Swimming trainer resistance control system based on training data

Abstract

The invention belongs to the technical field of swimming assistance, and discloses a swimming trainer resistance control system based on training data, comprising a multi-source data capture layer for capturing swimmer training data in real time through an integrated sensor network, including paddling propulsive force data, action trajectory data, water flow resistance distribution data and physiological parameter data; a recognition layer is extracted, a motion decomposition algorithm is applied to carry out recognition and segmentation on the paddling propulsive force data and the motion trajectory data, n paddling periods and corresponding paddling technical parameters are obtained, a muscle fluid coupling matrix is constructed, and a muscle fluid phase matching curve in each paddling period is obtained; individualized training data are predicted and generated based on the muscle fluid phase matching curve and the physiological parameter data; according to the invention, comprehensive, intelligent and personalized training can be provided for swimmers, the training quality is improved, sports injuries are reduced, and the swimmers are helped to achieve the best training effect.

Description

Technical Field

[0001] The present invention relates to the technical field of swimming assistance, and more specifically, to a resistance control system for a swimming trainer based on training data. Background Art

[0002] A patent with the publication number CN115738163A discloses a magnetic control resistance system for fitness equipment, which includes a human-machine interface, a main controller, a rotational speed sensor, a programmed constant current source, an electromagnet, and a magnetic control flywheel. The human-machine interface is used to set the resistance control mode and input the control mode parameters into the main controller. The rotational speed sensor is used to collect the rotational speed of the magnetic control flywheel and transmit the rotational speed parameters to the main controller. The main controller is used to calculate and generate control parameters based on the control mode parameters and the rotational speed parameters of the magnetic control flywheel, and transmit the control parameters to the programmed constant current source. The programmed constant current source is electrically connected to the electromagnet and is used to control the excitation current of the electromagnet to change according to the control parameters calculated by the main controller, thereby controlling the magnetic field strength of the electromagnet. The electromagnet is magnetically attracted to the magnetic control flywheel and is used to control the resistance of the magnetic control flywheel. This invention not only has the advantages of quietness, smooth resistance, and small volume of magnetic resistance, but also can make the magnetic control flywheel provide a magnetic control fluid resistance close to the resistance of rowing in water.

[0003] The existing resistance control systems for swimming trainers mainly have the following problems:

[0004] The start and end of the stroke cycle are often determined by the zero-crossing point of the first derivative of the instantaneous stroke force. However, in actual training, especially for beginners or in a fatigued state, the movements of swimmers may be irregular, and the characteristics of the stroke cycle vary greatly. Traditional zero-crossing detection methods are prone to misjudgment. Specifically, slight force or trajectory changes may cause the derivative zero-crossing point, which is thus wrongly marked as the cycle boundary, or the cycle fluctuations caused by irregular movements are misidentified as complete stroke cycles.

[0005] The changes in stroke data of different swimmers or in different exercise states may cause the standardized cycle recognition threshold to be unable to adapt to the situation of each athlete. Existing technologies may define the cycle boundary through a unified threshold, but this method cannot flexibly adapt to factors such as individual differences and fatigue states. Irregular movements, slight movement changes, and static swings in the movements of beginners or in a fatigued state may interfere with the determination of the stroke cycle, making it difficult to accurately identify the boundary of the stroke cycle. Existing technologies have not effectively excluded the errors brought by these irregular movements.

[0006] In view of this, the present invention proposes a resistance control system for a swimming trainer based on training data to solve the above problems. Summary of the Invention

[0007] To overcome the above defects of the prior art and to achieve the above object, the present invention provides the following technical solution: A resistance control system for a swimming trainer based on training data, comprising:

[0008] A multi-source data capture layer that captures the training data of the swimmer in real time through an integrated sensor network, including stroke propulsion force data, motion trajectory data, water flow resistance distribution data, and physiological parameter data;

[0009] An extraction and recognition layer that applies an action decomposition algorithm to identify and segment the stroke propulsion force data and the motion trajectory data, obtains n stroke cycles and corresponding stroke technical parameters, constructs a muscle-fluid coupling matrix, and obtains a muscle-fluid phase matching curve in each stroke cycle; predicts and generates individualized training data based on the muscle-fluid phase matching curve and the physiological parameter data;

[0010] A magnetic control resistance generation layer that constructs a controllable magnetic field based on the individualized training data, the motion trajectory data, and the water flow resistance distribution data, and generates a resistance matching the stroke technical parameters based on the individualized training data to adjust the resistance on the swimmer's stroke path in real time;

[0011] A resistance dynamic adjustment layer that identifies the current stroke cycle, and based on the water flow resistance distribution data, controls the resistance to change dynamically with time in each stage within the current stroke cycle, obtains the resistance change rate, and compares it with a preset optimal resistance change rate threshold. When the preset optimal resistance change rate threshold is reached, the current resistance output is maintained until the end of the current stroke cycle;

[0012] A training mode matching layer that receives the individualized training data and the resistance change rate, matches and executes corresponding training strategies; automatically sets training target parameters and a phased resistance distribution curve according to the training strategies, and outputs a personalized control plan.

[0013] Preferably, the method for capturing the training data of the swimmer includes:

[0014] The integrated sensor network is composed of m types of sensors, including force sensors, accelerometers, gyroscopes, water flow velocity sensors, physiological parameter sensors, and position sensing devices; the m types of sensors are integrated at key parts of the trainer through waterproof encapsulation. The key parts include stroke handles, underwater brackets, and wearable parts worn on the swimmer's body, forming a multi-point collaborative integrated sensor network; the wearable parts include the chest, upper arms, ankles, and back;

[0015] The m types of sensors are connected to the multi-source data capture layer by wired or wireless means, and a unified clock synchronization protocol is set to keep the timestamps of the captured training data of the swimmer consistent, and the training data of the swimmer is collected through the m types of sensors;

[0016] The stroke propulsion data include instantaneous stroke force, average stroke force, maximum stroke force, action time, thrust direction and stroke power; the motion trajectory data include three-dimensional position data, acceleration, angular velocity, trajectory path length and stroke time; the water flow resistance distribution data include the total resistance size, the change in water flow resistance and the resistance size under different movements; the physiological parameter data include the swimmer's heart rate, respiratory rate, blood oxygen saturation, muscle force data and body surface temperature.

[0017] Preferably, the method for obtaining the n stroke cycles and stroke technical parameters includes:

[0018] The instantaneous stroke force in the propulsion data is defined as a time series. The sliding window method is used to analyze the instantaneous stroke force. The sliding window size is preset to w. Within each sliding window, the first-order derivative of the instantaneous stroke force is calculated to obtain the rate of change of the instantaneous stroke force over time.

[0019] Identify the beginning and end of the stroke cycle based on the rate of change of the propulsive force over time. Mark the candidate boundaries of the stroke cycle by finding the zero crossing point of the first-order derivative of the instantaneous stroke force. Preliminarily determine the beginning and end of the stroke cycle through the candidate boundaries of the stroke cycle. The zero crossing point indicates that the rate of change of the propulsive force changes from positive to negative or from negative to positive.

[0020] The stroke cycle begins when the rate of change of propulsion force with time changes from negative to positive for the first time; the stroke cycle ends when the rate of change of propulsion force with time changes from positive to negative for the first time.

[0021] A coupling similarity function between the instantaneous stroke force and the motion trajectory is introduced to improve the accuracy of stroke cycle boundary determination. The coupling similarity function is: Where α represents the weight coefficient of the instantaneous stroke force change; β represents the weight coefficient of the motion trajectory acceleration; dF(t) represents the first-order derivative of the instantaneous stroke force; F(t) represents the instantaneous stroke force; It represents the rate of change of instantaneous paddling force over time; represents the acceleration of the motion trajectory; P(t) represents the three-dimensional coordinates of the swimmer's body position, P(t) = [x(t), y(t), z(t)]; t represents the variable index of the time point; ||·|| represents the Euclidean norm of the acceleration of the motion trajectory;

[0022] A minimum stroke cycle duration threshold is preset, and stroke cycle identification results that are shorter than the preset minimum stroke cycle duration threshold are excluded. The difference between the maximum instantaneous stroke force and the minimum instantaneous stroke force is compared with the preset minimum stroke cycle duration threshold. When the difference between the maximum instantaneous stroke force and the minimum instantaneous stroke force is greater than the preset minimum stroke cycle duration threshold, it is determined that a valid stroke action exists within the current time sliding window;

[0023] Introduce the coefficient of variation of the current stroke cycle to dynamically adjust the preset minimum stroke cycle duration threshold. Define the duration of all identified stroke cycles as T, the standard deviation of the durations of all stroke cycles as σ T , and the mean of the durations of all stroke cycles as μ T . Divide the standard deviation σ of the durations of all stroke cycles T by the mean μ of the durations of all stroke cycles T to obtain the coefficient of variation CV;

[0024] When the coefficient of variation CV is greater than the preset minimum stroke cycle duration threshold, it is determined that the current stroke cycle has large fluctuations, and the preset minimum stroke cycle duration threshold is dynamically adjusted through a dynamic threshold adjustment function, θ(t) = γ·μ T +(1 - γ)·μ ΔF ; where θ(t) represents the adaptive cycle recognition threshold at time point t; γ represents the allocation adjustment factor; μ ΔF represents the mean of the instantaneous stroke force fluctuations; ΔF represents the instantaneous stroke force fluctuations;

[0025] For each stroke cycle, calculate the instantaneous stroke force peak, stroke cycle duration, maximum instantaneous stroke force, and minimum instantaneous stroke force to obtain stroke technical parameters.

[0026] Preferably, the method for obtaining the muscle-fluid phase matching curve includes:

[0027] Represent the muscle force data in the physiological parameter data as a time-series muscle activation vector M(t). Concatenate the stroke propulsion force data and the motion trajectory data, collectively referred to as fluid response data, and represent the fluid response data as a fluid response vector F(t). Calculate the degree of cooperation between the muscle force data and the fluid response data within each stroke cycle;

[0028] Through the inner product operation of the muscle activation vector M(t) and the fluid response vector F(t) for each stroke cycle, obtain the muscle-fluid coupling matrix within each stroke cycle, perform normalization processing on the muscle-fluid coupling matrix within each stroke cycle, calculate the Frobenius norm of the muscle-fluid coupling matrix within each stroke cycle, and unify the scales of the muscle-fluid coupling matrices of different stroke cycles to obtain the normalized muscle-fluid coupling matrix;

[0029] Perform eigenvalue decomposition on the muscle-fluid coupling matrix to extract the principal coupling direction between muscle force data and fluid response data. Obtain the eigenvalues and eigenvectors of the matrix through eigenvalue decomposition. The eigenvalues represent the coupling strength between muscle force data and fluid response data, and the eigenvectors represent the principal directions of coupling. Take the eigenvector corresponding to the largest eigenvalue as the principal coupling direction. The largest eigenvalue represents the strongest coupling strength between muscle force data and fluid response data at any time point t. Arrange the largest eigenvalues at each time point t in chronological order to obtain the muscle-fluid phase matching curve.

[0030] Preferably, the method for obtaining the individualized training data includes:

[0031] Use a multi-modal Transformer to construct a training data prediction model. The training data prediction model includes an input layer, an encoding layer, a multi-modal fusion layer, a decoding layer, and a multi-task output layer. Use the historical muscle-fluid phase matching curve and physiological parameter data as the input data of the input layer, and output the corresponding individualized training data through the multi-task output layer. The individualized training data includes action patterns, force curve change trends, and fatigue levels. The action patterns include different swimming strokes, stroke forces, and stroke frequencies.

[0032] Use multi-class cross-entropy as the loss function of the model to optimize the training data prediction model until it stops when reaching the preset number of iterations, and obtain the trained training data analysis model. Input the current muscle-fluid phase matching curve and physiological parameter data into the trained training data analysis model to predict the individualized training data.

[0033] Preferably, the method for constructing the controllable magnetic field includes:

[0034] Based on the action trajectory data and water flow resistance distribution data, calculate the influence of the swimmer on the water flow in each stroke cycle through the CFD fluid dynamics model. Calculate a water flow resistance pattern based on the influence of the swimmer on the water flow in each stroke cycle. The water flow resistance pattern includes the magnitude of the water flow resistance, the direction of the water flow resistance, and the three-dimensional coordinate area of the water flow resistance distribution. Adjust the training load of the swimmer according to the force curve change trend and fatigue level to match the physiological parameter data of each swimmer. Based on the water flow resistance pattern, construct a controllable magnetic field through the magnetic control resistance unit.

[0035] Preferably, the method for real-time adjustment of the resistance on the swimmer's stroke path includes:

[0036] Based on individualized training data, a resistance model matching the stroke technique parameters is constructed through the LFS hydrodynamic model. The motion trajectory data of the swimmer is input into the LFS hydrodynamic model to simulate the disturbance and flow velocity distribution of the water flow under the stroke action, analyze the influence of the swimmer on the water flow in each stroke cycle, and output water flow resistance data according to the water flow resistance pattern to match the resistance on the swimmer's stroke path. The resistance on the swimmer's stroke path is adjusted in real time through a fuzzy control algorithm, and fuzzy rules are defined. The fuzzy rules include Rule 1 and Rule 2;

[0037] Rule 1 is that if the change error between the change rate of the swimmer's force curve and the change rate of the preset swimmer's force curve is greater than the preset change error threshold, and the fatigue degree of the swimmer is greater than the preset fatigue degree threshold, then the resistance on the swimmer's stroke path is reduced; Rule 2 is that if the change error between the change rate of the swimmer's force curve and the change rate of the preset swimmer's force curve is less than or equal to the preset change error threshold, and the fatigue degree of the swimmer is less than or equal to the preset fatigue degree threshold, then the resistance on the swimmer's stroke path is increased.

[0038] Preferably, the method for obtaining the resistance change rate includes:

[0039] Based on the motion trajectory data and the stroke propulsion force data, the current stroke cycle is identified in real time, and the stroke cycle is divided into a stroke preparation stage, an entry stage, a pulling stage, a propulsion stage, and an exit stage;

[0040] Within each stage of each stroke cycle, by analyzing the change of the water flow resistance distribution data over time, the resistance change rate within each stage is calculated where, ΔR represents the change amount of the water flow resistance; Δt represents the time interval corresponding to the change amount of the water flow resistance.

[0041] Preferably, the method for matching and executing the corresponding training strategy includes:

[0042] According to the individualized training data of the swimmer and the resistance change rate, different training strategies are formulated; the resistance control terminal of the swimming trainer automatically selects the training strategy suitable for the current training stage and executes the corresponding training strategy, continuously monitors the physiological parameter data of the swimmer, and adjusts the training strategy according to the real-time physiological parameter data; the training strategies include technical optimization training strategies, strength enhancement training strategies, endurance maintenance training strategies, and rehabilitation training strategies.

[0043] Preferably, the method for outputting the personalized control scheme includes:

[0044] According to the real-time physiological parameter data of the swimmer, the training goals are automatically set, and the training goals include strength training goals, technical improvement goals, and fatigue recovery goals;

[0045] After determining the training goal, the resistance control terminal of the swimming trainer automatically generates a phased resistance distribution curve based on the real-time individualized training data and training goal of the swimmer; depicts the resistance changes in each stage within each stroke cycle through the phased resistance distribution curve, and automatically generates a corresponding personalized control plan based on the training goal and the phased resistance distribution curve; the personalized control plan includes resistance setting, training time, stage arrangement, and feedback mechanism.

[0046] Technical effects and advantages of the resistance control system of the swimming trainer based on training data of the present invention:

[0047] By combining the instantaneous stroke force and the change of the motion trajectory acceleration, the coupling similarity function is used to improve the accuracy of the stroke cycle boundary determination. This method effectively avoids the misjudgment problem caused by the zero crossing point of the first derivative of the instantaneous stroke force alone. By comprehensively considering the force change and the motion trajectory acceleration, the reliability and accuracy of the boundary determination are enhanced. By introducing the coupling similarity function of the instantaneous stroke force and the motion trajectory, compared with the prior art that solely relies on the zero crossing point of the first derivative of the instantaneous stroke force, it can better handle the interference caused by fatigue, beginners, or irregular movements, making the identification of the stroke cycle boundary more accurate.

[0048] The coefficient of variation of the stroke cycle duration is introduced to dynamically adjust the minimum stroke cycle duration threshold to adapt to the cycle fluctuations of different swimmers or different motion states. Based on the duration distribution and standard deviation of the stroke cycle, the system can flexibly adjust the threshold according to the current state, avoiding the deviation caused by a fixed threshold and improving the robustness of cycle identification. It overcomes the problem in the prior art that a fixed threshold cannot cope with different motion states and has higher adaptability and accuracy;

[0049] By analyzing the standard deviation and mean value of all stroke cycles, the threshold is dynamically adjusted, so that the cycle identification in different states (such as beginners, fatigue states, different swimming strokes, etc.) can be effectively optimized, avoiding the problem that a fixed threshold cannot adapt to different situations. By reasonably presetting the threshold limit for the stroke cycle duration, the stroke cycles shorter than the preset minimum stroke cycle duration are excluded, reducing the situation of misidentifying small cycle fluctuations as complete cycles. By using the coupling similarity function to process the similarity between the propulsive force and the trajectory data, irregular movements can be effectively identified and filtered, avoiding the problem of cycle misidentification caused by slight changes, especially suitable for training beginners or in a fatigue state. Brief Description of the Drawings

[0050] Figure 1 It is a schematic structural diagram of the resistance control system of the swimming trainer based on training data of the present invention;

[0051] Figure 2Schematic flow diagram of the resistance control method for a swimming trainer based on training data according to the present invention. Detailed implementation manners

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment 1

[0054] Please refer to Figure 1 As shown, the first embodiment further describes the resistance control system for a swimming trainer based on training data proposed by the present invention, including:

[0055] With the continuous development of swimming training technology, modern swimming training has gradually developed towards the direction of personalization, high efficiency and intelligence. In order to help swimmers improve their training effects, the prior art has begun to use various sensors and intelligent devices to monitor data such as the movements, strengths, and physical states of swimmers, and adjust the training intensity and training methods through the feedback of these data. However, the existing training systems still face the following technical challenges, especially in the aspect of stroke cycle recognition and precise adjustment of training data.

[0056] Difficulty in stroke cycle recognition: In swimming training, accurately recognizing the stroke cycle of a swimmer is a core issue. Traditional stroke cycle recognition methods usually rely on the zero crossing point of the first derivative of the instantaneous stroke force to determine the start and end of the stroke cycle. Although this method is effective in certain situations, in practical applications, especially in beginners or fatigue states, the movements of swimmers may be irregular, and the characteristics of the stroke cycle vary greatly, resulting in fuzzy boundary determination. The characteristics of the stroke cycle vary greatly under different swimming strokes, and it is difficult for traditional methods to uniformly recognize them, resulting in deviations in the candidate boundaries of the stroke cycle. In addition, in the case of fatigue or beginners, the stroke movements of swimmers may be irregular, resulting in difficulty in determining the boundaries of the stroke cycle.

[0057] Inability to adapt to individual differences and dynamic adjustment of training states: Traditional stroke cycle recognition methods usually use a unified threshold to define the cycle boundary and cannot be adaptively adjusted according to the individual differences and motion states (such as fatigue, stroke changes, etc.) of each swimmer. This makes it possible that the system may not obtain accurate cycle divisions when processing the stroke data of different athletes and different training stages, thus affecting subsequent training adjustment and effect evaluation.

[0058] Irregular motion interference period recognition: Beginners may have irregular movements during training, and movements in a fatigued state may also be irregular or loose, resulting in errors in the recognition of the stroke cycle. Existing technologies may not be able to effectively eliminate the misidentifications caused by these irregular movements, affecting the accuracy and effectiveness of training data.

[0059] Imprecise adjustment of training data: In addition to the cycle recognition problem, when existing training systems dynamically adjust the training load according to data such as the strength, fatigue level, and movement pattern of the swimmer, they may not fully consider the real-time changes in the motion state. Traditional systems usually rely on fixed adjustment rules and cannot adjust the training intensity and resistance distribution in real time and precisely according to the actual conditions and technical characteristics of the athlete, thus affecting the training effect.

[0060] To overcome the above problems, the present invention proposes a resistance control system for a swimming trainer based on training data, including:

[0061] A multi-source data capture layer that captures the training data of the swimmer in real time through an integrated sensor network, including stroke propulsion force data, motion trajectory data, water flow resistance distribution data, and physiological parameter data;

[0062] An extraction and recognition layer that applies an action decomposition algorithm to identify and segment the stroke propulsion force data and motion trajectory data to obtain n stroke cycles and corresponding stroke technical parameters, constructs a muscle-fluid coupling matrix, and obtains the muscle-fluid phase matching curve in each stroke cycle; predicts and generates individualized training data based on the muscle-fluid phase matching curve and physiological parameter data;

[0063] A magnetic control resistance generation layer that constructs a controllable magnetic field through the individualized training data, motion trajectory data, and water flow resistance distribution data, and generates a resistance matching the stroke technical parameters based on the individualized training data to adjust the resistance on the swimmer's stroke path in real time;

[0064] A resistance dynamic adjustment layer that identifies the current stroke cycle, and based on the water flow resistance distribution data, controls the resistance to change dynamically with time in each stage of the current stroke cycle, obtains the resistance change rate, and compares it with a preset optimal resistance change rate threshold. When the preset optimal resistance change rate threshold is reached, the current resistance output is maintained until the end of the current stroke cycle;

[0065] A training mode matching layer that receives the individualized training data and the resistance change rate, matches and executes corresponding training strategies; automatically sets training target parameters and a phased resistance distribution curve according to the training strategies, and outputs a personalized control plan.

[0066] The method for capturing the training data of the swimmer includes:

[0067] The integrated sensor network consists of m types of sensors, including force sensors, accelerometers, gyroscopes, water flow velocity sensors, physiological parameter sensors (such as heart rate monitors, blood oxygen sensors, and electromyography sensors), and position sensing devices (such as inertial measurement units IMU and UWB positioning modules); the m types of sensors are integrated at key parts of the trainer through waterproof encapsulation, and the key parts include the water - paddling handle, the underwater bracket, and the wearable parts worn on the swimmer's body, forming a multi - point collaborative integrated sensor network; the wearable parts include the chest, upper arm, ankle, and back.

[0068] The m types of sensors are connected to the multi - source data capture layer by wired or wireless means (such as Bluetooth, Wi - Fi, or ZigBee), and a unified clock synchronization protocol is set to keep the timestamps of the captured swimmer training data consistent, and the swimmer training data is collected through the m types of sensors.

[0069] The water - paddling propulsion force data includes instantaneous water - paddling force, average water - paddling force, maximum water - paddling force, action time, thrust direction, and water - paddling power; the action trajectory data includes three - dimensional position data, acceleration, angular velocity, trajectory path length, and the time taken for water - paddling; the water flow resistance distribution data includes the total resistance magnitude, the change in water flow resistance, and the resistance magnitudes under different actions; the physiological parameter data includes the swimmer's heart rate, respiratory rate, blood oxygen saturation, muscle force data, and body surface temperature.

[0070] The swimming trainer based on training data collects the following four types of key data through a multi - source fusion sensor network:

[0071] The water - paddling propulsion force data is used to quantify the actual thrust generated by each water - paddling, providing a physical basis for water - paddling efficiency; the action trajectory data provides a basis for accurate three - dimensional motion analysis and is the key to identifying the quality and rhythm of technical actions; the water flow resistance distribution data helps to understand the characteristics of fluid resistance suffered by the swimmer in different postures and speeds and is the core of constructing a resistance feedback model; the physiological parameter data is used to monitor the body load and fatigue state and supports personalized adjustment of training intensity; the multi - source data capture layer is the basis for constructing personalized resistance control and dynamic training feedback, truly realizing high - precision coupling modeling and real - time response of human - action - environment.

[0072] Multi - dimensional data collaborative acquisition:

[0073] The integrated sensor network combines kinetic, kinematic, hydrodynamic, and physiological data, breaking through the limitations of traditional single-data acquisition. Through a unified clock protocol and distributed layout, spatio-temporally unified and high-fidelity data fusion is achieved. The concepts of muscle-fluid coupling matrix and phase-matching curve are proposed to evaluate the quality of the stroke from the combined perspective of physics and physiology, going beyond the traditional single-factor analysis of trajectory or force. Using the captured data, a controllable magnetic field is constructed in real time for time-varying drag feedback adjustment, enabling the drag to be synchronized with the motion technical parameters in real time and simulating the real water drag state. The trainer is no longer a static device but automatically adjusts the training target and drag output curve according to data changes, supporting truly personalized training.

[0074] The method for obtaining n stroke cycles and stroke technical parameters includes:

[0075] Define the instantaneous stroke force in the stroke propulsion force data as a time series, and use the sliding window method to analyze the instantaneous stroke force. Preset the size of the sliding window as w. Within each sliding window, calculate the first derivative of the instantaneous stroke force to obtain the rate of change of the instantaneous stroke force over time;

[0076] Identify the start and end of the stroke cycle based on the rate of change of the propulsion force over time. Mark the candidate boundaries of the stroke cycle by finding the zero-crossing points of the first derivative of the instantaneous stroke force. The start and end of the stroke cycle are preliminarily determined through the candidate boundaries of the stroke cycle. The zero-crossing point indicates that the rate of change of the propulsion force changes from positive to negative or from negative to positive;

[0077] When the rate of change of the propulsion force over time first changes from negative to positive, it is recorded as the start of the stroke cycle; when the rate of change of the propulsion force over time first changes from positive to negative, it is recorded as the end of the stroke cycle;

[0078] However, there will be problems with ambiguous boundary determination in the existing technology. The characteristics of the stroke cycle vary greatly under different swimming strokes, making it difficult for traditional methods to uniformly identify. The strokes of beginners or in a fatigued state are irregular, resulting in difficulties in identification, and the candidate boundaries of the stroke cycle;

[0079] Introduce the coupling similarity function between the instantaneous stroke force and the motion trajectory to improve the accuracy of stroke cycle boundary determination. The coupling similarity function is: where α represents the weight coefficient of the instantaneous stroke force change, controlling the contribution of the instantaneous stroke force to the similarity function and reflecting the influence of the propulsion force change rate; β represents the weight coefficient of the motion trajectory acceleration, controlling the contribution of the motion trajectory data to the similarity function and reflecting the influence of the acceleration of the motion trajectory; dF(t) represents the first derivative of the instantaneous stroke force; F(t) represents the instantaneous stroke force, describing the instantaneous stroke force of the swimmer at each time point; represents the rate of change of the instantaneous stroke force over time; Denote the acceleration of the motion trajectory; P(t) represents the three-dimensional coordinates of the swimmer's body position, P(t) = [x(t), y(t), z(t)]; t represents the variable index of the time point; ||·|| represents the Euclidean norm of the acceleration of the motion trajectory;

[0080] Adjustment actions such as irregular actions of beginners or actions in a fatigued state, static swings, etc. may cause fluctuations similar to periods in the propulsion force or trajectory data, thus being misidentified as a complete stroke cycle; slight force or trajectory changes may also cause derivative zero-crossing points, thus being wrongly marked as cycle boundaries;

[0081] Preset the minimum stroke cycle duration threshold, exclude the stroke cycle recognition results less than the preset minimum stroke cycle duration threshold, compare the difference between the maximum instantaneous propulsion force and the minimum instantaneous propulsion force with the preset minimum stroke cycle duration threshold, and when the difference between the maximum instantaneous propulsion force and the minimum instantaneous propulsion force is greater than the preset minimum stroke cycle duration threshold, it is determined that there is an effective stroke action within the current time sliding window;

[0082] Different swimmers or different motion states (such as fatigued state, different swimming strokes, etc.) will lead to different change rates of the propulsion force data. Determining the candidate boundaries of the stroke cycle through a unified threshold parameter may not be applicable to all situations, resulting in deviations in cycle recognition under different stroke patterns and being unable to flexibly adjust the threshold according to the individual differences and training status of the athletes; Introduce the coefficient of variation of the current stroke cycle to dynamically adjust the preset minimum stroke cycle duration threshold. Define the duration of all recognized stroke cycles as T, the standard deviation of the durations of all stroke cycles as σ T , and the mean of the durations of all stroke cycles as μ T , divide the standard deviation σ T of the durations of all stroke cycles by the mean μ T of the durations of all stroke cycles to obtain the coefficient of variation CV;

[0083] When the coefficient of variation CV is greater than the preset minimum stroke cycle duration threshold, it is determined that the volatility of the current stroke cycle is large, and the preset minimum stroke cycle duration threshold is dynamically adjusted through a dynamic adjustment threshold function, θ(t) = γ·μ T +(1 - γ)·μ ΔF ; where θ(t) represents the adaptive cycle recognition threshold at time point t; γ represents the allocation adjustment factor, which is used to control the weight allocation of the two parts of the mean in the threshold calculation. According to the expert experience method, γ ∈ (0, 1]; μ ΔF represents the mean of the instantaneous propulsion force fluctuations; ΔF represents the instantaneous propulsion force fluctuation, that is, the difference between the instantaneous propulsion forces at adjacent time points;

[0084] Solve the following problems existing in the prior art:

[0085] The start and end of the stroke cycle are often determined by the zero-crossing points of the first derivative of the instantaneous stroke force. However, in actual training, especially for beginners or in a fatigued state, the movements of swimmers may be irregular, with significant differences in the characteristics of the stroke cycle. Traditional zero-crossing detection methods are prone to misjudgment. Specifically, minor force or trajectory changes may cause derivative zero-crossing points, which are then wrongly marked as cycle boundaries, or cycle fluctuations caused by irregular movements (such as adjustment movements, static swings, etc.) are misidentified as complete stroke cycles.

[0086] Variations in stroke data for different swimmers or in different motion states (such as different rates of force change, different swimming strokes, etc.) may result in the standardized cycle recognition threshold being unable to adapt to the situation of each athlete. Existing technologies may define cycle boundaries through a unified threshold, but this method cannot flexibly adapt to factors such as individual differences and fatigue states.

[0087] Irregular movements, minor movement changes, static swings, etc. in the movements of beginners or in a fatigued state may interfere with the determination of the stroke cycle, making it difficult to accurately identify the boundaries of the stroke cycle. Existing technologies may fail to effectively eliminate the errors brought about by these irregular movements.

[0088] Innovative points of this solution:

[0089] By combining the changes in instantaneous stroke force and the acceleration of the movement trajectory, a coupled similarity function is used to improve the accuracy of determining the boundaries of the stroke cycle. This method effectively avoids the misjudgment problems caused by solely relying on the zero-crossing points of the first derivative of the instantaneous stroke force. By comprehensively considering force changes and the acceleration of the movement trajectory, the reliability and accuracy of boundary determination are enhanced.

[0090] The coefficient of variation of the stroke cycle duration is introduced to dynamically adjust the minimum stroke cycle duration threshold to adapt to the cycle fluctuations in different swimmers or different motion states. Based on the duration distribution and standard deviation of the stroke cycle, the system can flexibly adjust the threshold according to the current state, avoiding the deviation caused by a fixed threshold and improving the robustness of cycle recognition.

[0091] By analyzing the standard deviation and mean of all stroke cycles, the threshold is dynamically adjusted, enabling effective optimization of cycle recognition in different states (such as beginners, fatigued states, different swimming strokes, etc.), avoiding the problem that a fixed threshold cannot adapt to different situations. By reasonably presetting the threshold limit for the stroke cycle duration, stroke cycles shorter than the preset minimum stroke cycle duration are excluded, reducing the situation of misidentifying small cycle fluctuations as complete cycles.

[0092] Beneficial effects compared to existing technologies are:

[0093] By introducing a coupling similarity function of instantaneous rowing force and motion trajectory, compared with the prior art that solely relies on the zero-crossing point of the first derivative of the instantaneous rowing force, it can better handle the interference caused by fatigue, beginners, or irregular motions, making the identification of the rowing cycle boundary more accurate.

[0094] By dynamically adjusting the minimum rowing cycle duration threshold, the identification criteria can be flexibly adjusted according to the specific state of the swimmer (such as fatigue, different swimming strokes, etc.), overcoming the problem in the prior art that a fixed threshold cannot cope with different motion states, and having higher adaptability and accuracy.

[0095] By excluding rowing cycles shorter than the minimum rowing cycle duration threshold, the system can effectively filter out invalid cycle fluctuations, reduce unnecessary misjudgments, and thus ensure the high quality and reliability of the training data.

[0096] Dynamically adjusting the threshold and introducing the coefficient of variation mechanism can adjust the boundary identification criteria in real time according to the volatility of the rowing cycle, effectively avoiding identification problems caused by athlete differences or changes in motion states, and improving the stability of the system.

[0097] By using the coupling similarity function to process the similarity between the propulsion force and trajectory data, it can effectively identify and filter out irregular motions, avoiding the problem of cycle misidentification caused by slight changes, and is especially suitable for training beginners or those in a fatigued state.

[0098] For each rowing cycle, calculate the peak value of the instantaneous rowing force, the rowing cycle duration, the maximum instantaneous rowing force, and the minimum instantaneous rowing force to obtain the rowing technique parameters.

[0099] The method for obtaining the muscle-fluid phase matching curve includes:

[0100] Express the muscle force data in the physiological parameter data as a time-series muscle activation vector M(t), splice the rowing propulsion force data and the motion trajectory data, collectively referred to as fluid response data, express the fluid response data as a fluid response vector F(t), and calculate the degree of coordination between the muscle force data and the fluid response data within each rowing cycle.

[0101] Through the inner product operation of the muscle activation vector M(t) and the fluid response vector F(t) for each rowing cycle, obtain the muscle-fluid coupling matrix within each rowing cycle, perform normalization processing on the muscle-fluid coupling matrix within each rowing cycle, calculate the Frobenius norm of the muscle-fluid coupling matrix within each rowing cycle, and unify the scales of the muscle-fluid coupling matrices of different rowing cycles to obtain the normalized muscle-fluid coupling matrix.

[0102] Perform eigenvalue decomposition on the muscle-fluid coupling matrix to extract the principal coupling direction between muscle force data and fluid response data. Obtain the eigenvalues and eigenvectors of the matrix through eigenvalue decomposition. The eigenvalue represents the strength of the coupling between muscle force data and fluid response data, and the eigenvector represents the principal direction of the coupling. Take the eigenvector corresponding to the largest eigenvalue as the principal coupling direction. The largest eigenvalue represents the strongest coupling strength between muscle force data and fluid response data at any time point t. Arrange the largest eigenvalues at each time point t in chronological order to obtain the muscle-fluid phase matching curve.

[0103] The method for obtaining individualized training data includes:

[0104] Use a multi-modal Transformer to construct a training data prediction model. The training data prediction model includes an input layer, an encoding layer, a multi-modal fusion layer, a decoding layer, and a multi-task output layer. Use the historical muscle-fluid phase matching curve and physiological parameter data as the input data of the input layer, and output the corresponding individualized training data through the multi-task output layer. The individualized training data includes movement patterns, strength curve change trends, and fatigue levels. Movement patterns include different swimming strokes (such as freestyle, breaststroke, butterfly stroke, etc.), stroke force, and stroke frequency.

[0105] Use multi-class cross-entropy as the loss function of the model to optimize the training data prediction model until it stops when reaching the preset number of iterations, and obtain the trained training data analysis model. Input the current muscle-fluid phase matching curve and physiological parameter data into the trained training data analysis model to predict the individualized training data.

[0106] The method for constructing a controllable magnetic field includes:

[0107] Based on the action trajectory data and water flow resistance distribution data, calculate the influence of the swimmer on the water flow in each stroke cycle through the CFD fluid dynamics model. Calculate a water flow resistance pattern based on the influence of the swimmer on the water flow in each stroke cycle. The water flow resistance pattern includes the magnitude of the water flow resistance, the direction of the water flow resistance, and the three-dimensional coordinate area of the water flow resistance distribution. Adjust the training load of the swimmer according to the strength curve change trend and fatigue level to match the physiological parameter data of each swimmer. Based on the water flow resistance pattern, construct a controllable magnetic field through a magnetic control resistance unit. The magnetic control resistance unit is a device that uses the principle of magnetic field to control and adjust the resistance, which is composed of a permanent magnet or an electromagnet and a magnetic conductive material (such as metal, magnetic fluid, etc.), and affects the magnitude of the resistance by adjusting the strength and distribution of the magnetic field. The working principle of the magnetic control resistance unit usually involves the following aspects:

[0108] Relationship between magnetic field and resistance: When a magnetic field acts on a specific magnetic material or liquid, a resistance proportional to the magnetic field strength is generated. Usually, by changing the strength or direction of the magnetic field, the magnitude of the resistance can be precisely adjusted. For example, in a swimming trainer, the magnetic field can be adjusted electromagnetically to generate different reaction forces on the contact surface with the water flow, thereby adjusting the resistance in the water.

[0109] Use of permanent magnets and electromagnets:

[0110] Permanent magnet: Provides stable resistance through a fixed magnetic field strength. In some sports equipment, permanent magnets are usually used to generate a constant magnetic field and provide a constant resistance.

[0111] Electromagnet: Adjusts the strength of the magnetic field by changing the current intensity, and then adjusts the resistance. Electromagnets can usually precisely control the change of the magnetic field, so they are suitable for scenarios that require dynamic adjustment of resistance.

[0112] Resistance adjustment method: In a magnetic control resistance unit, the generated resistance is usually controlled by adjusting the relative position of the magnetic field (for example, changing the distance between the magnet and the magnetic material), the magnetic field strength (for example, adjusting the current magnitude), or changing the concentration of magnetic particles in the fluid, etc. These changes will directly affect the exercise load of the swimmer in the water and help adjust the training intensity;

[0113] Adaptive adjustment: In a swimming trainer, the magnetic control resistance unit adjusts the strength and distribution of the magnetic field in real time according to training data (such as the swimmer's movement pattern, power output, fatigue level, etc.) to ensure that the resistance change meets the individualized training needs. This adjustment is usually achieved by sensors (such as speed sensors, force sensors, electromyogram sensors, etc.) to collect data, and then the control system dynamically calculates and adjusts the magnetic field.

[0114] Methods for real-time adjustment of the resistance on the swimmer's stroke path include:

[0115] Based on individualized training data, construct a resistance model that matches the stroke technical parameters through the LFS hydrodynamic model. Input the swimmer's movement trajectory data into the LFS hydrodynamic model to simulate the disturbance and flow velocity distribution of the water flow under the stroke action, analyze the influence of the swimmer on the water flow in each stroke cycle, and output the water flow resistance data according to the water flow resistance pattern to match the resistance on the swimmer's stroke path. For example, if the swimmer exerts greater force in a certain stroke cycle, the system can calculate the required increased resistance value to ensure that the training intensity matches. Real-time adjustment of the resistance on the swimmer's stroke path is carried out through a fuzzy control algorithm, and fuzzy rules are defined. The fuzzy rules include Rule 1 and Rule 2;

[0116] Rule 1 is that if the change rate of the swimmer's force curve has a change error greater than the preset change error threshold compared to the change rate of the preset swimmer's force curve, and the swimmer's fatigue level is greater than the preset fatigue level threshold, then the resistance on the swimmer's stroke path is reduced; Rule 2 is that if the change rate of the swimmer's force curve has a change error less than or equal to the preset change error threshold compared to the change rate of the preset swimmer's force curve, and the swimmer's fatigue level is less than or equal to the preset fatigue level threshold, then the resistance on the swimmer's stroke path is increased.

[0117] The methods for obtaining the resistance change rate include:

[0118] Based on the motion trajectory data and the stroke propulsion force data, the current stroke cycle is identified in real time. For example, the instantaneous stroke force of the stroke propulsion force data changes with different stages of the stroke motion. Using these data, the start and end of the current cycle can be accurately delimited. And the stroke cycle is divided into a stroke preparation stage, an entry stage, a pulling stage, a propulsion stage, and an exit stage; for example, in the stroke entry stage, the disturbance of the water flow is small, and the water flow resistance may be low; while in the pulling and propulsion stages, the water flow resistance will be greater because the water flow has a large disturbance under the thrust of the swimmer's arm or body.

[0119] Within each stage of each stroke cycle, the magnitude of the resistance changes over time. By analyzing the change of the water flow resistance distribution data over time, the resistance change rate within each stage is calculated. Where, ΔR represents the change amount of the water flow resistance; Δt represents the time interval corresponding to the change amount of the water flow resistance.

[0120] The methods for matching and executing corresponding training strategies include:

[0121] According to the swimmer's individualized training data and the resistance change rate, different training strategies are formulated; the resistance control terminal of the swimming trainer automatically selects the training strategy suitable for the current training stage and executes the corresponding training strategy, continuously monitors the swimmer's physiological parameter data, and adjusts the training strategy according to the real-time physiological parameter data; the training strategies include technical optimization training strategies, strength enhancement training strategies, endurance maintenance training strategies, and rehabilitation training strategies.

[0122] The methods for outputting a personalized control plan include:

[0123] According to the swimmer's real-time physiological parameter data, the training goals are automatically set. The training goals include strength training goals, technical improvement goals, and fatigue recovery goals; according to the swimmer's real-time individualized training data and the resistance change rate, the strategy suitable for the current training stage is selected. For example:

[0124] Technical improvement goal: If the swimmer's stroke is not smooth enough, the system may choose to increase the resistance, forcing the swimmer to use greater force and more standardized movements, thus improving the technique.

[0125] Strength enhancement goal: If the swimmer has a strong force curve and low fatigue level, the system may choose to increase the resistance to strengthen muscle strength and increase the training intensity.

[0126] Fatigue management goal: If the swimmer's fatigue level is high, the system will automatically reduce the resistance to help the swimmer adapt to the training load and avoid overtraining.

[0127] After determining the training goal, the resistance control terminal of the swimming trainer automatically generates a phased resistance distribution curve based on the swimmer's real-time individualized training data and training goal; depicts the resistance changes in each stage within each stroke cycle through the phased resistance distribution curve, and automatically generates a corresponding personalized control plan based on the training goal and the phased resistance distribution curve; the personalized control plan includes resistance setting, training time, stage arrangement, and feedback mechanism;

[0128] Resistance setting: Specify the resistance change range within each stroke cycle to ensure that the training intensity matches the goal. For example, during the strength enhancement stage, the system may set a gradually increasing resistance; during the recovery stage, the system may set a gradually decreasing resistance.

[0129] Training time and stage arrangement: The duration of each training stage and the specific resistance adjustment plan within each stage. For example, the strength training stage may last longer, while the technical training stage may last shorter.

[0130] Feedback mechanism: The system will set a feedback mechanism to dynamically adjust the training goal and resistance setting according to the swimmer's real-time performance. For example, if the swimmer shows extreme fatigue during the strength training stage, the system will reduce the resistance intensity to avoid overtraining.

[0131] According to the matched training strategy, the system will adjust the magnetic control resistance unit in real time to ensure that during the training process, the water flow resistance matches the swimmer's training goal.

[0132] The preset optimal resistance change rate threshold is set by the staff. Different optimal resistance change rates are collected through the resistance control terminal of the swimming trainer, and the average value of multiple optimal resistance change rates is taken as the preset optimal resistance change rate threshold; similarly, the preset minimum stroke cycle duration threshold, preset change error threshold, and preset fatigue level threshold are set.

[0133] In this embodiment, by combining the instantaneous propulsion force and the change of the acceleration of the movement trajectory, a coupling similarity function is used to improve the accuracy of the determination of the boundary of the stroke cycle. This method effectively avoids the misjudgment problem caused by the zero-crossing point of the first derivative of the instantaneous propulsion force alone. By comprehensively considering the force change and the acceleration of the movement trajectory, the reliability and accuracy of the boundary determination are enhanced. By introducing the coupling similarity function of the instantaneous propulsion force and the movement trajectory, compared with the prior art that solely relies on the zero-crossing point of the first derivative of the instantaneous propulsion force, it can better handle the interference caused by fatigue, beginners, or irregular movements, making the identification of the stroke cycle boundary more accurate.

[0134] The coefficient of variation of the stroke cycle duration is introduced to dynamically adjust the threshold of the minimum stroke cycle duration to adapt to the cycle fluctuations of different swimmers or different exercise states. Based on the duration distribution and standard deviation of the stroke cycle, the system can flexibly adjust the threshold according to the current state, avoiding the deviation caused by a fixed threshold and improving the robustness of cycle identification. It overcomes the problem in the prior art that a fixed threshold cannot cope with different exercise states and has higher adaptability and accuracy.

[0135] By analyzing the standard deviation and mean value of all stroke cycles, the threshold is dynamically adjusted, so that the cycle identification in different states (such as beginners, fatigue states, different swimming strokes, etc.) can be effectively optimized, avoiding the problem that a fixed threshold cannot adapt to different situations. By reasonably presetting the threshold limit for the stroke cycle duration, the stroke cycles shorter than the preset minimum stroke cycle duration are excluded, reducing the situation of misidentifying small cycle fluctuations as complete cycles. By using the coupling similarity function to process the similarity between the propulsion force and the trajectory data, irregular movements can be effectively identified and filtered, avoiding the problem of cycle misidentification caused by slight changes, especially suitable for the training of beginners or in a fatigue state.

[0136] Embodiment 2

[0137] Please refer to Figure 2 As shown, for the parts not described in detail in this embodiment, refer to the description content of Embodiment 1. A collaborative management method for an Internet of Things gateway based on edge computing is provided, including:

[0138] S1. Real-time capture of swimmer training data through an integrated sensor network, including stroke propulsion force data, movement trajectory data, water flow resistance distribution data, and physiological parameter data;

[0139] S2. Apply an action decomposition algorithm to identify and segment the stroke propulsion force data and the movement trajectory data to obtain n stroke cycles and corresponding stroke technical parameters, and construct a muscle-fluid coupling matrix to obtain the muscle-fluid phase matching curve in each stroke cycle; predict and generate personalized training data based on the muscle-fluid phase matching curve and the physiological parameter data;

[0140] S3. Construct a controllable magnetic field based on the individualized training data, motion trajectory data, and water flow resistance distribution data. Generate a resistance that matches the rowing technique parameters based on the individualized training data, and adjust the resistance on the swimmer's rowing path in real time.

[0141] S4. Identify the current rowing cycle. Based on the water flow resistance distribution data, control the resistance to change dynamically with time according to each stage within the current rowing cycle, obtain the resistance change rate, and compare it with the preset optimal resistance change rate threshold. When the preset optimal resistance change rate threshold is reached, maintain the current resistance output until the end of the current rowing cycle.

[0142] S5. Receive the individualized training data and the resistance change rate, match and execute the corresponding training strategy; automatically set the training target parameters and the phased resistance distribution curve according to the training strategy, and output a personalized control plan.

[0143] Since the electronic device introduced in this embodiment is the electronic device used in the resistance control system of the swimming trainer based on training data in the embodiments of the present application, based on the resistance control system of the swimming trainer based on training data introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the resistance control system of the swimming trainer based on training data in the embodiments of the present application, it falls within the protection scope of the present application.

[0144] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0145] The above is only the preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for ordinary technical users in the technical field of the present invention, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A resistance control system for a swimming trainer based on training data, characterized in that, Comprising: A multi-source data capture layer that captures the training data of swimmers in real time through an integrated sensor network, including stroke propulsion force data, motion trajectory data, water flow resistance distribution data, and physiological parameter data; An extraction and recognition layer that applies an action decomposition algorithm to identify and segment the stroke propulsion force data and motion trajectory data, obtains n stroke cycles and corresponding stroke technique parameters, constructs a muscle-fluid coupling matrix, and obtains the muscle-fluid phase matching curve in each stroke cycle; predicts and generates individualized training data based on the muscle-fluid phase matching curve and physiological parameter data; A magnetic control resistance generation layer that constructs a controllable magnetic field through individualized training data, motion trajectory data, and water flow resistance distribution data, and generates a resistance matching the stroke technique parameters based on the individualized training data to adjust the resistance on the swimmer's stroke path in real time; A resistance dynamic adjustment layer that identifies the current stroke cycle, and based on the water flow resistance distribution data, controls the resistance to change dynamically with time in each stage of the current stroke cycle, obtains the resistance change rate, and compares it with a preset optimal resistance change rate threshold. When the preset optimal resistance change rate threshold is reached, the current resistance output is maintained until the end of the current stroke cycle; A training mode matching layer that receives individualized training data and resistance change rate, matches and executes corresponding training strategies; automatically sets training target parameters and a phased resistance distribution curve according to the training strategies, and outputs a personalized control scheme.

2. The resistance control system of the swimming trainer based on training data according to claim 1, wherein The method for capturing the training data of swimmers includes: The integrated sensor network is composed of m types of sensors, including force sensors, accelerometers, gyroscopes, water flow velocity sensors, physiological parameter sensors, and position perception devices; the m types of sensors are integrated at key parts of the trainer through waterproof packaging to form a multi-point collaborative integrated sensor network; The m types of sensors are connected to the multi-source data capture layer by wired or wireless means, a unified clock synchronization protocol is set, and the training data of swimmers is collected through the m types of sensors; The stroke propulsion force data includes instantaneous stroke force, average stroke force, maximum stroke force, action time, thrust direction, and stroke power; the motion trajectory data includes three-dimensional position data, acceleration, angular velocity, trajectory path length, and time taken for the stroke; the water flow resistance distribution data includes the total resistance magnitude, water flow resistance change amount, and resistance magnitudes under different actions; the physiological parameter data includes the swimmer's heart rate, respiratory rate, blood oxygen saturation, muscle force data, and body surface temperature.

3. The resistance control system of the swimming trainer based on training data according to claim 2, characterized in that, The method for obtaining the n stroke cycles and stroke technique parameters includes: Define the instantaneous stroke force in the stroke propulsion force data as a time series, analyze the instantaneous stroke force using the sliding window method, preset the sliding window size as w, and calculate the first derivative of the instantaneous stroke force within each sliding window to obtain the rate of change of the instantaneous stroke force with time; Identify the start and end of the stroke cycle according to the rate of change of the propulsion force with time, mark the candidate boundaries of the stroke cycle by finding the zero crossing points of the first derivative of the instantaneous stroke force, and preliminarily determine the start and end of the stroke cycle through the candidate boundaries of the stroke cycle. The zero crossing point indicates that the rate of change of the propulsion force changes from positive to negative or from negative to positive; When the rate of change of the propulsion force from negative to positive for the first time, it is recorded as the start of the stroke cycle; when the rate of change of the propulsion force from positive to negative for the first time, it is recorded as the end of the stroke cycle; Introduce the coupling similarity function between the instantaneous rowing force and the movement trajectory. The coupling similarity function is as follows: where α represents the weight coefficient of the instantaneous rowing force change; β represents the weight coefficient of the movement trajectory acceleration; dF(t) represents the first derivative of the instantaneous rowing force; F(t) represents the instantaneous rowing force; represents the change rate of the instantaneous rowing force with respect to time; represents the acceleration of the movement trajectory; P(t) represents the three-dimensional coordinates of the swimmer's body position, P(t) = [x(t), y(t), z(t)]; t represents the variable index of the time point; ||·|| represents the Euclidean norm of the acceleration of the movement trajectory; A preset minimum stroke cycle duration threshold is set to exclude the stroke cycle recognition results shorter than the preset minimum stroke cycle duration threshold. The difference between the maximum instantaneous propulsion force and the minimum instantaneous propulsion force is compared with the preset minimum stroke cycle duration threshold. When the difference between the maximum instantaneous propulsion force and the minimum instantaneous propulsion force is greater than the preset minimum stroke cycle duration threshold, it is determined that there is an effective stroke action within the current time sliding window; Introduce the coefficient of variation of the current stroke cycle to dynamically adjust the preset minimum stroke cycle duration threshold. Define the duration of all identified stroke cycles as T, the standard deviation of the durations of all stroke cycles as σ T , and the mean of the durations of all stroke cycles as μ T , divide the standard deviation σ of the durations of all stroke cycles T by the mean μ of the durations of all stroke cycles T to obtain the coefficient of variation CV; When the coefficient of variation CV is greater than the preset minimum stroke cycle duration threshold, it is determined that the volatility of the current stroke cycle is large, and the preset minimum stroke cycle duration threshold is dynamically adjusted through a dynamic adjustment threshold function, θ(t) = γ·μ T +(1 - γ)·μ ΔF ; where θ(t) represents the adaptive cycle recognition threshold at time point t; γ represents the allocation adjustment factor; μ ΔF represents the mean value of the instantaneous stroke force fluctuation; ΔF represents the instantaneous stroke force fluctuation; For each stroke cycle, calculate the peak value of the instantaneous propulsion force, the stroke cycle duration, the maximum instantaneous propulsion force, and the minimum instantaneous propulsion force to obtain the stroke technique parameters.

4. The resistance control system of the swimming trainer based on training data according to claim 3, characterized in that The method for obtaining the muscle-fluid phase matching curve includes: Represent the muscle force data in the physiological parameter data as a time series of muscle activation vectors M(t). Concatenate the stroke propulsion force data and the motion trajectory data, which are collectively called fluid response data, and represent the fluid response data as a fluid response vector F(t). Calculate the degree of cooperation between the muscle force data and the fluid response data within each stroke cycle; Through the inner product operation of the muscle activation vector M(t) and the fluid response vector F(t) for each stroke cycle, obtain the muscle-fluid coupling matrix for each stroke cycle, and perform normalization processing on the muscle-fluid coupling matrix for each stroke cycle. Calculate the Frobenius norm of the muscle-fluid coupling matrix for each stroke cycle, and unify the scales of the muscle-fluid coupling matrices of different stroke cycles to obtain the normalized muscle-fluid coupling matrix; Perform eigenvalue decomposition on the muscle-fluid coupling matrix to extract the main coupling direction between the muscle force data and the fluid response data. Obtain the eigenvalues and eigenvectors of the matrix through eigenvalue decomposition; the eigenvalues represent the strength of the coupling between the muscle force data and the fluid response data, and the eigenvectors represent the main direction of the coupling; take the eigenvector corresponding to the largest eigenvalue as the main coupling direction, and the largest eigenvalue represents the strongest coupling strength between the muscle force data and the fluid response data at any time point t; arrange the largest eigenvalues at each time point t in chronological order to obtain the muscle-fluid phase matching curve.

5. The resistance control system of the swimming trainer based on training data according to claim 4, wherein The method for obtaining the individualized training data includes: Use a multi-modal Transformer to construct a training data prediction model. The training data prediction model includes an input layer, an encoding layer, a multi-modal fusion layer, a decoding layer, and a multi-task output layer; use the historical muscle-fluid phase matching curve and physiological parameter data as the input data of the input layer, and output the corresponding individualized training data through the multi-task output layer; the individualized training data includes action patterns, the change trend of the force curve, and the degree of fatigue; Use multi-class cross-entropy as the loss function of the model to optimize the training data prediction model until it stops when reaching the preset number of iterations, and obtain a trained training data analysis model; input the current muscle fluid phase matching curve and physiological parameter data into the trained training data analysis model to predict individualized training data.

6. The resistance control system of the swimming trainer based on training data according to claim 5, characterized in that The method for constructing the controllable magnetic field includes: Based on the action trajectory data and water flow resistance distribution data, calculate the influence of the swimmer on the water flow in each stroke cycle through the CFD hydrodynamics model. Based on the influence of the swimmer on the water flow in each stroke cycle, calculate a water flow resistance pattern, which includes the magnitude of the water flow resistance, the direction of the water flow resistance, and the three-dimensional coordinate area of the water flow resistance distribution; and adjust the training load of the swimmer according to the change trend of the force curve and the degree of fatigue to match the physiological parameter data of each swimmer; based on the water flow resistance pattern, construct a controllable magnetic field through the magnetic control resistance unit.

7. The resistance control system of the swimming trainer based on training data according to claim 6, characterized in that, The method for real-time adjusting the resistance on the swimmer's stroke path includes: Based on the individualized training data, construct a resistance model matching the stroke technical parameters through the LFS hydrodynamics model. Input the action trajectory data of the swimmer into the LFS hydrodynamics model to simulate the disturbance and flow velocity distribution of the water flow under the stroke action, analyze the influence of the swimmer on the water flow in each stroke cycle, and output the water flow resistance data according to the water flow resistance pattern to match the resistance on the swimmer's stroke path. Real-time adjust the resistance on the swimmer's stroke path through the fuzzy control algorithm, and define the fuzzy rules, which include Rule 1 and Rule 2; Rule 1 is that if the change error between the change rate of the swimmer's force curve and the change rate of the preset swimmer's force curve is greater than the preset change error threshold, and the degree of fatigue of the swimmer is greater than the preset fatigue degree threshold, then reduce the resistance on the swimmer's stroke path; Rule 2 is that if the change error between the change rate of the swimmer's force curve and the change rate of the preset swimmer's force curve is less than or equal to the preset change error threshold, and the degree of fatigue of the swimmer is less than or equal to the preset fatigue degree threshold, then increase the resistance on the swimmer's stroke path.

8. The resistance control system of the swimming trainer based on training data according to claim 7, characterized in that, The method for obtaining the resistance change rate includes: Based on the action trajectory data and stroke propulsion force data, identify the current stroke cycle in real time, and divide the stroke cycle into a stroke preparation stage, an entry stage, a pulling stage, a propulsion stage, and an exit stage; During each stage of each stroke cycle, by analyzing the variation of the water flow resistance distribution data over time, the rate of change of resistance within each stage is calculated. Wherein, ΔR represents the change in water flow resistance; Δt represents the time interval corresponding to the change in water flow resistance.

9. The resistance control system of the swimming trainer based on training data according to claim 8, wherein The method for matching and executing corresponding training strategies includes: Formulate different training strategies according to the individualized training data and resistance change rate of the swimmer; the resistance control terminal of the swimming trainer automatically selects the training strategy suitable for the current training stage and executes the corresponding training strategy, continuously monitors the physiological parameter data of the swimmer, and adjusts the training strategy according to the real-time physiological parameter data; the training strategies include technical optimization training strategies, strength enhancement training strategies, endurance maintenance training strategies, and rehabilitation training strategies.

10. The resistance control system of the swimming trainer based on training data according to claim 9, wherein The method for outputting a personalized control plan includes: Automatically set training goals according to the real-time physiological parameter data of the swimmer, and the training goals include strength training goals, technical improvement goals, and fatigue recovery goals; After determining the training objective, the resistance control terminal of the swimming trainer automatically generates a phased resistance distribution curve based on the real-time individualized training data and training objective of the swimmer; the phased resistance distribution curve depicts the resistance changes in each stage within each stroke cycle, and based on the training objective and the phased resistance distribution curve, an individualized control plan is automatically generated.

Images