Wearable real-time feedback and motion correction system for swimming posture

Abstract

The present application belongs to the technical field of swimming posture correction, and specifically discloses a wearable swimming posture real-time feedback and action correction system. The present application comprehensively and stereoscopically evaluates the swimming posture from the dimensions of kinematics, hydrodynamics and biomechanics by fusing multi-modal sensor data and personalized hydrodynamic models, thereby improving the accuracy of swimming action analysis and effectively distinguishing between actions with high propulsion efficiency and invalid actions that generate excess resistance. Based on the tolerance range of dynamically adjusted action correction, the present application realizes an innovative change from fixed standards to adaptive judgment, provides more detailed guidance when the swimmer is in good physical condition, and prioritizes error correction when the swimmer is tired, thereby benefiting the scientificity and safety of the swimming process. By constructing a complete collaborative system from data acquisition, intelligent analysis, dynamic judgment to closed-loop feedback, the present application ensures the continuous effectiveness of the feedback and improves the overall training efficiency and intelligent level.

Description

Technical Field

[0001] This invention belongs to the field of swimming posture correction technology, specifically involving a wearable swimming posture real-time feedback and motion correction system. Background Technology

[0002] Swimming is a sport that demands extremely high technical skill; correct body posture, stroke trajectory, and leg kick rhythm are key to improving speed and conserving energy. To assist swimmers in improving their technique, various auxiliary training methods and equipment have emerged in the field of sports training. Among these, utilizing wearable electronic devices to monitor athletes' various physiological and athletic parameters in real time and provide data support has become an important development direction in modern scientific training.

[0003] Currently, common technical analysis methods in swimming training include visual observation by coaches and video playback analysis. In addition, wearable swimming devices have emerged on the market. These devices typically have built-in inertial sensors such as accelerometers, capable of recording basic data such as the number of strokes, stroke frequency, and distance per stroke. Some devices can sync data to an application after training for user review and analysis. A few devices have simple real-time reminder functions, such as vibration prompts to maintain the set stroke frequency.

[0004] However, existing technical means have obvious technical defects: (1) Traditional video analysis methods have high latency and cannot achieve real-time correction of actions.

[0005] (2) Most existing wearable devices are limited to recording and displaying data in a single dimension, lacking the ability to comprehensively analyze information from multiple sources. Furthermore, the feedback information they provide is relatively simple, usually using fixed standards to evaluate all users. They fail to consider individual physiological differences among swimmers and changes in their state due to decreased physical strength during a single training session, which greatly reduces the relevance and effectiveness of the feedback. Summary of the Invention

[0006] In view of this, in order to solve the problems mentioned in the background technology, a wearable swimming posture real-time feedback and motion correction system is proposed.

[0007] The objective of this invention can be achieved through the following technical solution: This invention provides a wearable swimming posture real-time feedback and motion correction system, including: a data acquisition module, a simulation model construction module, a theoretical motion data generation module, a correction command acquisition module, and a multimodal feedback signal acquisition module.

[0008] The data acquisition module is connected to the simulation model construction module, the simulation model construction module is connected to the theoretical motion data generation module, the theoretical motion data generation module is connected to the correction instruction acquisition module, and the correction instruction acquisition module is connected to the multimodal feedback signal acquisition module.

[0009] The data acquisition module collects real-time motion data, water resistance data, and muscle activity data through multimodal sensors deployed on the body of the target swimmer.

[0010] The simulation model building module constructs a personalized hydrodynamic simulation model of the pool where the target swimmer is located, based on real-time motion data, water resistance data, and muscle activity data.

[0011] The theoretical motion data generation module generates theoretical optimal motion data for the target swimmer in the current state based on a personalized hydrodynamic simulation model, and also generates dynamic tolerance thresholds.

[0012] The correction instruction acquisition module compares the real-time motion data with the theoretical optimal motion data to obtain the target swimmer's motion data deviation value, and determines the target swimmer's motion correction needs based on the dynamic tolerance threshold. If there is a motion correction need, a correction instruction is generated.

[0013] The multimodal feedback signal acquisition module generates multimodal feedback signals for real-time motion correction based on the correction command.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By integrating multimodal sensor data and personalized hydrodynamic models, the present invention overcomes the limitations of the prior art which relies solely on single kinematic data for analysis. It provides a comprehensive and three-dimensional evaluation of swimming posture from three dimensions: kinematics, hydrodynamics, and biomechanics, thereby improving the accuracy and depth of swimming motion analysis and effectively distinguishing between movements with high propulsion efficiency and ineffective movements that generate excessive resistance.

[0015] 2. Based on the dynamic adjustment of the tolerance range for motion correction, this invention achieves an innovative shift from fixed standards to adaptive evaluation, providing more refined guidance when swimmers are in good physical condition and prioritizing error correction when they are fatigued, thus making the swimming process more scientific and safer.

[0016] 3. This invention constructs a complete collaborative system from data collection, intelligent analysis, dynamic evaluation to closed-loop feedback, which helps to continuously evolve with the swimmer's technical progress and changes in condition, ensuring the continuous effectiveness of feedback and improving the overall training efficiency and intelligence level. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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 these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system module connections of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, this invention provides a wearable swimming posture real-time feedback and motion correction system, with the following specific module distribution: data acquisition module, simulation model construction module, theoretical motion data generation module, correction command acquisition module, and multimodal feedback signal acquisition module.

[0021] The data acquisition module is connected to the simulation model construction module, the simulation model construction module is connected to the theoretical motion data generation module, the theoretical motion data generation module is connected to the correction instruction acquisition module, and the correction instruction acquisition module is connected to the multimodal feedback signal acquisition module.

[0022] The data acquisition module collects real-time motion data, water resistance data, and muscle activity data through multimodal sensors deployed on the body of the target swimmer.

[0023] In a preferred feasible example of the present invention, the data acquisition module specifically includes: acquiring limb motion data as real-time motion data by deploying inertial measurement units at key parts of the target swimmer's body.

[0024] It should be noted that the inertial measurement unit integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. It outputs high-precision attitude information through sensor fusion algorithms such as Kalman filtering. It is fixed to key parts of the target swimmer's body, such as the wrist, upper arm, waist, lower leg, and ankle, and can capture the three-dimensional motion information of the target swimmer's limbs in real time.

[0025] The real-time motion data includes, but is not limited to, roll angle, pitch angle, yaw angle, elbow angle, shoulder angle, knee angle, stroke frequency, stroke cycle, and motion phase recognition.

[0026] The roll angle, or body roll angle, measures the angle of rotation of the swimmer's body around its longitudinal axis (head to toes) and is a key indicator for freestyle and backstroke techniques. The pitch angle, or body rotation around its lateral axis (left to right), reflects the body's streamlined horizontal position and the degree of head rise during breathing. The yaw angle, or body rotation around its vertical axis, is typically small; an excessively large yaw angle may indicate a serpentine forward motion. The elbow angle is used to determine whether a high elbow position is achieved during the catch phase; an ideal angle is approximately 90°-120°. The shoulder angle is used to assess the stroke amplitude and entry extension. The knee angle is used to analyze the amplitude and efficiency of the leg kick. The stroke frequency is the number of strokes per unit time. The stroke cycle is the time of one complete stroke. The motion phase recognition automatically identifies the stage of the stroke, such as entry, catch, push, exit, and recovery, using acceleration and angular velocity patterns.

[0027] Pressure sensors are deployed at key points on the swimmer's inward-facing surface to obtain water pressure information as data on water resistance changes.

[0028] It should be noted that the pressure sensor is fixed to key areas on the front side of the swimmer's swimsuit, such as the chest or back, hands, and forearms, where the swimsuit fits snugly against the body, and can capture the water pressure information of the pool where the swimmer is located in real time.

[0029] The water resistance change data includes relative pressure difference, pressure distribution map, pressure center trajectory, pressure change gradient, peak pressure, and pressure duration.

[0030] The relative pressure difference is the difference between the absolute pressure value and the hydrostatic pressure, directly reflecting the dynamic pressure generated by the movement. The pressure distribution map is the pressure distribution across the entire palm surface, plotted after detecting pressure from an array of pressure sensors placed on the hand or arm, used to find the optimal angle of attack, i.e., the palm orientation. The pressure center trajectory is the path of the pressure center point during the stroke, reflecting the stability of the propulsive force. The pressure change gradient is the rate of pressure change; a steep gradient usually indicates an efficient catch and hold motion. The peak pressure is the maximum pressure value that occurs during a single stroke, related to the magnitude of the propulsive force generated. The pressure duration is the duration of the efficient pressure phase, such as the push phase.

[0031] Electromyography (EMG) sensors are deployed in key muscle groups of target swimmers to acquire muscle electrical signals as muscle activity data.

[0032] It should be noted that the electromyography sensor electrode pads are attached to the surface of key muscle groups used by the target swimmer to drive swimming movements, such as the latissimus dorsi, deltoid, and quadriceps femoris in freestyle swimming, so that the surface electromyography sensor can non-invasively detect the weak bioelectrical signals generated by the muscles during contraction.

[0033] The muscle activity data includes rectified EMG signals, integrated electromyography values, root mean square values, muscle activation timing, and muscle synergy patterns.

[0034] The rectified EMG signal is obtained by inverting the negative portion of the original signal to a positive value, facilitating analysis. The integrated electromyography (iEMG) value is obtained by integrating the rectified EMG signal over a period of time, reflecting the total load of muscle activity and used to assess fatigue levels. The root mean square (RMS) value is calculated by taking the root mean square of the signal within a time window, reflecting the average amplitude of muscle activity during that time period. The muscle activation sequence is used to determine when a muscle begins to activate, reaches its peak, and deactivates during the stroke cycle. This is crucial for analyzing movement coordination, such as the activation sequence of the pectoralis major and latissimus dorsi. The muscle synergy mode analyzes the activation coordination relationship between different muscle groups through multi-channel EMG analysis.

[0035] The simulation model building module constructs a personalized hydrodynamic simulation model of the pool where the target swimmer is located, based on real-time motion data, water resistance data, and muscle activity data.

[0036] In a preferred feasible example of the present invention, the specific method of constructing the personalized hydrodynamic simulation model corresponding to the pool where the target swimmer is located includes: A1. Geometric construction: Based on the obtained human morphological parameters of the target swimmer and the basic parameters of the pool where it is located, a virtual avatar of the target swimmer and its virtual pool are constructed using professional software, and these are integrated to obtain a basic model of personalized hydrodynamic simulation.

[0037] A2. Coefficient Calibration: The resistance coefficient is calibrated in the virtual swimming pool based on water resistance data, and the propulsion coefficient of the virtual avatar is calibrated in the virtual swimming pool based on real-time motion data and muscle activity data.

[0038] A3. Simulation Model Determination: The personalized hydrodynamic simulation basic model calibrated with drag coefficient and propulsion coefficient is denoted as the personalized hydrodynamic model.

[0039] This invention overcomes the limitations of existing technologies that rely solely on single kinematic data for analysis by integrating multimodal sensor data with personalized hydrodynamic models. It provides a comprehensive and three-dimensional evaluation of swimming posture from three dimensions: kinematics, hydrodynamics, and biomechanics, improving the accuracy and depth of swimming motion analysis. It can effectively distinguish between movements with high propulsion efficiency and ineffective movements that generate unnecessary resistance.

[0040] The theoretical motion data generation module generates theoretical optimal motion data for the target swimmer in the current state based on a personalized hydrodynamic simulation model, and also generates dynamic tolerance thresholds.

[0041] In a preferred feasible example of the present invention, the theoretical motion data generation module specifically includes: importing the real-time motion data and water resistance data of the target swimmer as model inputs into a personalized hydrodynamic simulation model, performing simulation calculations, and outputting the theoretical optimal motion data of the target swimmer in the current state.

[0042] Specifically, the real-time motion data, water resistance data, and muscle activity data acquired in the previous stage are continuously input into the initialized personalized hydrodynamic simulation model. Real-time motion data drives the posture and movements of the virtual digital model, water resistance data is used to calibrate the fluid environment parameters in the model in real time, and muscle activity data provides the model with physiological constraints on the target swimmer's power efficiency and fatigue state. Driven by this multi-source data, the model performs simulation calculations, solving fluid dynamics control equations such as the Navier-Stokes equations to calculate the theoretically ideal motion parameters that achieve maximum propulsion efficiency and minimum resistance under the current swimming speed and body state. These parameters include, for example, the optimal stroke trajectory, entry angle, or kick frequency, and are recorded as the theoretically optimal motion data for the target swimmer in the current state.

[0043] Based on muscle activity data, the physiological fatigue state of the target swimmer is assessed, and a dynamic tolerance threshold is generated accordingly.

[0044] Specifically, the fatigue level of the target swimmer is assessed by analyzing muscle activity data and real-time motion data.

[0045] For example, when a shift in the electromyographic signal spectrum to lower frequencies is detected in muscle activity data, or when irregular fluctuations in stroke frequency are observed in real-time motion data without an increase in power, the swimmer is determined to be in a state of fatigue. Based on this judgment, the tolerance range of the theoretically optimal motion data is widened, thereby generating a dynamic tolerance threshold that better reflects the current physical condition. The adjustment of this threshold can follow the following logical relationship: ,in, This is the final generated dynamic tolerance threshold. The preset basic tolerance threshold, This is a dimensionless fatigue factor calculated based on real-time data. It is obtained by comprehensively analyzing the rate of change of signal characteristics in muscle activity data and the motion stability index in real-time motion data, and its value increases with the degree of fatigue. At the same time, the personalized hydrodynamic simulation model fine-tunes the theoretically optimal motion data itself to adapt to the motion mode under fatigue, and finally outputs the final theoretically optimal motion data that takes into account both theoretical optimization and practical feasibility.

[0046] It should be noted that the specific calculation process of the dimensionless fatigue factor includes: (1) calculating the real-time rate of change of parameters in muscle activity data: assuming that the median frequency baseline value of parameters in muscle activity data in the initial stage of training, i.e. when physical strength is sufficient, is The median frequency at the current moment is Then the real-time rate of change of this parameter for: During fatigue Decrease, therefore The value is positive, and the more severe the fatigue, The larger.

[0047] (2) Normalize the real-time rate of change of parameters in muscle activity data: ,in The minimum rate of change of this parameter recorded in historical training corresponds to the fatigue-free state. The maximum rate of change of this parameter recorded in historical training corresponds to the state of severe fatigue. This is the normalized real-time rate of change of the parameter, with a value range of [0,1]. The closer it is to 1, the higher the degree of fatigue reflected by the parameter.

[0048] (3) The real-time rate of change of each parameter in the muscle activity data after normalization is weighted and summed to obtain the dimensionless fatigue factor: ,in For the first The weights of each parameter, and satisfying Such as the median frequency weight of electromyography spectrum Stroke frequency variation coefficient weight wait, For the first Real-time rate of change of each parameter after normalization The value range is [0,1], dimensionless, and the fatigue degree is related to... Positive correlation, such as during mild fatigue When severely fatigued .

[0049] The correction instruction acquisition module compares the real-time motion data with the theoretical optimal motion data to obtain the target swimmer's motion data deviation value, and determines the target swimmer's motion correction needs based on the dynamic tolerance threshold. If there is a motion correction need, a correction instruction is generated.

[0050] In a preferred feasible example of the present invention, the specific method for obtaining the motion correction requirements of the target swimmer based on the dynamic tolerance threshold includes: comparing the differences between each parameter of the real-time motion data of the target swimmer and the corresponding parameters of the theoretical optimal motion data to obtain the initial deviation value of each parameter of the real-time motion data of the target swimmer.

[0051] The initial deviation values ​​of each parameter in the real-time motion data are filtered by a dynamic tolerance threshold. If the initial deviation value of any parameter is a valid deviation value, it is determined that the target swimmer has a need for motion correction; otherwise, it is determined that the target swimmer does not have a need for motion correction.

[0052] It should be noted that the specific process of the filtering is as follows: when the absolute value of the initial deviation of a certain parameter... When the initial deviation of the parameter is equal to the effective deviation, it is recorded as: When the absolute value of the initial deviation of a certain parameter If the initial deviation value of the parameter is not a valid deviation value, it is recorded as follows:

[0053] in It is a dynamic tolerance threshold, which reflects the current tolerance for errors in the operational data. This represents the absolute magnitude of the original deviation value. When the absolute value of a deviation exceeds the current tolerance threshold, it is determined to be a significant deviation requiring attention and is retained as a valid action deviation value. Conversely, if the absolute value of the deviation is within the tolerance threshold, it is considered a normal fluctuation in motion or an acceptable minor flaw, and its value is set to zero and ignored.

[0054] In a preferred feasible example of the present invention, the specific method of generating the correction instruction includes: B1, deviation root cause analysis: screening the parameters of the real-time motion data of the target swimmer whose initial deviation value is a valid deviation value, recording them as the abnormal motion parameters of the target swimmer, and matching them with the preset deviation root causes corresponding to each motion parameter, thereby determining the deviation root causes corresponding to each abnormal motion parameter of the target swimmer.

[0055] B2. Instruction content generation: Based on the root cause of the deviation, generate instruction content with clear guiding significance, and record it as a correction instruction.

[0056] B3. Command Priority Sorting: When multiple sources of deviation occur simultaneously, such as elbow dragging, insufficient leg kick amplitude, and breathing head lifting, the deviations are sorted according to the absolute value of the initial deviation and their weight in terms of impact on exercise efficiency, and the most critical deviations that have the greatest impact on overall performance are corrected first.

[0057] For example, correcting "raising the head too high when breathing" has a higher priority than correcting "slightly smaller leg kicks".

[0058] B4. Instruction Encoding: The generated correction instructions are encoded into a machine-readable standardized format for transmission to the feedback actuator.

[0059] The multimodal feedback signal acquisition module generates multimodal feedback signals for real-time motion correction based on the correction command.

[0060] In a preferred feasible example of the present invention, the specific method for generating a multimodal feedback signal for real-time motion correction includes: matching in a feedback strategy library and selecting a feedback mode according to the correction instruction.

[0061] Based on the selected feedback mode, a multimodal feedback signal containing either speech or tactile signals is generated.

[0062] After feeding back the multimodal feedback signal to the execution device, the data acquisition module is executed again, and this process is repeated until the target swimmer finishes swimming.

[0063] In a preferred feasible example of the present invention, the specific method of generating a multimodal feedback signal containing a voice signal or a tactile signal includes: when the feedback mode is a voice signal feedback, generating text prompt information based on a correction instruction and converting it into an audio signal.

[0064] When the feedback mode is tactile signal feedback, the vibration parameters are determined based on the correction command, and a drive signal is generated.

[0065] When the correction instruction has multiple associated error types, the audio signal and the driving signal are combined to generate a composite multimodal feedback signal.

[0066] This invention, based on dynamically adjusting the tolerance range of motion correction, achieves an innovative shift from fixed standards to adaptive evaluation. It provides more refined guidance when swimmers are energetic and prioritizes correcting errors when they are fatigued, making the swimming process more scientific and safer.

[0067] In a preferred feasible example of the present invention, after generating the theoretically optimal motion data and dynamic tolerance threshold, the method further includes: recording the historical abnormal motion parameters and multimodal feedback signal records of the target swimmer, generating long-term training data, and identifying the target swimmer's motion improvement trend and fatigue evolution pattern accordingly.

[0068] In a specific example, the process for identifying the trend of improvement in the action is as follows: track the change curve of the effective deviation value of a specific key action feature, such as the stroke angle, over time, and use trend analysis algorithms such as linear regression or moving average to determine whether the specific key action feature is improving, stagnating, or regressing.

[0069] The specific identification process of the fatigue evolution pattern is as follows: Analyze the correlation between muscle activity data such as changes in electromyographic signal power spectrum and real-time motion data such as movement frequency and amplitude stability during long-term training of the target swimmer, and find typical data patterns that predict the occurrence of fatigue.

[0070] For example, if the median frequency of the quadriceps electromyography signal of a target swimmer drops to a certain threshold, the consistency of its leg kick amplitude will be significantly reduced. In this case, the median frequency of the quadriceps electromyography signal and the leg kick amplitude are considered to be correlated.

[0071] Based on the trend of motion improvement and fatigue evolution, the parameters of the personalized hydrodynamic simulation model and the generation logic of the dynamic tolerance threshold are updated.

[0072] Specifically, if a certain action trend is identified as stagnant for a long period, the corresponding parameters of the personalized hydrodynamic simulation model will be adjusted. This involves attempting to find a breakthrough from a new biomechanical perspective or adjusting the feedback strategy for that action. The identified fatigue patterns will be solidified into new rules or model parameters to update the calculation logic of the dynamic tolerance threshold.

[0073] In a preferred feasible example of the present invention, the specific steps of identifying the target swimmer's movement improvement trend and fatigue evolution pattern include: extracting movement stability indicators for characterizing technical stability and fatigue indicators for characterizing physical condition from the long-term training data.

[0074] As a specific example, the motion stability indicators used to characterize technical stability include, but are not limited to, the coefficient of variation of key motion parameters, motion phase consistency indicators, and water resistance data stability indicators.

[0075] The coefficient of variation of the key action parameters is obtained by calculating the ratio of the standard deviation to the mean of the parameters in the historical data of the training data. A smaller coefficient of variation indicates more stable parameters; conversely, a consistently high coefficient of variation indicates random bias in the action.

[0076] The motion phase consistency index is obtained by measuring the duration fluctuation range of the motion phases of parameters in the training data, or the stability of the time proportion of each phase. High phase consistency indicates a standardized motion process; low consistency reflects a lack of smooth motion transitions.

[0077] The water resistance data stability index is obtained by analyzing the overlap of the pressure center trajectory of the movement in the training data, such as trajectory deviation during multiple water entry, or the variation range of the relative pressure difference peak value, such as peak fluctuation. Stable water resistance data indicates that the disturbance of the water flow by the movement is controllable; large fluctuations indicate that there is excessive resistance in the movement.

[0078] The fatigue indicators used to characterize physical fitness include, but are not limited to, electromyographic signal characteristics, exercise efficiency decay indicators, and movement stability correlation indicators.

[0079] The muscle electromyography (EMG) characteristic index is obtained by summing the iEMG values ​​of key muscle groups in a single training session. The higher the cumulative value, the greater the muscle load.

[0080] The exercise efficiency decay index compares the single stroke distance between the early and middle stages of training. For example, if the stroke distance is 1.8m / stroke in the early stage and decreases to 1.5m / stroke in the middle stage, the decay rate is 16.7%. Alternatively, it can be calculated as the ratio of stroke frequency to swimming speed. For example, if the stroke frequency increases while the swimming speed remains unchanged when fatigued, the ratio will increase. The faster the exercise efficiency decays, the faster the physical energy is consumed; if the decay rate is low, it indicates that the physical energy reserves are sufficient.

[0081] The movement stability correlation index is obtained by the proportion of low-frequency electromyographic signals of muscles, reflecting the degree of impact of fatigue on technical movements, and is a key basis for determining whether the feedback strategy needs to be adjusted.

[0082] The motion stability index and the fatigue index are processed using a pattern recognition algorithm to output a structured motion trend type and fatigue state level.

[0083] As a specific example, the pattern recognition algorithm includes, but is not limited to, time series regression algorithms, clustering and threshold determination algorithms.

[0084] It should be noted that the specific process of outputting the structured trend of action improvement includes: sorting the action stability index according to the training time axis to form time series data, and using time series regression algorithms, such as linear regression and moving average, to fit the above sequence data; if the fitted curve shows a continuous downward trend, such as the coefficient of variation decreasing from 0.3 to 0.1, and the phase consistency fluctuation range shrinking from ±0.5 seconds to ±0.2 seconds, it is determined to be an action improvement trend, indicating that the stability of the action is improving.

[0085] If the fitted curve shows a horizontal fluctuation trend, such as the coefficient of variation remaining at 0.25±0.02 for a long time, it is judged as a stagnant trend of movement, indicating that the movement technique has not broken through the bottleneck.

[0086] If the fitted curve shows a continuous upward trend, such as the coefficient of variation increasing from 0.2 to 0.35, it is determined to be a regression trend of the action, indicating that the stability of the action is decreasing.

[0087] The specific process of outputting the structured fatigue state level includes: normalizing the fatigue index to form a multi-dimensional fatigue feature vector; and using a clustering algorithm to perform unsupervised clustering on the fatigue feature vectors accumulated in historical training, automatically generating three clusters corresponding to mild fatigue, moderate fatigue, and severe fatigue, respectively.

[0088] Based on the clustering results, the threshold range of indicators for each cluster is extracted to form a judgment rule: if the current fatigue feature vector falls within the threshold range of the mild fatigue cluster, such as iEMG cumulative load 0.2-0.4 and paddling distance attenuation rate 0.1-0.2, it is judged as mild fatigue.

[0089] If it falls within the threshold range of moderate fatigue, such as iEMG cumulative load of 0.4-0.7 and attenuation rate of 0.2-0.5, it is judged as moderate fatigue.

[0090] If it falls within the threshold range of severe fatigue cluster, and if the iEMG cumulative load is >0.7 and the attenuation rate is >0.5, it is judged as severe fatigue.

[0091] The feedback priority in the feedback strategy library is adjusted based on the action trend type and fatigue level.

[0092] As a specific example, the fatigue state levels include mild fatigue, moderate fatigue, and severe fatigue.

[0093] It should be noted that the priority adjustment based on the action improvement trend specifically includes: (1) Action improvement trend: If the trend of an action is an upward trend, the feedback priority of the action is reduced, and resources are instead allocated to other stagnant or regressing indicators.

[0094] For example, the original priority of "stroke rhythm deviation" feedback was 3, with 1 being the highest. After improvement, it was reduced to 5, and the priority of feedback was "holding phase duration deviation", i.e. stagnation trend, which was raised to 2.

[0095] (2) Stagnation trend of a certain action: If the trend of a certain action is to remain unchanged for a long time, increase its feedback frequency and priority, and adjust the feedback content.

[0096] (3) Action regression trend: If the trend of a certain action is regression, it will be given the highest priority feedback, and the feedback content will focus on root cause investigation.

[0097] The priority adjustment based on fatigue level specifically includes: (1) Mild fatigue: Maintain normal priority. (2) Moderate fatigue: Increase the weight of safety indicators and prioritize feedback on deviations that may lead to excessive physical exertion or imbalance. (3) Severe fatigue: Only retain feedback on key safety or energy-saving deviations and simplify the feedback content.

[0098] This invention constructs a complete collaborative system from data acquisition, intelligent analysis, dynamic evaluation to closed-loop feedback, which helps to continuously evolve with the swimmer's technical progress and changes in condition, ensuring the continuous effectiveness of feedback and improving the overall training efficiency and intelligence level.

[0099] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A wearable swimming posture real-time feedback and motion correction system, characterized in that: include: The data acquisition module collects real-time motion data, water resistance data, and muscle activity data through multimodal sensors deployed on the body of the target swimmer; The simulation model building module constructs a personalized hydrodynamic simulation model of the pool where the target swimmer is located, based on real-time motion data, water resistance data, and muscle activity data. The theoretical motion data generation module generates theoretical optimal motion data of the target swimmer in the current state based on a personalized hydrodynamic simulation model, and at the same time generates dynamic tolerance thresholds. The specific contents of the theoretical motion data generation module include: The real-time motion data and water resistance data of the target swimmer are imported into the personalized hydrodynamic simulation model as model input, and simulation calculations are performed to output the theoretical optimal motion data of the target swimmer in the current state. Based on muscle activity data, assess the physiological fatigue state of the target swimmer and generate a dynamic tolerance threshold accordingly. The dynamic tolerance threshold reflects the current tolerance for action errors in the operating data; The correction instruction acquisition module compares the real-time motion data with the theoretical optimal motion data to obtain the motion data deviation value of the target swimmer, and determines the motion correction needs of the target swimmer based on the dynamic tolerance threshold. If there are motion correction needs, a correction instruction is generated. The specific methods for obtaining the target swimmer's motion correction needs based on dynamic tolerance thresholds include: The parameters of the target swimmer's real-time motion data are compared with the corresponding parameters of the theoretical optimal motion data to obtain the initial deviation values ​​of each parameter of the target swimmer's real-time motion data; The initial deviation values ​​of each parameter of the real-time motion data are filtered by a dynamic tolerance threshold. If the initial deviation value of any parameter is a valid deviation value, it is determined that the target swimmer has a need for motion correction. Otherwise, it is determined that the target swimmer does not have a need for motion correction. The specific process of the filtering is as follows: when the absolute value of the initial deviation of a certain parameter is... When the initial deviation of the parameter is equal to the effective deviation, it is recorded as: When the absolute value of the initial deviation of a certain parameter If the initial deviation value of the parameter is not a valid deviation value, it is recorded as follows: ; in It is a dynamic tolerance threshold, which reflects the current tolerance for action errors in the running data; This represents the absolute magnitude of the original deviation value. When the absolute value of a deviation exceeds the current tolerance threshold, it is determined to be a significant deviation that requires attention and is retained as a valid action deviation value. ; The multimodal feedback signal acquisition module generates multimodal feedback signals for real-time motion correction based on the correction command.

2. The wearable swimming posture real-time feedback and motion correction system according to claim 1, characterized in that: The specific contents of the data acquisition module include: The inertial measurement units deployed at key parts of the target swimmer's body acquire limb motion data as real-time motion data. Pressure sensors are deployed at key points on the frontal surface of the target swimmer to obtain water pressure information as data on water resistance changes. Electromyography (EMG) sensors are deployed in key muscle groups of target swimmers to acquire muscle electrical signals as muscle activity data.

3. The wearable swimming posture real-time feedback and motion correction system according to claim 1, characterized in that: The specific methods for constructing a personalized hydrodynamic simulation model corresponding to the swimming pool of the target swimmer include: A1. Geometric Construction: Based on the acquired anthropometric parameters of the target swimmer and the basic parameters of the pool in which they are located, a virtual avatar of the target swimmer and the virtual pool in which they are located are constructed using professional software. These are then integrated to obtain a personalized hydrodynamic simulation basic model. A2. Coefficient calibration: The resistance coefficient is calibrated in the virtual swimming pool based on water resistance data, and the propulsion coefficient of the virtual avatar is calibrated in the virtual swimming pool based on real-time motion data and muscle activity data. A3. Simulation Model Determination: The personalized hydrodynamic simulation basic model calibrated with drag coefficient and propulsion coefficient is denoted as the personalized hydrodynamic model.

4. The wearable swimming posture real-time feedback and motion correction system according to claim 1, characterized in that: The specific methods for generating correction instructions include: B1. Deviation Root Cause Analysis: Screen the parameters belonging to the real-time motion data of the target swimmer whose initial deviation value is a valid deviation value, record them as the abnormal motion parameters of the target swimmer, and match them with the preset deviation root causes of each motion parameter to determine the deviation root causes of each abnormal motion parameter of the target swimmer. B2. Instruction content generation: Based on the root causes of the deviation, generate instruction content with clear guiding significance and record it as a correction instruction; B3. Command Priority Sorting: When multiple sources of deviation occur simultaneously, they are sorted according to the absolute value of the initial deviation and the weight of their impact on motion efficiency, with priority given to correcting the most critical deviation that has the greatest impact on overall performance. B4. Instruction Encoding: The generated correction instructions are encoded into a machine-readable standardized format for transmission to the feedback actuator.

5. The wearable swimming posture real-time feedback and motion correction system according to claim 1, characterized in that: The specific methods for generating multimodal feedback signals for real-time motion correction include: Based on the correction instructions, match the feedback strategy library and select the feedback mode; Based on the selected feedback mode, a multimodal feedback signal containing either voice or tactile signals is generated. After feeding back the multimodal feedback signal to the execution device, the data acquisition module is executed again, and this process is repeated until the target swimmer finishes swimming.

6. The wearable swimming posture real-time feedback and motion correction system according to claim 5, characterized in that: The specific methods for generating multimodal feedback signals containing speech or tactile signals include: When the feedback mode is voice signal feedback, text prompts are generated based on the correction instructions and converted into audio signals. When the feedback mode is tactile signal feedback, the vibration parameters are determined based on the correction command, and a drive signal is generated; When the correction instruction has multiple associated error types, the audio signal and the driving signal are combined to generate a composite multimodal feedback signal.

7. The wearable swimming posture real-time feedback and motion correction system according to claim 5, characterized in that: After generating the theoretically optimal motion data and dynamic tolerance threshold, the following is also included: Record the historical abnormal motion parameters and multimodal feedback signals of the target swimmer to generate long-term training data, and identify the trend of motion improvement and fatigue evolution patterns of the target swimmer based on this data. Based on the trend of motion improvement and fatigue evolution, the parameters of the personalized hydrodynamic simulation model and the generation logic of the dynamic tolerance threshold are updated.

8. The wearable swimming posture real-time feedback and motion correction system according to claim 7, characterized in that: The specific steps for identifying the movement improvement trend and fatigue evolution pattern of the target swimmer include: From the long-term training data, we extract motion stability indicators to characterize technical stability and fatigue indicators to characterize physical condition. The motion stability index and the fatigue index are processed using a pattern recognition algorithm to output a structured motion improvement trend and fatigue level. The feedback priority in the feedback strategy library is adjusted based on the action trend type and fatigue level.

Images