error correction device of the present invention;

[0049] Figure 2 This is a flowchart illustrating an embodiment of the real-time motion error correction method of the present invention.

[0050] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0051] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0052] To better understand the above technical solutions, exemplary embodiments of this disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art.

[0053] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The quantifier "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of "first," "second," and "third," etc., does not indicate any order and can be interpreted as names.

[0054] like Figure 1 As shown, Figure 1 This is a schematic diagram of the hardware operating environment server 1 (also called a real-time motion error correction device) involved in the embodiment of the present invention.

[0055] The server in this embodiment of the invention includes devices with display functions such as "Internet of Things devices", smart air conditioners, smart lights, smart power supplies with network connectivity, AR / VR devices with network connectivity, smart speakers, autonomous vehicles, PCs, smartphones, tablets, e-book readers, and portable computers.

[0056] like Figure 1 As shown, the server 1 includes: a memory 11, a processor 12, and a network interface 13.

[0057] The memory 11 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the server 1, such as the hard disk of the server 1. In other embodiments, the memory 11 can also be an external storage device of the server 1, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the server 1.

[0058] Furthermore, the memory 11 may include both internal storage units of the server 1 and external storage devices. The memory 11 can be used not only to store application software and various types of data installed on the server 1, such as the code of the real-time motion error correction program 10, but also to temporarily store data that has been output or will be output.

[0059] In some embodiments, processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 11 or process data, such as executing real-time motion correction program 10.

[0060] The network interface 13 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface), which is typically used to establish communication connections between the server 1 and other electronic devices.

[0061] The network can be the Internet, a cloud network, a Wi-Fi network, a Personal Area Network (PAN), a Local Area Network (LAN), and / or a Metropolitan Area Network (MAN). Various devices in the network environment can be configured to connect to the communication network according to various wired and wireless communication protocols. Examples of such wired and wireless communication protocols may include, but are not limited to, at least one of the following: Transmission Control Protocol and Internet Protocol (TCP / IP), User Datagram Protocol (UDP), Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), ZigBee, EDGE, IEEE 802.11, Li-Fi, 802.16, IEEE 802.11s, IEEE 802.11g, multi-hop communication, wireless access points (APs), device-to-device communication, cellular communication protocols, and / or Bluetooth communication protocols, or combinations thereof.

[0062] Optionally, the server may also include a user interface, which may include a display, an input unit such as a keyboard, and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be referred to as a screen or display unit, used to display information processed in server 1 and to display a visual user interface.

[0063] Figure 1 Only server 1, which includes components 11-13 and a real-time motion correction program 10, is shown. Those skilled in the art will understand that... Figure 1 The structure shown does not constitute a limitation on server 1 and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0064] In this embodiment, the processor 12 can be used to call the real-time motion correction program stored in the memory 11 and perform the following operations:

[0065] The initial skeletal key point coordinates of the user in a preset calibration posture are obtained. Based on the preset virtual biomechanical constraint model, the initial length of each virtual elastic connection unit in the virtual biomechanical constraint model is determined as a personalized zero-load reference.

[0066] Obtain the current real-time coordinates of the skeletal key points, calculate the real-time length of all virtual elastic connection units in the virtual biomechanical constraint model, and calculate the elastic deformation load index characterizing the tension of each virtual elastic connection unit based on the deformation degree of the real-time length relative to the personalized zero-load reference.

[0067] For a preset joint region, the elastic deformation load index of all virtual elastic connection units associated with the joint region is weighted and aggregated to generate a regional aggregated pressure value that characterizes the current stress state of the joint region.

[0068] Based on the regional aggregated pressure value, the instantaneous fatigue generation rate is calculated using a nonlinear mapping function that characterizes the tolerance properties of biological tissues, wherein the nonlinear mapping function defines an exponential growth relationship between the fatigue generation rate and the pressure intensity.

[0069] Discrete integral operations are performed in the time dimension. When the aggregated pressure value of the region is higher than the preset safety tolerance threshold, the cumulative damage risk value at the current moment is updated based on the difference between the instantaneous fatigue generation rate and the preset physiological recovery rate.

[0070] The elastic deformation load index and the cumulative damage risk value are compared with the corresponding preset thresholds to generate hierarchical causal feedback information containing the source identifier of elastic deformation load or the warning of cumulative damage risk, and then rendered and displayed.

[0071] Based on the hardware architecture of the aforementioned real-time motion correction device, an embodiment of the real-time motion correction method of the present invention is proposed. The real-time motion correction method of the present invention aims to achieve real-time assessment and causal feedback of latent biomechanical loads and time-cumulative injury risks, relying solely on a visual approach.

[0072] Reference Figure 2 , Figure 2 In one embodiment of the real-time motion correction method of the present invention, the real-time motion correction method includes the following steps:

[0073] S10. Obtain the initial skeletal key point coordinates of the user in the preset calibration posture, and determine the initial length of each virtual elastic connection unit in the preset virtual biomechanical constraint model as a personalized zero-load reference based on the preset virtual biomechanical constraint model.

[0074] In this embodiment of the application, step S10 is the initialization and benchmark construction stage of the entire real-time motion correction process, which aims to eliminate individual differences among different users in terms of height, limb proportions (such as arm span and leg length), etc., and to establish a unified evaluation reference system for subsequent mechanical calculations.

[0075] Specifically, the processing flow for this step is as follows:

[0076] First, the system acquires real-time video streams or image data from the user via the terminal device's camera component. Upon system startup or after the user triggers a calibration command, the system guides the user to maintain a "preset calibration posture." This preset calibration posture is a predefined, standard static movement used to capture the basic skeletal structure of the human body. For example, the system can detect whether the user is in an upright, static state or a T-pose with arms outstretched. The system uses human pose estimation algorithms (such as deep learning-based visual skeleton extraction algorithms) to analyze video frames. When it detects that the user's posture meets the preset conditions and remains stable, it extracts the initial skeletal keypoint coordinates (denoted as P) of the user's body joints at that moment. init It typically includes three-dimensional spatial coordinates (x, y, z).

[0077] Secondly, the system loads a "preset virtual biomechanical constraint model." This model is a data structure or topological network used to simulate the biomechanical characteristics of the human body, and it predefines the connection relationships between key points of various bones in the human body. In this model, the paths connecting specific key points of bones are defined as "virtual elastic connection units." These units logically represent soft tissue structures with elastic properties in the human body (such as muscle bundles, tendons, or ligaments).

[0078] Finally, the system calculates and locks the "personalized zero-load baseline." Based on the extracted initial skeletal keypoint coordinates, the system maps them to corresponding nodes in the virtual biomechanical constraint model. Next, the system iterates through all virtual elastic connection units defined in the model, using the Euclidean distance formula to calculate the spatial distance between the two ends of each unit in the current calibration posture. This calculated distance is the "initial length" of the virtual elastic connection unit. Since the user is in a specific calibration posture (e.g., naturally relaxed standing), the system assumes that each elastic unit is in a zero-potential-energy state without external stretching. Therefore, this set of initial length data is stored and locked as the user's personalized zero-load baseline. During subsequent real-time monitoring, this baseline will be used to measure the relative deformation of each connection unit.

[0079] In some embodiments of this application, the preset calibration posture is strictly defined in order to ensure the physical accuracy of the "personalized zero-load reference".

[0080] Specifically, this posture is defined as the user being in a naturally upright state without active muscle contraction (i.e., a "zero potential energy state"). This posture was chosen because, when standing naturally and relaxed, the body's antagonistic muscle groups are in a balanced and low-tension state, at which point the measured length of the virtual elastic connection unit is closest to its biologically anatomically resting length. This minimizes the baseline deviation caused by muscle pre-contraction, ensuring that the subsequently calculated tensile deformation rate accurately reflects passive or active tensile loads.

[0081] Furthermore, in order to simulate the internal stress state of the human body, the preset virtual biomechanical constraint model is not a simple geometric connection, but a physical topology constructed through the following steps S11 to S12:

[0082] Step S11: Define the human skeleton vector as a rigid connection unit with constant length.

[0083] Specifically, the system defines the vector connecting two adjacent skeletal key points (e.g., the hip joint and the knee joint) as a rigid body in the model. During subsequent calculations, the system enforces that the magnitude of this vector (i.e., the bone length) remains constant, preventing axial deformation. This approach is based on the biological fact that "human bones undergo very little observable deformation under normal exercise loads." While ensuring the physical reliability of the model, it significantly reduces the degrees of freedom and the number of variables in the system's calculations, thereby reducing the computational demands on the terminal device.

[0084] Step S12: Based on human anatomy, establish virtual elastic connection units between adjacent joints and along cross-joint paths.

[0085] Specifically, the system loads a pre-defined anatomical topology mapping table, generating elastic connection lines between specific pairs of skeletal nodes. These connection lines are established not only between adjacent joint nodes (e.g., connecting the femur and tibia to simulate the collateral ligaments or joint capsule of the knee), but also along paths spanning multiple joints (e.g., directly connecting the pelvic node to the tibia to simulate biarticular muscles such as the rectus femoris). These virtual elastic connection units are logically endowed with physical properties similar to springs or polymers, simulating the flexible connections of soft tissues such as muscles, tendons, or ligaments during movement. They can lengthen or shorten with the relative movement of skeletal nodes, serving as the core physical carrier for calculating internal tension loads in this system.

[0086] It is understandable that defining the skeleton as a rigid unit of constant length in step S11 aligns with the physical fact that human bones are not easily deformed during movement. Furthermore, by locking the skeleton length as a variable, the system's computational freedom is significantly reduced, thereby drastically decreasing computational power consumption and making real-time processing on mobile devices possible. Simultaneously, by introducing virtual elastic connection units in step S12, the internal soft tissues (muscles, tendons) invisible in anatomy are materialized into computable topological connections. Using these as the sole physical carrier for subsequent calculations of the "elastic deformation load index" allows this solution to overcome the limitations of visual geometry, achieving a technological leap from "external posture monitoring" to "internal tension monitoring."

[0087] S20. Obtain the current real-time coordinates of the skeletal key points, calculate the real-time length of all virtual elastic connection units in the virtual biomechanical constraint model, and calculate the elastic deformation load index characterizing the tension of each virtual elastic connection unit based on the deformation degree of the real-time length relative to the personalized zero-load reference.

[0088] Specifically, after completing the initial baseline construction, step S20 enters the real-time monitoring phase. The system continuously extracts the real-time skeletal keypoint coordinates of each frame from the video stream at a preset frame rate (e.g., 30fps). Subsequently, the system traverses each virtual elastic connection unit in the virtual biomechanical constraint model, using the real-time coordinates of the two end nodes to calculate its current Euclidean distance, i.e., the real-time length.

[0089] To quantify the tension state within these connecting units, this application introduces the "elastic deformation load index". load This physical quantity reflects not only the geometric elongation of the connecting unit but also the material properties of biological tissue, thus characterizing the current tension of the unit (i.e., the simulated muscle or tendon).

[0090] In some embodiments of this application, the elastic deformation load index, which characterizes the tension of each virtual elastic connection unit, is calculated using the following mathematical model:

[0091] For each virtual elastic connection unit j in the virtual biomechanical constraint model, the system calculates the elastic deformation load index using the following formula:

[0092]

[0093] Among them, Index load,j Let be the elastic deformation load index of the j-th virtual elastic connection unit. This is a dimensionless value used to intuitively reflect the relative load intensity of the unit.

[0094] L j The real-time length is the straight-line distance between the two nodes of the virtual elastic connection unit at the current moment.

[0095] L base,j This is the corresponding initial length in the personalized zero-load baseline, that is, the unit length of the user in a natural relaxed state as measured in step S10.

[0096] K j This is a preset tissue stiffness coefficient. This coefficient is pre-set based on human anatomy and is used to differentiate the elastic modulus of different soft tissues. For example, when simulating the connecting units of harder tissues such as the Achilles tendon or collateral ligament, its K... j The value will be set relatively large (e.g., 15), meaning that the same deformation rate will produce a larger load index; while for connecting units simulating softer muscles such as "abdominal muscles," their K... j The value is set to a smaller value (e.g., 08).

[0097] The `max(0,…)` function ensures that a positive load exponent is generated only when the real-time length exceeds the zero-load reference length; it takes the value 0 when the length shortens (compressed or relaxed), thus focusing only on the tensile state that poses a risk of damage to biological tissue. Since soft tissues such as muscles and ligaments primarily generate tension through contraction or stretching, and typically do not generate significant structural stress under compression (relaxation), when the real-time length L... j Less than the initial length L base,j When the value is in a relaxed / compressed state, the function forces the calculation result to be set to 0, thereby avoiding negative values ​​from interfering with subsequent risk assessment.

[0098] For example, suppose the system is monitoring a user's "squat" motion and is focusing on a virtual connecting unit (simulating the medial collateral ligament) on the inside of the knee joint. When the user stands, this unit is at a reference length L. base =10cm, at this time Index load =0.

[0099] When a user commits the "knee valgus" error during a squat, the medial joint space of the knee is abnormally widened. At this point, the system detects the real-time length L of this virtual unit. j Increased to 11cm. Assume the ligament's stiffness coefficient K... j =2.0. The system executes the above formula to calculate:

[0100]

[0101] This non-zero index loadA value of 0.2 quantitatively indicates that the ligament is currently under significant tensile load. Conversely, if the user's knee widens outward, causing a narrowing of the medial collateral space, L... j If the value becomes 9cm, then the result of max(0,9-10) in the formula is 0, i.e., the index... load =0 accurately reflects that the part is in a relaxed and safe state at this time.

[0102] It is understandable that by calculating the uniaxial tensile deformation of each virtual elastic connection unit relative to a personalized zero-load benchmark in real time and introducing a tissue stiffness coefficient for weighting, this step transforms pure geometric displacement into a load distribution with biomechanical significance. This enables high-fidelity inference of the user's internal tension state without the need for any force sensors, thus providing a core data foundation for identifying implicit compensatory actions.

[0103] S30. For a preset joint region, the elastic deformation load index of all virtual elastic connection units associated with the joint region is weighted and aggregated to generate a regional aggregated pressure value that characterizes the current stress state of the joint region.

[0104] After calculating the independent load index of each virtual elastic connection unit, step S30 aims to assess the overall stress state of the entire joint region (e.g., the entire left knee joint, the entire right shoulder joint). This is because tension in a single ligament or muscle does not necessarily indicate an immediate risk of joint injury, but if the soft tissue group surrounding the joint is under high tension as a whole, or if the joint itself is in a dangerous geometric configuration, the probability of injury will increase significantly.

[0105] In some embodiments, to more comprehensively quantify this combined risk, a regional aggregate pressure characterizing the current stress state of the joint region is generated, specifically through the following sub-steps S31 to S33:

[0106] S31. Obtain the elastic deformation load index of all virtual elastic connection units associated with the joint region, and perform weighted summation using preset weighting coefficients to obtain the basic pressure value.

[0107] The system first determines the set Ω of virtual elastic connection units (e.g., virtual units simulating the medial collateral ligament, lateral collateral ligament, patellar ligament, etc.) encompassing the joint region of interest (e.g., "left knee"). Then, the system calculates the elastic deformation load index (Index) for each unit in this set. load,j A weighted sum is performed. A weighting coefficient (w) is introduced. j The purpose of this study is to reflect the differences in the contribution of different soft tissues to joint stability. For example, in a knee joint model, the medial collateral ligament (MCL) is crucial for preventing genu valgum, and therefore it is given a high weight (e.g., w). MCL=0.4); while the weight of certain accessory muscle groups is relatively low (e.g., w aux =0.1). Through weighted summation, the system calculates a "basic pressure value" that reflects the pure tension dimension.

[0108] S32. Real-time detection of the geometric angle between rigid connection units within the joint area. When the geometric angle exceeds the preset safe angle range, a shear force penalty term is calculated based on the magnitude of the angle exceeding the limit.

[0109] Specifically, considering only tensile tension is insufficient to fully cover the causes of sports injuries. Shear force is often another major culprit in meniscus wear or ligament tears, and it is usually directly related to poor joint geometry. Therefore, step S32 introduces a detection mechanism that couples geometry and mechanics.

[0110] The system calculates the angle θ between adjacent rigid connection units (bone vectors) within the joint region in real time. For example, during a squat, it calculates the projected angle between the femur and tibia in the coronal plane (i.e., genu valgum / genu valgum angle). The system presets a "safe angle range" (e.g., [-5°, 5°]). When the measured angle θ falls within the safe range, the system determines there is no abnormal shear risk, and the shear force penalty term is set to 0. When the measured angle θ exceeds the safe range (e.g., the genu valgum angle reaches 15°), the system determines that the joint is in a dangerous configuration, with huge internal shear stress. At this time, the system calculates the angle based on the excess amplitude |θ-θ safe Calculate a nonlinear "shear force penalty term".

[0111] S33. The shear force penalty term is superimposed on the base pressure value to obtain the regional cohesive pressure.

[0112] Finally, the system superimposes the "basic pressure value" (tension contribution) calculated in step S31 with the "shear force penalty term" (geometric shear contribution) calculated in step S32 to obtain the final regional cohesion pressure (P). region This indicator combines the dual destructive factors of "tension" and "shear," and can more accurately reflect biomechanical load.

[0113] Specifically, the regional polymerization pressure is calculated using the following formula:

[0114] P region =∑ j∈Ω (w j Index load,j )+Penalty shear (θ),

[0115] In the formula, P regionAggregate pressure in the region is a core indicator for assessing the immediate risk to the joint.

[0116] Ω represents the set of virtual elastic connection units associated with the corresponding joint region.

[0117] w j The preset weighting coefficients satisfy ∑w j =1 or set according to the model requirements.

[0118] Index load,j The elastic deformation load index of each virtual elastic connection unit (from step S20).

[0119] Penalty shear (θ) is the shear force penalty function based on the detected joint geometry angle θ.

[0120] In a preferred embodiment of this application, to highlight the harmfulness of poor geometry, the shear force penalty function is designed as an exponentially growing function, for example:

[0121] Penalty shear (θ)=λ·exp(α·max(0,|θ|-θ thresh ))-λ,

[0122] In the formula, θ thresh Here, λ and α are the safety angle thresholds, and λ and α are constants that adjust the intensity of the penalty. This means that as long as the angle slightly exceeds the limit, the penalty term may be small; however, once it severely exceeds the limit (such as severe inward deduction), the penalty term will increase explosively, directly dominating P. region The value triggers a high-priority alarm.

[0123] Continuing with the squat example, let's assume the index of the user's medial collateral ligament at the knee joint... load =0.2 (weight w=0.5), Lateral Collateral Ligament Index load =0 (weight w = 0.5).

[0124] At this point, the basic pressure value = 0.5 × 0.2 + 0.5 × 0 = 0.1.

[0125] However, the system also detected that the user's knee valgus angle θ = 15°, which far exceeds the safety threshold θ. thresh =5°.

[0126] Based on the shear force penalty function (assuming λ = 1, α = 05), the shear force penalty term is calculated as ≈ exp(0.5 × (15-5)) = exp(5) ≈ 148 (which may be 1.48 after normalization adjustment).

[0127] Ultimately, the regional aggregation pressure P region=0.1+1.48=1.58.

[0128] As can be seen, although the simple ligament tensile load (0.1) is not high, the final aggregation pressure soars due to the huge shear force penalty (1.48) caused by poor posture. This accurately simulates the real biomechanical scenario of "the meniscus may be damaged by shear force even though the ligament is not strained".

[0129] It is understandable that by weighting and aggregating the dispersed elastic deformation load indices according to anatomical significance and superimposing shear force penalty terms for typical injury mechanisms, this step S30 transforms point-like tension information into a comprehensive stress state at the joint level, thereby achieving precise quantification of the most common high-risk patterns in clinical practice (such as knee valgus and lumbar compensation).

[0130] S40. Based on the regional aggregated pressure value, the instantaneous fatigue generation rate is calculated using a nonlinear mapping function that characterizes the tolerance properties of biological tissues, wherein the nonlinear mapping function defines an exponential growth relationship between the fatigue generation rate and the pressure intensity.

[0131] Specifically, the regional aggregate pressure value (P) reflecting the macroscopic stress state of the joint was obtained. region After that, the core task of step S40 is to complete the mapping from the "mechanical dimension" to the "physiological damage dimension".

[0132] Because the risk of injury to human soft tissue does not increase linearly with load, but rather exhibits a significant nonlinear abrupt change. That is, within a safe load range, the microscopic damage to the tissue is almost zero; once the tolerance threshold is exceeded, the risk of injury explodes exponentially with increasing load. To accurately simulate this biological characteristic, this step introduces the intermediate variable of "instantaneous fatigue generation rate".

[0133] Specifically, the instantaneous fatigue generation rate is calculated using the following formula:

[0134] Rate fatigue =A·exp(β·(P) region -P threshold )),

[0135] In the formula:

[0136] Rate fatigue This represents the instantaneous fatigue generation rate. It quantifies the intensity of the damage "flow" experienced by the joint at the current time frame (e.g., frame t). The higher the value, the greater the potential damage to the body from maintaining that action at that instant.

[0137] P regionThe region's polymerization pressure is calculated from the preceding step S30 and includes tensile load and shear force penalty terms.

[0138] P threshold This refers to the preset safety tolerance threshold. This threshold is a constant (e.g., a normalized 1.0) set based on human anatomical data, representing the maximum physiological stress that the joint region can withstand without microstructural damage.

[0139] A is the base frequency factor used to adjust the order of magnitude of the output rate to match the time scale of subsequent integration operations. For example, the value of A can be 0.02 to ensure that when P... region,k It is exactly equal to P threshold At that time, Rate fatigue,k ≈0.

[0140] β is the risk sensitivity coefficient. This is a key adjustment parameter used to set the non-linear steep increase in the fatigue generation rate with increasing pressure, in order to simulate the damage characteristics of biological tissues under overload conditions.

[0141] The above formula constructs a nonlinear dynamic model describing the damage evolution of biological tissues under overload conditions. In this model, the regional cohesion pressure (P) region ) is defined as the external driving load that causes damage to the microstructure of biological soft tissues, while the tolerance threshold (P) is... threshold This represents the maximum physiological limit of a specific biological tissue's ability to maintain structural integrity and elastic recovery.

[0142] By utilizing the monotonically increasing and nonlinear characteristics of the exponential function (exp), this application can accurately simulate the explosive increase in the rate of microscopic damage generation in biological tissues as the load intensity increases after the external load exceeds the physiological limit. This modeling method overcomes the shortcomings of traditional linear models in characterizing the "brittle fracture" or "overload collapse" characteristics of biological tissues, and conforms to the real biomechanical response mechanism of human soft tissues under ultimate stress.

[0143] Regarding the role of the risk sensitivity coefficient β, this coefficient is usually set to a value greater than 1 (e.g., β = 25). This ensures that the model has "overload sensitivity".

[0144] When P region Slightly lower than or equal to P threshold When the exponent term is negative or zero, the calculated Rate fatigue Extremely low, simulating the "harmless" state of the human body during safe movement.

[0145] When P region Only exceeding P thresholdWhen the amount is small (e.g., exceeding 20%), due to the amplification effect of β and the nonlinear characteristics of the exp function, the Rate... fatigue It will increase significantly, not just a simple 20% increase.

[0146] When P region When the rate significantly exceeds the limit (e.g., exceeding 50%), fatigue This will result in explosive growth. This precisely simulates the physical process of rapid structural collapse of biological tissues under extreme loads (e.g., ligament tears often occur in an instantaneous overload).

[0147] Let's continue with the squatting example mentioned above.

[0148] Scenario 1 (Safety): Assume that in the initial stage of the action, P region =0.8. Set threshold P threshold =1.0, sensitivity β = 25. At this time, P region <P threshold In actual algorithm implementation, a conditional check is usually added: if the pressure is below the threshold, the Rate is directly set. fatigue =0. Even when calculated using a formula, the exponent is negative, and the result is close to 0. This is consistent with common sense (easy movements do not lead to fatigue accumulation).

[0149] Scenario 2 (Slight Compensation): The user begins to fatigue, P region It rose to 1.2 (20% above the standard).

[0150] Calculate: Rate fatigue ∝exp(2.5×(1.2-1.0))=exp(0.5)≈1.65. At this point, damage begins to form at a moderate rate.

[0151] Scenario 3 (Severe Shear): The user experiences severe knee valgus, leading to a surge in shear force penalty, causing P... region It surged to 1.6 (which seems to be only 0.4 more than scenario two).

[0152] Calculate: Rate fatigue ∝exp(2.5×(1.6-1.0))=exp(1.5)≈4.48.

[0153] As can be seen from the comparison, the pressure value only increased from 1.2 to 1.6 (an increase of about 33%), but the fatigue generation rate soared from 1.65 to 4.48 (an increase of about 171%). This calculation result of "a slight increase in pressure, but a double increase in risk" is precisely the technical effect that this application aims to achieve using a nonlinear mapping function, which successfully reproduces the vulnerability characteristics of physical biological tissues in the digital world.

[0154] S50. Perform discrete integral operation in the time dimension. When the aggregated pressure value of the region is higher than the preset safety tolerance threshold, accumulate the difference between the instantaneous fatigue generation rate and the preset physiological recovery rate to update the cumulative damage risk value at the current moment.

[0155] Specifically, although the instantaneous fatigue generation rate calculated in step S40 can reflect the intensity of the action danger at the current moment, in biomechanics, soft tissue failure is often not caused by a single instantaneous overload, but by fatigue accumulation due to prolonged sub-healthy stress conditions. Therefore, this step introduces discrete integral calculations in the time dimension to simulate the damage accumulation effect of human tissues during continuous stress and the self-repair mechanism during rest.

[0156] In some embodiments, discrete integration is performed in the time dimension using the following update rule (difference equation):

[0157] Rish cum (t)=Risk cum (t-1)+(Rate fatigue (t)-Rate recovery )·Δt,

[0158] In the formula:

[0159] Risk cum (t) represents the cumulative damage risk value at the current moment. This is a dimensionless state variable used to quantify the current overall fatigue level or damage risk level of the joint region.

[0160] Risk cum (t-1) represents the cumulative damage risk value at the previous moment, which is initialized to 0 when the system starts or after the user completes a full rest.

[0161] Rate fatigue (t) represents the instantaneous fatigue generation rate at the current moment, calculated in step S40 based on the current regional aggregation pressure value. When the regional aggregation pressure value is higher than the preset safety tolerance threshold, this rate is positive, driving the risk value to increase.

[0162] Rate recovery The preset physiological recovery rate represents a constant or variable used to simulate the body's metabolic recovery (such as lactic acid clearance and microstructural rebound) under low-load conditions. Physically, it represents the body's own "negative entropy" capacity. The introduction of this parameter gives the model a "competitive accumulation" characteristic, meaning the final accumulation of damage depends on the dynamic interplay between "generated damage" and "recovered damage."

[0163] Δt is the sampling time interval, which corresponds to the frame interval of the video stream (e.g., for a 30fps video stream, Δt≈0.033s).

[0164] To facilitate subsequent graded feedback and threshold determination, the cumulative damage risk value is limited to a preset range [0,1] (i.e., 0% to 100%). After each integral update, the system performs boundary checks (Clamping):

[0165] If the calculation result is less than 0 (i.e., the recovery amount exceeds the cumulative amount), force Risk to be reset. cum (t) = 0 indicates that the body is in a safe state of full recovery and no negative accumulation occurs.

[0166] If the calculation result is greater than 1 (i.e., the accumulated amount overflows), force Risk to be reset. cum (t) = 1 indicates that the risk has reached the critical value, and the system will trigger the highest level of warning.

[0167] We will continue to use the aforementioned "squat" example.

[0168] Phase A (Continuous Compensation): The user maintains a squatting posture with slight knee valgus (P region (Slightly above the threshold). Assuming the Rate is calculated at this point... fatigue =0.5 units / s, while the body's natural recovery rate (Rate_{recovery}) = 0.1 units / s. According to the formula, for every second that passes, the cumulative risk value will increase by (0.5 - 0.1) × 1 = 0.4. This means that although the instantaneous intensity of the action is not high, if the user persists for 25 seconds, the cumulative risk value (Risk) will increase significantly. cum The energy level will accumulate from 0 to 10 (full bar), triggering an alarm. This accurately simulates the risk of chronic overuse injury, similar to the "boiling frog" effect.

[0169] Phase B (Rest and Recovery): The user stands and rests after receiving the alarm (P region (Below the threshold). At this time, the Rate output in step S40 is... fatigue Zeroing out. The integral formula becomes: Risk cum (t)=Risk cum (t-1)+(0-0.1)·Δt. It can be seen that the cumulative risk value will decrease linearly at a rate of 0.1 units / s. This simulates the physiological process of muscle fatigue gradually subsiding during rest.

[0170] It is understandable that by introducing a discrete integral mechanism with a recovery term, step S50 transforms the instantaneous load in the spatial dimension into a dynamic risk pool in the temporal dimension. This allows for both rapid response to acute high loads and precise capture of chronic cumulative damage caused by minor errors over a long period, thereby achieving dynamic risk quantification that aligns with the actual physiological process of fatigue and recovery in the human body. Thus, this application can not only identify whether the current action is standard but also answer the question of whether the user can continue, providing scientific guidance with temporal depth.

[0171] S60. Compare the elastic deformation load index and the cumulative damage risk value with the corresponding preset thresholds respectively, generate hierarchical causal feedback information containing the source identifier of elastic deformation load or the warning of cumulative damage risk, and render and display it.

[0172] Step S60 is the decision-making and output layer of the technical solution of this application. Unlike the traditional approach that only provides a general "error" message or "score," this step constructs a "hierarchical causal feedback mechanism" based on the results of previous calculations. The core logic of this mechanism lies in distinguishing between "latent structural hazards (cause)" and "explicit damage crises (effect)," and providing differentiated visual guidance according to the urgency of the risk.

[0173] Specifically, the system sets a first load threshold (for the spatial dimension index). load For example, setting it to 0.8 to characterize soft tissue entering a state of stress) and the second load threshold (Risk for the time dimension). cum For example, setting it to 0 or 9 indicates that fatigue accumulation is approaching the critical point.

[0174] The system monitors the load index of all virtual elastic connection units and the cumulative risk value of the joint area in real time, and triggers the following rendering logic based on the comparison results (steps S61-S62):

[0175] S61. When any of the elastic deformation load indices exceeds the first load threshold, a visual connection marker line is rendered between the two skeletal key points connected by the corresponding virtual elastic connection unit in the user motion video stream.

[0176] This step aims to provide "implicit risk feedback." When the system detects the index of a specific virtual elastic connection unit (e.g., a unit simulating the medial collateral ligament of the knee joint),... load When the load exceeds the first load threshold, but the overall cumulative risk has not yet reached its limit, the system determines that there is a specific structural tension hazard.

[0177] To allow users to intuitively perceive "where it's tight" or "where the force is incorrect," the system will accurately locate the two key skeletal points (such as the medial epicondyle of the femur and the medial condyle of the tibia) connected by the virtual unit in the video stream and draw a visual connection line between the two points.

[0178] This marker line is typically rendered as a yellow or orange dashed line, and the line thickness or opacity can be adjusted according to the index. load The numerical dynamic changes. This visually simulates a "tight rubber band", intuitively revealing to the user the specific anatomical source (i.e. "cause") that leads to the current abnormal force, thereby guiding the user to adjust their posture in a targeted manner (e.g., "relax the inside of the knee").

[0179] S62. When the cumulative damage risk value exceeds the second load threshold, a warning rendering effect covering the corresponding joint area is generated in the user motion video stream.

[0180] This step aims to provide "explicit damage feedback." When the system detects the Risk after time integration... cum When the second load threshold is exceeded, it means that the soft tissue fatigue in the joint area has reached a dangerous level. Regardless of whether the current instantaneous posture is standard, continuing to move will easily cause injury.

[0181] At this point, the system no longer emphasizes a specific connection, but instead escalates the feedback to a regional blocking warning. The system will generate a semi-transparent layer in the video stream, covering the entire corresponding joint area (e.g., the entire knee joint).

[0182] This warning rendering effect typically manifests as a bright red pulsed flash (e.g., alternating between bright and dark at a frequency of 2Hz) or as a diffused ripple halo around the joint. This strong visual stimulus is designed to deliver a "stop immediately" or "must rest" signal (i.e., the "effect") to the user, forcibly interrupting the user's movement to prevent substantial injury.

[0183] For example, all rendering instructions can be drawn directly onto the original camera image using Unity or OpenGL, with rendering latency controlled within 10 milliseconds, ensuring that feedback is strictly synchronized with user actions.

[0184] Through the coordinated operation of steps S61 and S62, this application achieves a three-dimensional guidance system that "from point to surface, from cause to effect": in the early stages of risk, the "connecting marker lines" accurately point out the specific soft tissue locations where excessive force is applied, helping users to fine-tune and correct errors; in the later stages of risk, the "regional warnings" enforce the control of fatigue risk. This hierarchical feedback mechanism perfectly solves the shortcomings of existing technologies that "only report errors without providing guidance" and "ignore the accumulation of fatigue."

[0185] It is understood that the real-time motion error correction method of this application has the following beneficial effects:

[0186] 1. Achieved accurate identification of implicit compensatory risks of "correct form but incorrect force": By employing a "virtual biomechanical constraint model" and establishing a "personalized zero-load benchmark," this application can calculate the "elastic deformation load index," reflecting the tension of internal soft tissues, in real time based on sparse skeletal coordinates obtained from monocular vision. Thus, in the absence of contact sensors (such as electromyography patches), this application can penetrate the appearance of external geometric posture to infer the relative tension state of the human body's internal structures, effectively identifying implicit biomechanical risks caused by users who, although conforming to standards in geometric appearance, actually rely on incorrect muscle group exertion (such as compensatory mechanisms), overcoming the limitations of simply comparing geometric angles.

[0187] 2. Achieved fatigue accumulation and chronic injury prediction consistent with human physiological characteristics: By introducing a "damage dynamics evolution model," utilizing a "nonlinear mapping function" characterizing the tolerance properties of biological tissues, and "discrete integral operations" including "physiological recovery rate," this application can simulate the exponential growth of nonlinear damage in biological soft tissues under overload conditions, as well as the metabolic recovery mechanism under rest conditions, thereby establishing a risk assessment system with "time memory" capabilities. This enables the system not only to alert to instantaneous large-amplitude errors but also to keenly detect the "fatigue accumulation effect" caused by maintaining small deviations in posture for extended periods, effectively preventing chronic strain.

[0188] 3. Real-time biomechanical simulation with extremely low computational power consumption was achieved: By constructing a topological network composed of "rigid connection units" and "virtual elastic connection units," the traditional finite element analysis (FEM) mesh model was replaced. This application transforms the solution of complex differential equations into lightweight linear algebra and matrix operations, successfully solving the technical challenge of real-time internal force analysis under engineering constraints of "sparse input data (only coordinates)" and "limited computational resources (mobile terminal)," enabling high-precision biomechanical assessments to run smoothly at high frame rates on ordinary consumer electronic devices (such as mobile phones and smart TVs).

[0189] 4. Provides hierarchical visualization guidance with causal explanation: By establishing a dual evaluation system of "elastic deformation load index (space / cause)" and "cumulative damage risk value (time / effect)," and generating "connection marker lines" and "regional warning effects" respectively. This application breaks away from the black-box mode of traditional technologies that only provide binary "right / wrong" feedback, and can intuitively show users the specific anatomical source of the risk (i.e., "which muscle is tight") and the current overall fatigue level, realizing a leap from "simple error reporting" to "root cause error correction," significantly improving the user's training efficiency and safety.

[0190] Furthermore, this invention also proposes a computer-readable storage medium, which can be any one or any combination of several of the following: hard disk, multimedia card, SD card, flash memory card, SMC, read-only memory (ROM), erasable programmable read-only memory (EPROM), portable compact disc read-only memory (CD-ROM), USB memory, etc. The computer-readable storage medium includes a real-time motion correction program 10. The specific implementation of the computer-readable storage medium of this invention is largely the same as the specific implementation of the real-time motion correction method and the server 1 described above, and will not be repeated here.

[0191] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0192] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0193] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0194] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0195] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0196] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A real-time motion error correction method, characterized in that, include: The initial skeletal key point coordinates of the user in a preset calibration posture are obtained. Based on the preset virtual biomechanical constraint model, the initial length of each virtual elastic connection unit in the virtual biomechanical constraint model is determined as a personalized zero-load reference. Obtain the current real-time coordinates of the skeletal key points, calculate the real-time length of all virtual elastic connection units in the virtual biomechanical constraint model, and calculate the elastic deformation load index characterizing the tension of each virtual elastic connection unit based on the deformation degree of the real-time length relative to the personalized zero-load reference. For a preset joint region, the elastic deformation load index of all virtual elastic connection units associated with the joint region is weighted and aggregated to generate a regional aggregated pressure value that characterizes the current stress state of the joint region. Based on the regional aggregated pressure value, the instantaneous fatigue generation rate is calculated using a nonlinear mapping function that characterizes the tolerance properties of biological tissues, wherein the nonlinear mapping function defines an exponential growth relationship between the fatigue generation rate and the pressure intensity. Discrete integral operations are performed in the time dimension. When the aggregated pressure value of the region is higher than the preset safety tolerance threshold, the cumulative damage risk value at the current moment is updated based on the difference between the instantaneous fatigue generation rate and the preset physiological recovery rate. The elastic deformation load index and the cumulative damage risk value are compared with the corresponding preset thresholds to generate hierarchical causal feedback information containing the source identifier of elastic deformation load or the warning of cumulative damage risk, and then rendered and displayed.

2. The real-time motion error correction method as described in claim 1, characterized in that, The preset calibration posture is defined as the user being in a state of natural uprightness without active muscle contraction; The preset virtual biomechanical constraint model is constructed in the following way: The human skeleton vector is defined as a rigid connection unit with constant length; Based on human anatomy, virtual elastic connection units are established between adjacent joints and along cross-joint pathways, wherein the virtual elastic connection units are used to simulate the flexible connection relationships of human muscles, tendons or ligaments.

3. The real-time motion error correction method as described in claim 1, characterized in that, The elastic deformation load index, which characterizes the tension of each virtual elastic connection unit, is calculated, including: For each virtual elastic connection unit j in the virtual biomechanical constraint model, the elastic deformation load index is calculated using the following formula: In the formula, Index load,j Let L be the elastic deformation load index of the j-th virtual elastic connection unit. j L is the real-time length. base,j K is the corresponding initial length in the personalized zero-load reference. j This is the preset tissue stiffness coefficient.

4. The real-time motion error correction method as described in claim 1, characterized in that, Generating a region-specific aggregate pressure characterizing the current stress state of the joint region includes: The elastic deformation load index of all virtual elastic connection units associated with the joint region is obtained, and the basic pressure value is obtained by weighted summation using preset weighting coefficients. The geometric angle between rigid connection units within the joint area is detected in real time. When the geometric angle exceeds the preset safety angle range, a shear force penalty term is calculated based on the magnitude of the excess angle. The shear force penalty term is superimposed on the base pressure value to obtain the regional cohesive pressure.

5. The real-time motion error correction method as described in claim 4, characterized in that, The polymerization pressure in the region is calculated using the following formula: P region =∑ j∈Ω (w j ·Index load,j )+Penalty shear (θ), In the formula, P region The region represents aggregated pressure, Ω represents the set of virtual elastic connection units associated with the corresponding joint region, and w j For preset weight coefficients, Index load,j Penalty represents the elastic deformation load index of each virtual elastic connection unit. shear (θ) is the shear force penalty function based on the detected joint geometry angle θ.

6. The real-time motion error correction method as described in claim 1, characterized in that, The instantaneous fatigue generation rate is calculated using the following formula: Rate fatigue =A·exp(β·(P region -P threshold )), In the formula, Rate fatigue P represents the instantaneous fatigue generation rate; region P represents the polymerization pressure in the region. threshold The preset safety tolerance threshold is represented by A; the basic frequency factor is β; and the risk sensitivity coefficient is β.

7. The real-time motion error correction method as described in claim 1, characterized in that, Discrete integration is performed in the time dimension, using the following update rules: Risk cum (t)=Risk cum (t-1)+(Rate fatigue (t)-Rate recovery )·Δt, In the formula, Risk cum (t) represents the cumulative damage risk value at the current moment; Risk cum (t-1) represents the cumulative damage risk value at the previous time step; Rate fatigue (t) represents the instantaneous fatigue generation rate at the current moment; Rate recovery The preset physiological recovery rate is used to simulate the metabolic recovery process of the human body under low load conditions; Δt is the sampling time interval. The cumulative damage risk value is limited to a preset range [0,1].

8. The real-time motion error correction method as described in claim 1, characterized in that, Generate hierarchical causal feedback information that includes identification of elastic deformation load sources or warnings of cumulative damage risks, including: When any of the elastic deformation load indices exceeds the first load threshold, a visual connection marker line is rendered between the two skeletal key points connected by the corresponding virtual elastic connection unit in the user motion video stream. When the cumulative damage risk value exceeds the second load threshold, a warning rendering effect covering the corresponding joint area is generated in the user's motion video stream.

9. A real-time motion error correction device, characterized in that, The method includes a memory, a processor, and a real-time motion correction program stored in the memory and executable on the processor. When the processor executes the real-time motion correction program, it implements the real-time motion correction method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a real-time motion correction program, which, when executed by a processor, implements the real-time motion correction method as described in any one of claims 1-8.

Images