A Smart Matching Method for Ab Wheel Training Intensity Based on Adaptive Adjustment of Elastic Resistance

Abstract

This application relates to the field of intelligent control technology for fitness equipment, and provides a method for intelligent matching of training intensity for an ab wheel based on adaptive adjustment of elastic resistance. The method acquires real-time motion state data of the user during ab wheel training through a sensor module; inputs the motion state data and physiological state data into a training intensity matching model, which calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user's input basic body parameters; when the matching degree is lower than a preset matching threshold, a resistance adjustment command is generated according to the training intensity matching model. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel rolls. This method fills the technological gaps in dynamic matching of training intensity, nonlinear resistance adjustment, and closed-loop feedback control, providing a new approach for the design of intelligent fitness equipment.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology for fitness equipment, and in particular to an intelligent matching method for training intensity of an ab wheel based on adaptive adjustment of elastic resistance. Background Technology

[0002] The ab wheel is a widely used core muscle training tool, and its training effect is closely related to the proper matching of resistance intensity. Traditional ab wheels usually use a fixed resistance design (such as constant spring resistance or damping structure), which cannot dynamically adjust the resistance according to the user's real-time physical condition, movement accuracy, and training goals, leading to the following problems:

[0003] 1. Insufficient precision in matching training intensity: Individual differences among users (such as height, weight, body fat percentage, and athletic ability) and changes in physical fitness during training (such as muscle fatigue and heart rate fluctuations) are not effectively quantified. Resistance adjustment relies solely on the user's subjective feelings or preset levels, which can easily lead to problems such as excessive resistance causing movement distortion (such as lumbar compensation and improper core muscle activation) or insufficient resistance leading to inadequate training effects.

[0004] 2. Single resistance adjustment mechanism: The resistance adjustment of existing smart ab rollers is mostly based on linear adjustment of a single motion parameter (such as rolling displacement or speed), lacking real-time monitoring and comprehensive analysis of the user's physiological state (such as heart rate and electromyographic signals), and unable to form a closed-loop feedback of "data acquisition-model calculation-resistance adjustment", making it difficult to achieve dynamic adaptation of training intensity.

[0005] 3. Lack of nonlinear resistance characteristics: During the forward push of a traditional ab wheel, the resistance usually increases linearly with the displacement or remains constant, which does not match the force characteristics of the human core muscle group (the initial force needs to overcome inertia, the middle stage needs to maintain stable force, and the extreme displacement needs to control balance); during the pull-back process, the resistance adjustment is not dynamically coupled with the pull-back speed, and it is impossible to intervene in real time for users whose movement speed is too fast (which may lead to muscle strain) or too slow (which results in low training efficiency).

[0006] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0007] This application provides an intelligent matching method for ab wheel training intensity based on adaptive elastic resistance adjustment. It aims to address the shortcomings of existing technologies, which, while proposing sensor-based resistance adjustment schemes for fitness equipment, fail to deeply integrate motion state data (rolling displacement, velocity, acceleration, force frequency) with physiological state data (heart rate, electromyographic signals), and lack a multi-dimensional training intensity matching model incorporating user's basic body parameters and training target parameters. This results in resistance adjustment strategies failing to consider both personalized needs and the biomechanical characteristics of movement. Furthermore, existing solutions lack the design for the non-linear increasing characteristics of forward resistance and the dynamic adaptive mechanism of pullback resistance, making it difficult to maximize training effectiveness while ensuring proper movement.

[0008] In a first aspect, embodiments of this application provide a method for intelligent matching of training intensity for an ab wheel based on adaptive adjustment of elastic resistance, applied to an intelligent ab wheel. The intelligent ab wheel includes an ab wheel body, an elastic resistance adjustment device, a sensor module, and a control module. The elastic resistance adjustment device is connected to the axle of the ab wheel body to provide variable elastic resistance. The sensor module is used to collect motion state data and physiological state data of the user during training. The control module pre-stores a training intensity matching model. The method includes:

[0009] The sensor module acquires real-time motion data of the user during ab wheel training. The motion data includes at least the rolling displacement, rolling speed, pull-back acceleration, and force frequency of the ab wheel.

[0010] The exercise status data and physiological status data are input into the training intensity matching model. The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user's input basic body parameters. The training target parameters include preset training duration, preset standard range of motion, and preset energy consumption threshold. The basic body parameters include height, weight, body fat percentage, and exercise ability level.

[0011] When the matching degree is lower than the preset matching threshold, a resistance adjustment command is generated according to the training intensity matching model. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel body rolls, so that the resistance of the ab wheel increases nonlinearly with the increase of the rolling displacement when the user pushes forward, and the resistance during the pull-back is dynamically and adaptively adjusted as the pull-back speed increases.

[0012] In some embodiments, the training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and user-input basic body parameters. This includes: normalizing real-time collected motion state data and physiological state data; weighting the ab wheel rolling displacement, rolling speed, pull-back acceleration, force frequency, heart rate signal, and electromyography signal according to preset weights to obtain the current training intensity feature value; establishing a user fitness baseline model based on the basic body parameters; calculating the difference between the current training intensity feature value and the target intensity feature value corresponding to the preset training target parameters; and using the difference as the matching degree between the current training intensity and the target training intensity.

[0013] In some embodiments, the dynamic generation of a phased training plan based on user motion data, personalized parameters, historical training fitness, and training goals, including movement types, training duration per set, rest intervals, and weight adjustment strategies, includes: determining the initial training difficulty level based on the user's input basic body parameters and historical training records; combining preset training goal parameters to generate a basic training plan including standard movement demonstrations, training duration per set, and rest intervals; during training, evaluating the user's movement standardization and physical exertion based on real-time collected motion and physiological data; after each training cycle, iteratively optimizing the movement types and resistance adjustment strategies in the basic training plan based on historical training fitness parameters to form a phased training plan adapted to the user's current physical fitness; the historical training fitness parameters include resistance adjustment frequency, pullback acceleration achievement rate, and heart rate recovery rate.

[0014] In some embodiments, when the matching degree is lower than a preset matching threshold, generating a resistance adjustment instruction based on the training intensity matching model includes: if the matching degree is lower than a first preset threshold, increasing the deformation degree of the elastic element by a preset ratio according to the current rolling displacement and pull-back speed, so that the forward resistance increases by 5%-10% in the current displacement range; if the matching degree is lower than a second preset threshold, determining whether the user is in an overload state by combining real-time electromyography signals and heart rate signals, and after eliminating the risk of overload, switching the resistance level by controlling the number of meshing teeth of the elastic resistance adjustment device, so that the elastic resistance torque during pull-back increases exponentially with the increase of pull-back speed, while limiting the peak value of the forward resistance to not exceed the safe load threshold corresponding to the user's basic body parameters.

[0015] In some embodiments, the training intensity matching model establishes a multi-dimensional state-space model that includes motion state parameters, physiological state parameters, and resistance adjustment parameters. It then uses fuzzy logic control or neural network algorithms to dynamically optimize the real-time resistance output of the elastic resistance adjustment device. This allows the adjusted elastic resistance to match the user's current physical condition and target training intensity, forming a closed-loop feedback adjustment. This enables the adaptive adjustment of elastic resistance and intelligent matching of training intensity during ab wheel training.

[0016] For example, the neural network algorithm uses a long short-term memory network to construct a training intensity prediction model. The training intensity prediction model is trained offline using the user's historical training data. The input parameters include historical motion state data, physiological state data, resistance adjustment parameters, and training effect feedback. The output parameter is the resistance adjustment benchmark value for the next training cycle. During real-time training, the training intensity prediction model dynamically corrects the resistance adjustment benchmark value based on the current training data. The correction strategy includes: when the pull-back acceleration is lower than the benchmark value for three consecutive times, an automatic resistance reduction mechanism is triggered to reduce the elastic resistance torque by 10%-15% to avoid the user's movement being deformed due to excessive resistance. The training effect feedback includes the degree of muscle soreness and the movement completion score.

[0017] For example, the fuzzy logic control includes multiple sets of fuzzy rules. The fuzzy rules use the user's heart rate change rate, force frequency fluctuation amplitude, and rolling displacement achievement rate as input variables, and resistance adjustment level as output variables. The fuzzy rules include: if the heart rate change rate exceeds a preset safety threshold and the force frequency decreases significantly, it is determined that the user is in a fatigued state, and a resistance reduction command is output to reduce the deformation of the elastic element by 20%-30%; if the rolling displacement achievement rate exceeds 90% for three consecutive times and the electromyographic signal intensity is stable, it is determined that the user's physical fitness has improved, and a resistance increase command is output to increase the pull-back resistance torque by 15%-20% by increasing the number of meshing teeth.

[0018] For example, the training intensity matching model uses the number of times the user completes the standard movement, energy consumption efficiency, and movement standardization as reward functions, and resistance adjustment parameters as the movement space. It optimizes the elastic resistance adjustment strategy by constructing a Markov decision process. Specifically, in the early stage of training, an ε-greedy strategy is used to explore the optimal resistance adjustment range. After the user's cumulative training time exceeds 5 hours, the model switches to a strategy gradient-based optimization method. The response sensitivity of the resistance adjustment device is adjusted according to real-time reward feedback so that the nonlinear increasing curve of the forward resistance matches the user's muscle exertion characteristics.

[0019] In some embodiments, the electromyographic signals acquired by the sensor module include the surface electromyographic amplitudes of the rectus abdominis and external oblique muscles; the method further includes: extracting features from the electromyographic signals using a convolutional neural network to assess the fatigue state of the target muscle group in real time; when the amplitude of the electromyographic signals of the target muscle group decreases by more than 30% and lasts for more than 15 seconds, the training intensity matching model automatically triggers a protective resistance adjustment mechanism, causing the elastic resistance torque during pullback to immediately decrease by 25%-30%, and prompting the user to adjust the range of motion through an interactive interface to avoid excessive muscle strain.

[0020] In some embodiments, the method further includes: after each training session, performing multi-dimensional analysis on the motion state data, physiological state data, and resistance adjustment data of that training session to generate a personalized training report; the personalized training report includes: an assessment of the balance of force exertion of the target muscle group, an effectiveness score of the resistance adjustment strategy, and a suggested range of training intensity for the next stage; wherein, the effectiveness score of the resistance adjustment strategy is obtained by calculating the degree of agreement between the actual resistance adjustment timing and the theoretical optimal adjustment timing based on a time series prediction model.

[0021] This application provides an intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance. It collects real-time motion state data (rolling displacement, speed, pull-back acceleration, force frequency) and physiological state data (heart rate, electromyography signals) using a sensor module. Combined with the user's basic body parameters (height, weight, body fat percentage, exercise ability level) and preset training target parameters (training duration, range of motion, energy consumption threshold), a training intensity matching model is constructed. This achieves closed-loop control of "data acquisition - state assessment - resistance adjustment," overcoming the limitations of existing technologies that rely on adjusting resistance using only a single or a few parameters. The innovative design features a strategy where the forward resistance increases non-linearly with increasing rolling displacement, and the pull-back resistance dynamically and adaptively adjusts with the pull-back speed. This matches the force characteristics of the human core muscle groups (moderate initial resistance during forward push, significantly increased resistance at the limit of displacement to enhance control, and dynamic resistance compensation based on speed during pull-back to avoid inertial impact or insufficient efficiency), solving the problem of mismatch between the linear resistance adjustment of traditional ab wheels and human biomechanics. By training an intensity matching model to calculate the matching degree between the current intensity and the target intensity in real time, and dynamically optimizing the resistance adjustment parameters based on algorithms such as fuzzy logic and neural networks, a personalized adaptive adjustment system is formed. This ensures that the resistance output always matches the user's real-time physical condition and training goals, thus avoiding the risk of overload and improving training efficiency.

[0022] By integrating multi-dimensional data and employing intelligent model computation, the system assesses user status in real time and dynamically adjusts resistance, overcoming the shortcomings of traditional ab wheel designs with a "one-size-fits-all" resistance. This allows training intensity to be tailored to individual differences and real-time changes in physical fitness, improving training effectiveness. A non-linear increasing resistance mechanism guides users to control the range of motion, preventing excessive displacement that could lead to lumbar compensation. Dynamically adjusted pull resistance suppresses the inertial impact of excessively rapid pullback. Combined with a fatigue protection mechanism based on electromyography (EMG) monitoring, this effectively reduces the risk of muscle strain. An algorithm matching training target parameters (such as energy expenditure thresholds) with user baseline parameters automatically optimizes resistance adjustment strategies, ensuring optimal load stimulation within a safe range for each training session. Furthermore, a phased training plan generation function (extended by dependent claims) enables continuous improvement in long-term training results.

[0023] In summary, this invention fills the technological gaps in existing intelligent ab wheel systems in terms of dynamic matching of training intensity, nonlinear resistance adjustment, and closed-loop feedback control through technological innovation, providing a brand-new approach for the design of intelligent fitness equipment.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic flowchart illustrating the steps of an intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance, provided in an embodiment of this application.

[0027] Figure 2 This is a schematic diagram of the structure of an intelligent ab wheel provided in one embodiment of this application;

[0028] Figure 3 This is a schematic block diagram of a smart matching system for ab wheel training intensity based on adaptive adjustment of elastic resistance, provided in one embodiment of this application.

[0029] Figure 4 This is a schematic block diagram of the structure of an intelligent ab wheel provided in one embodiment of this application.

[0030] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0033] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0034] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0035] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0036] The ab wheel is a widely used core muscle training tool, and its training effect is closely related to the proper matching of resistance intensity. Traditional ab wheels usually use a fixed resistance design (such as constant spring resistance or damping structure), which cannot dynamically adjust the resistance according to the user's real-time physical condition, movement accuracy, and training goals, leading to the following problems:

[0037] Insufficient precision in matching training intensity: Individual differences among users (such as height, weight, body fat percentage, and athletic ability) and changes in physical fitness during training (such as muscle fatigue and heart rate fluctuations) are not effectively quantified. Resistance adjustment relies solely on the user's subjective feelings or preset levels, which can easily lead to problems such as excessive resistance causing movement distortion (such as lumbar compensation and improper core muscle activation) or insufficient resistance leading to inadequate training effects.

[0038] The resistance adjustment mechanism is singular: the resistance adjustment of existing smart ab rollers is mostly based on linear adjustment of a single motion parameter (such as rolling displacement or speed), lacking real-time monitoring and comprehensive analysis of the user's physiological state (such as heart rate and electromyographic signals), and unable to form a closed-loop feedback of "data acquisition-model calculation-resistance adjustment", making it difficult to achieve dynamic adaptation of training intensity.

[0039] Lack of nonlinear resistance characteristics: During the forward push of a traditional ab wheel, the resistance usually increases linearly with the displacement or remains constant, which does not match the force characteristics of the human core muscle group (the initial force needs to overcome inertia, the middle stage needs to maintain stable force, and the extreme displacement needs to control balance); during the pull-back process, the resistance adjustment is not dynamically coupled with the pull-back speed, and it is impossible to intervene in real time when the user's movement speed is too fast (which may lead to muscle strain) or too slow (which results in low training efficiency).

[0040] While existing technologies have proposed sensor-based resistance adjustment schemes for fitness equipment, none have deeply integrated motion state data (rolling displacement, velocity, acceleration, force frequency) with physiological state data (heart rate, electromyographic signals), nor have they constructed a multi-dimensional training intensity matching model that includes the user's basic body parameters and training target parameters. This results in resistance adjustment strategies failing to take into account both personalized needs and the biomechanical characteristics of the movement. Furthermore, existing solutions lack the design for the non-linear increasing characteristics of forward resistance and the dynamic adaptive mechanism of pullback resistance, making it difficult to maximize training effectiveness while ensuring proper movement.

[0041] To resolve the above issues, please refer to... Figure 1 This application provides a method for intelligently matching the intensity of ab wheel training based on adaptive adjustment of elastic resistance, applicable to, for example... Figure 2 The smart ab wheel shown includes a wheel body, an elastic resistance adjustment device, a sensor module, and a control module. The elastic resistance adjustment device is connected to the wheel axle of the wheel body to provide variable elastic resistance. The sensor module is used to collect motion and physiological data of the user during training. The control module has a pre-stored training intensity matching model. It should be noted that all information involved in the method provided in this application was extracted with the authorization of the relevant user and in accordance with relevant regulations, and will not infringe on user privacy.

[0042] The provided intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance includes steps S101 to S103. Details are as follows:

[0043] Step S101. The sensor module acquires the user's motion state data during ab wheel training in real time. The motion state data includes at least the rolling displacement, rolling speed, pull-back acceleration, and force frequency of the ab wheel.

[0044] Specifically, the smart ab wheel uses its built-in sensor module to acquire real-time motion state data directly related to the mechanical properties of the user's movements during training. This includes: rolling displacement: the straight-line distance the ab wheel rolls forward from its initial position (close to the body), reflecting the range of motion; rolling speed: the rate of change of displacement per unit time during the push or pull-back process, reflecting the movement rate; pull-back acceleration: the acceleration value during the pull-back phase (returning from the extreme displacement to the initial position), used to determine whether the user is using inertial compensation to generate force; and force frequency: the number of complete movement cycles (push-back + pull-back) per unit time, reflecting the training rhythm.

[0045] By integrating a high-precision encoder or Hall sensor on the ab wheel axle, the number of wheel rotations is measured in real time, and the rolling displacement and speed are calculated by combining the wheel diameter; a triaxial accelerometer (such as a MEMS chip) is installed on the ab wheel body or grip, and the acceleration signal during the pull-back phase is separated by a Kalman filter algorithm to eliminate gravity interference; the frequency of grip force change when the user exerts force is monitored by a pressure sensor or strain gauge built into the grip, or the force frequency is calculated by the periodic characteristics of the displacement-time curve.

[0046] Data preprocessing ensures that the data accuracy meets the requirements of real-time control by performing noise reduction (such as low-pass filtering), time synchronization, and normalization on the raw sensor signals.

[0047] Breaking away from the limitations of traditional ab rollers that rely solely on monitoring a single displacement or speed, this device uses coupled analysis of displacement, speed, acceleration, and frequency to accurately characterize user movement features (such as whether the forward push is uniform and whether there is inertial impact during the pullback). This provides dynamic mechanical feedback for subsequent resistance adjustment. For example, if the pullback acceleration is too high, it can indicate that the user has a risk of lumbar compensation and trigger a resistance enhancement mechanism to suppress dangerous movements.

[0048] Step S102. Input the exercise state data and physiological state data into the training intensity matching model. The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on the preset training target parameters and the user's input basic body parameters. The training target parameters include preset training duration, preset standard range of motion, and preset energy consumption threshold. The basic body parameters include height, weight, body fat percentage, and exercise ability level.

[0049] Specifically, the motion state data and physiological state data (such as heart rate and electromyography signals) collected in step S101 are input into the training intensity matching model pre-stored in the control module. The model combines the user-input basic body parameters (height, weight, body fat percentage, and exercise ability level) and preset training target parameters (preset training duration, preset standard range of motion, and preset energy consumption threshold) to calculate the matching degree between the current training intensity and the target training intensity through a multi-dimensional weighted algorithm.

[0050] Data Input: Physiological Status Data: Heart rate and electromyographic signals of core muscle groups (such as rectus abdominis and erector spinae) are collected through a wristband heart rate sensor or electromyographic electrodes (surface EMG) integrated into the grip, reflecting real-time energy consumption and muscle activation level; Basic Body Parameters: Users input these parameters through the APP or device touchscreen upon first use, and the data is automatically synchronized with the body fat scale. Exercise ability level can be automatically assessed through initial movement tests (such as maximum forward push distance and number of standard movement repetitions); Training Goal Parameters: Users can customize these parameters in the APP, such as "30 minutes of core endurance training, standard range of motion 50cm, target energy consumption 200kcal".

[0051] Model Architecture: A hierarchical computational framework is adopted: Bottom Feature Layer: Standardizes motion data such as displacement and velocity, as well as heart rate and electromyography signals; Intermediate Mapping Layer: Establishes a mapping relationship between "basic parameters → target intensity threshold" through a regression model (such as random forest or BP neural network) (e.g., the reasonable resistance range is higher for users with larger body weight); Top Matching Layer: Calculates the Euclidean distance between the intensity index (such as actual energy consumption and movement standardization) under the current resistance and the preset target based on a multi-objective optimization algorithm (such as NSGA-II), and obtains the matching degree (0-100%) after normalization.

[0052] Breaking away from the traditional ab wheel's crude model of "preset gears + subjective adjustment," it achieves "personalized solutions" based on basic parameters such as height and weight. For example, it automatically increases the resistance ramp rate for users with high body fat percentage to enhance training stimulation. It integrates physiological data (such as judging declining physical fitness when heart rate exceeds a threshold) and exercise data (such as judging muscle fatigue when the range of motion decreases) to provide a scientific basis for dynamic resistance adjustment, avoiding the problem of "fixed resistance leading to inefficient training or injury."

[0053] Step S103. When the matching degree is lower than the preset matching threshold, a resistance adjustment command is generated according to the training intensity matching model. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel body rolls, so that the resistance of the ab wheel increases nonlinearly with the increase of the rolling displacement when the user pushes forward, and the resistance during the pull-back is dynamically and adaptively adjusted as the pull-back speed increases.

[0054] Specifically, when the matching degree is lower than a preset threshold (such as 80%), the control module drives the elastic resistance adjustment device to change the deformation degree of the elastic element (such as variable stiffness spring, shape memory alloy) according to the resistance adjustment instruction output by the training intensity matching model, or adjusts the number of meshing teeth through a gear meshing mechanism, thereby changing the elastic resistance moment when the abdominal wheel rolls, achieving: Forward push stage: The resistance increases non-linearly with the increase of the rolling displacement (such as in the form of a quadratic function or a piecewise function), matching the force characteristics of the core muscle group "initially overcoming inertia → stably exerting force in the middle → ultimately controlling balance"; Pull-back stage: The resistance is dynamically coupled with the pull-back speed (such as the resistance linearly increases with the increase of speed), suppressing the inertial compensation (such as borrowing force from the waist) caused by the user's excessive speed, or the low training efficiency caused by the excessive slowness of the speed.

[0055] Design of the resistance adjustment device: Control of the deformation of the elastic element: A servo motor is used to drive a screw-nut mechanism to adjust the spring compression amount or the stretching length in real time, and the non-linear adjustment of the resistance moment T = k⋅x (x is the displacement) is achieved by changing the spring stiffness coefficient (k) (such as making k increase with the increase of the displacement through a piecewise function); Gear meshing adjustment: Multiple gear sets are arranged on the wheel shaft, and the number of meshing teeth is controlled by an electromagnetic clutch. The more teeth are meshed, the greater the resistance moment, realizing discrete non-linear resistance gears, and dynamically switching in combination with the signal of the displacement sensor.

[0056] Control strategy: Forward push resistance curve: A non-linear function F(x) = ax 2 + bx + c (a > 0 to achieve increasing acceleration) is preset, and the coefficients are dynamically adjusted according to the user's movement ability level (such as for novice users, aa is smaller to avoid excessive initial resistance); Pull-back resistance strategy: When the pull-back speed v > v threshold, the resistance F(v) = kv linearly increases, forcing the user to reduce the speed; When v < v minimum efficient speed, the resistance is appropriately reduced to maintain the smoothness of the movement.

[0057] The non-linear resistance during the forward push conforms to the force characteristics of the core muscle group. The resistance is gentle in the initial stage (reducing the starting difficulty), and significantly increases at the limit position (strengthening the balance control ability), avoiding the problems of "too difficult in the early stage or too easy in the later stage" of traditional linear resistance; The dynamic resistance in the pull-back stage can intervene in dangerous actions in real time (such as a sharp increase in resistance during high-speed pull-back to protect the muscles), and at the same time avoid inefficient slow training, improving the training effect per unit time; Through the real-time closed-loop of "data collection - model calculation - resistance adjustment", the defects of the single resistance adjustment mechanism and the inability to adapt to physical fitness changes in the existing technology are solved. For example, when the user's fatigue causes movement deformation, the resistance threshold is automatically reduced to ensure the standardization of the movement.

[0058] Through the synergy of steps S101-S103, this method achieves a leap from "extensive subjective adjustment" to "precise dynamic adaptation": Data dimension: Integrating three types of data—mechanics, physiological state, and individual characteristics—to construct a multi-dimensional training model; Adjustment mechanism: Breaking through the linear adjustment of a single parameter, achieving composite control of nonlinear resistance in forward thrust and coupled resistance in pullback speed; User value: Maximizing training efficiency (matching resistance intensity as needed to avoid energy waste or insufficient stimulation) while ensuring proper movement (reducing lumbar compensation and muscle strain), truly realizing "personalized intelligent fitness".

[0059] In some embodiments, the training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and user-input basic body parameters. This includes: normalizing real-time collected motion state data and physiological state data; weighting the ab wheel rolling displacement, rolling speed, pull-back acceleration, force frequency, heart rate signal, and electromyography signal according to preset weights to obtain the current training intensity feature value; establishing a user fitness baseline model based on the basic body parameters; calculating the difference between the current training intensity feature value and the target intensity feature value corresponding to the preset training target parameters; and using the difference as the matching degree between the current training intensity and the target training intensity.

[0060] The training intensity matching model calculates the current training intensity feature value by weighting multidimensional data and compares it with the target intensity feature value to obtain the matching degree. Specifically, this includes: data normalization: standardizing motion state (displacement, velocity, acceleration, frequency) and physiological state (heart rate, electromyography) data to eliminate the influence of dimensions; weighted calculation: calculating the current training intensity feature value according to preset weights (e.g., core movement parameters weight 0.6, physiological parameters weight 0.4); fitness baseline model: establishing an individual fitness benchmark based on the user's basic body parameters (height, weight, etc.) and defining a reasonable intensity range; and difference calculation: calculating the difference between the current feature value and the target feature value using Euclidean distance or cosine similarity as the matching degree.

[0061] The normalization method employs Z-score standardization (x′=x−μ / σ), where μ and σ are the mean and standard deviation of historical training data. Weights are determined through expert experience or the analytic hierarchy process (AHP), such as rolling displacement (0.25), heart rate (0.2), and electromyography (EMG) signal (0.2). Baseline fitness modeling uses a linear regression model to fit the relationship between baseline parameters and maximum safe resistance and optimal training intensity range. Matching degree is calculated as: Matching degree = 1−||Xcurrent−Xtarget|| / ||Xtarget||, ranging from 0-100%. This avoids evaluation bias caused by a single parameter, achieving more accurate intensity matching through exercise-physiological data fusion. Reasonable intensity ranges are set based on individual differences to address the shortcomings of "one-size-fits-all" adjustments; for example, the resistance sensitivity coefficient is automatically reduced for target intensity characteristics of weak individuals. Mathematical quantification intuitively reflects the gap between current training and the target, providing a clear basis for dynamic adjustment.

[0062] In some embodiments, the dynamic generation of a phased training plan based on user motion data, personalized parameters, historical training fitness, and training goals, including movement types, training duration per set, rest intervals, and weight adjustment strategies, includes: determining the initial training difficulty level based on the user's input basic body parameters and historical training records; combining preset training goal parameters to generate a basic training plan including standard movement demonstrations, training duration per set, and rest intervals; during training, evaluating the user's movement standardization and physical exertion based on real-time collected motion and physiological data; after each training cycle, iteratively optimizing the movement types and resistance adjustment strategies in the basic training plan based on historical training fitness parameters to form a phased training plan adapted to the user's current physical fitness; the historical training fitness parameters include resistance adjustment frequency, pullback acceleration achievement rate, and heart rate recovery rate.

[0063] Dynamically generate phased training plans, comprising three core components: Initial plan generation: Determine the initial difficulty based on basic parameters and historical records, and generate a basic plan that includes movement demonstration, duration, and intervals; Real-time evaluation and optimization: Evaluate the effectiveness during training through movement standardization (e.g., whether the pullback acceleration exceeds the limit) and physical exertion (heart rate changes); Periodic iterative optimization: Adjust the movement type and resistance strategy based on historical fitness parameters (resistance adjustment frequency, acceleration achievement rate, heart rate recovery rate).

[0064] The initial difficulty level is divided into beginner (push distance <60cm), intermediate (60-80cm), and advanced (>80cm), corresponding to different initial resistance levels;

[0065] The training cycle is set up with 5-10 sets of exercises per cycle, each set lasting 5-10 minutes, and the rest interval is dynamically adjusted according to the heart rate recovery rate (e.g., the rest interval ends when the heart rate drops to 120% of the resting heart rate).

[0066] The fitness parameters include: resistance adjustment frequency: the number of times resistance is automatically adjusted per unit time, reflecting the needs of training intensity changes; pullback acceleration achievement rate: the percentage of times the pullback acceleration is ≤ the safety threshold, assessing the standardization of the movement; heart rate recovery rate: the rate at which the heart rate decreases during the rest period, reflecting the physical recovery ability.

[0067] Starting with a basic program, the difficulty is gradually increased through cyclical iterations to prevent beginners from getting injured due to a sudden increase in intensity; the program is adjusted based on real training feedback (such as reducing pull resistance when the pass rate is low) to solve the problem of "poor adaptability" of traditional fixed programs; not only is the resistance adjusted, but also the movement types (such as adding kneeling / standing variations) and rest intervals are optimized to improve the diversity and scientific nature of training.

[0068] In some embodiments, when the matching degree is lower than a preset matching threshold, generating a resistance adjustment instruction based on the training intensity matching model includes: if the matching degree is lower than a first preset threshold, increasing the deformation degree of the elastic element by a preset ratio according to the current rolling displacement and pull-back speed, so that the forward resistance increases by 5%-10% in the current displacement range; if the matching degree is lower than a second preset threshold, determining whether the user is in an overload state by combining real-time electromyography signals and heart rate signals, and after eliminating the risk of overload, switching the resistance level by controlling the number of meshing teeth of the elastic resistance adjustment device, so that the elastic resistance torque during pull-back increases exponentially with the increase of pull-back speed, while limiting the peak value of the forward resistance to not exceed the safe load threshold corresponding to the user's basic body parameters.

[0069] The resistance is adjusted according to the degree of mismatch, in two scenarios: mild mismatch (below the first threshold, such as 80%): increase the elastic deformation according to the ratio of displacement and velocity, and increase the incremental gradient of the forward resistance (5%-10%); severe mismatch (below the second threshold, such as 60%): first determine whether there is excessive load by electromyography / heart rate, and after eliminating the risk, switch gears so that the pullback resistance increases exponentially with the speed, while limiting the forward resistance to not exceed the safe load.

[0070] Threshold division: The first threshold (80%) corresponds to "fine-tuning required", and the second threshold (60%) corresponds to "significant adjustment required"; Slight adjustment: Increase the spring compression by using the servo motor to make the resistance function F(x)=(1+δ)F0(x), where δ is a 5%-10% adjustment coefficient;

[0071] Severe Adjustment: Overload Judgment: Risk is assessed when the electromyographic signal amplitude suddenly drops and the heart rate exceeds 85% of the maximum heart rate; adjustment is then paused. Resistance Gear Switching: The electromagnetic clutch switches to a higher gear set, with the pull-back resistance F(v) = kv. 2 (Exponentially increasing), the peak forward resistance is ≤ body weight × safety factor (e.g., 0.3).

[0072] To avoid overreaction caused by a "one-size-fits-all" approach to adjustment, mild problems can be addressed by continuous fine-tuning to maintain training fluidity. Before making severe adjustments, the risk of overload should be eliminated to prevent muscle strain caused by blindly increasing resistance. During the pullback phase, the exponential increase in resistance effectively suppresses inertial compensation. For example, during high-speed pullback, the sudden increase in resistance forces the user to actively control the speed.

[0073] In some embodiments, the training intensity matching model establishes a multi-dimensional state-space model that includes motion state parameters, physiological state parameters, and resistance adjustment parameters. It then uses fuzzy logic control or neural network algorithms to dynamically optimize the real-time resistance output of the elastic resistance adjustment device. This allows the adjusted elastic resistance to match the user's current physical condition and target training intensity, forming a closed-loop feedback adjustment. This enables the adaptive adjustment of elastic resistance and intelligent matching of training intensity during ab wheel training.

[0074] A multi-dimensional state-space model is constructed that includes motion (displacement, velocity), physiology (heart rate, electromyography), and resistance (degree of deformation, number of meshing teeth). Fuzzy logic or neural networks are used to dynamically optimize the resistance output, forming a closed loop of "data acquisition - model calculation - adjustment execution".

[0075] State space definition: State vector S=[x,v,a,f,HR,EMG,k,n], where k is the spring stiffness and nn is the number of meshing teeth; Fuzzy logic control: Input fuzzification: Displacement is divided into "short", "medium", and "long", and velocity is divided into fuzzy sets such as "slow", "moderate", and "fast"; Rule base design: such as "if the displacement is long and the velocity is fast, then the resistance is high"; Neural network control: A BP neural network is used, with the input layer being state parameters and the output layer being the resistance adjustment amount, and the weights are trained using historical data. It breaks through the limitations of single-parameter (such as displacement only) adjustment, simultaneously considering physiological state and resistance device parameters to achieve more precise dynamic balance; it continuously optimizes resistance through real-time feedback, for example, automatically reducing resistance when the user is fatigued (heart rate increases, electromyography decreases) to maintain proper movement; it supports the fast response of fuzzy logic and the nonlinear fitting of neural networks, adapting to complex training scenarios for different users.

[0076] For example, the neural network algorithm uses a long short-term memory network to construct a training intensity prediction model. The training intensity prediction model is trained offline using the user's historical training data. The input parameters include historical motion state data, physiological state data, resistance adjustment parameters, and training effect feedback. The output parameter is the resistance adjustment benchmark value for the next training cycle. During real-time training, the training intensity prediction model dynamically corrects the resistance adjustment benchmark value based on the current training data. The correction strategy includes: when the pull-back acceleration is lower than the benchmark value for three consecutive times, an automatic resistance reduction mechanism is triggered to reduce the elastic resistance torque by 10%-15% to avoid the user's movement being deformed due to excessive resistance. The training effect feedback includes the degree of muscle soreness and the movement completion score.

[0077] A training intensity prediction model is built using a Long Short-Term Memory (LSTM) network, trained offline using historical data, and the resistance adjustment baseline value is adjusted in real time. The adjustment strategy is as follows: if the pull-back acceleration is lower than the baseline value for three consecutive times, the resistance is reduced by 10%-15% to avoid distorted movement.

[0078] Model Construction: Input parameters: historical displacement, velocity, heart rate, resistance adjustment records, training effect feedback (soreness level, movement completion rate); Output parameters: next cycle resistance adjustment baseline value (e.g., the slope of the forward resistance increase); Offline training: using the Adam optimizer, the loss function is the mean squared error (MSE) between the predicted value and the actual adjusted value; Real-time correction: Baseline value calculation: based on the user's historical best movement data (e.g., the resistance value when acceleration reaches the target); Triggering condition: if the pull-back acceleration is less than 80% of the baseline value for three consecutive times, the resistance is judged to be too high, triggering a reduction. By capturing the time-series characteristics of the training data (e.g., fatigue accumulation trends) through LSTM, resistance is adjusted in advance rather than after the fact; precise intervention for abnormal pull-back acceleration effectively reduces dangerous movements such as lumbar compensation, especially suitable for safety protection during fatigue periods; baseline values ​​are generated based on the user's own historical data, solving the adjustment bias caused by differences in "force exertion habits" among different users.

[0079] For example, the fuzzy logic control includes multiple sets of fuzzy rules. The fuzzy rules use the user's heart rate change rate, force frequency fluctuation amplitude, and rolling displacement achievement rate as input variables, and resistance adjustment level as output variables. The fuzzy rules include: if the heart rate change rate exceeds a preset safety threshold and the force frequency decreases significantly, it is determined that the user is in a fatigued state, and a resistance reduction command is output to reduce the deformation of the elastic element by 20%-30%; if the rolling displacement achievement rate exceeds 90% for three consecutive times and the electromyographic signal intensity is stable, it is determined that the user's physical fitness has improved, and a resistance increase command is output to increase the pull-back resistance torque by 15%-20% by increasing the number of meshing teeth.

[0080] Multiple adjustment rules are designed based on fuzzy logic. The inputs are heart rate change rate, force frequency fluctuation, and displacement target achievement rate, and the output is the resistance adjustment level. For example, the resistance is lowered when the heart rate is too fast and the force frequency decreases, and the resistance is raised when the displacement target achievement rate is high and the electromyography is stable.

[0081] Input variable definitions: Heart rate change rate = (current heart rate − resting heart rate) / resting heart rate, divided into "low", "medium" and "high"; Force frequency fluctuation amplitude = |current frequency − average frequency| / average frequency, divided into "small", "medium" and "large"; Displacement target achievement rate = number of times the target was achieved / total number of times, divided into "<70%", "70%-90%" and ">90%";

[0082] Example of fuzzy rule: If (heart rate change rate is high AND force frequency fluctuates greatly) Then reduce resistance (20%-30%); If (displacement target achievement rate > 90% AND electromyography is stable) Then increase resistance (15%-20%); Output defuzzification: The specific adjustment amount is calculated using the center of gravity method to drive the elastic element or gear set to move.

[0083] No precise mathematical model is required; rules are formulated based on expert experience to adapt to nonlinear changes in sensor noise and user movements. Simultaneously, physical fitness (heart rate), movement stability (frequency fluctuations), and completion quality (achievement rate) are considered to avoid misjudgment based on a single indicator. When the achievement rate is consistently excellent, the resistance is automatically increased to motivate users to advance, which is in line with the "zone of proximal development" training theory.

[0084] For example, the training intensity matching model uses the number of times the user completes the standard movement, energy consumption efficiency, and movement standardization as reward functions, and resistance adjustment parameters as the movement space. It optimizes the elastic resistance adjustment strategy by constructing a Markov decision process. Specifically, in the early stage of training, an ε-greedy strategy is used to explore the optimal resistance adjustment range. After the user's cumulative training time exceeds 5 hours, the model switches to a strategy gradient-based optimization method. The response sensitivity of the resistance adjustment device is adjusted according to real-time reward feedback so that the nonlinear increasing curve of the forward resistance matches the user's muscle exertion characteristics.

[0085] A reinforcement learning model is constructed based on Markov Decision Process (MDP). The reward functions are the number of times the action is completed, energy efficiency, and standardization. The optimal resistance adjustment strategy is found through ε-greedy strategy (initial exploration) and policy gradient (later optimization) so that the forward resistance curve matches the muscle force characteristics.

[0086] MDP element definition: State St: current displacement, velocity, electromyography, heart rate; Action At: resistance adjustment parameters (degree of deformation, number of teeth engaged); Reward Rt: +10 points for completing the standard action, -5 points for action deformation, energy consumption efficiency (actual consumption / target consumption) ×10; Initial stage of strategy optimization (<5 hours): ε-greedy strategy (ε=0.2), 20% probability of randomly adjusting resistance to explore the optimal range; Later stage: strategy gradient algorithm (such as PPO), adjust the resistance adjustment sensitivity according to real-time rewards (such as adaptive resistance increase slope in the early stage of exertion).

[0087] Reinforcement learning automatically discovers the optimal resistance adjustment pattern for each individual, without relying on preset rules or expert experience; the forward resistance curve gradually fits the user's "start-stabilize-limit" force characteristics through reward feedback, for example, users with strong explosive power experience a greater increase in resistance in the limit displacement zone; as training time accumulates, the strategy becomes more and more in line with user habits, solving the problem of "decreased efficiency in the later stages" of traditional fixed strategies.

[0088] In some embodiments, the electromyographic signals acquired by the sensor module include the surface electromyographic amplitudes of the rectus abdominis and external oblique muscles; the method further includes: extracting features from the electromyographic signals using a convolutional neural network to assess the fatigue state of the target muscle group in real time; when the amplitude of the electromyographic signals of the target muscle group decreases by more than 30% and lasts for more than 15 seconds, the training intensity matching model automatically triggers a protective resistance adjustment mechanism, causing the elastic resistance torque during pullback to immediately decrease by 25%-30%, and prompting the user to adjust the range of motion through an interactive interface to avoid excessive muscle strain.

[0089] Electromyographic (EMG) signal features of the rectus abdominis and external oblique muscles are extracted using a convolutional neural network (CNN) to assess the fatigue status of the target muscle groups in real time. When the EMG amplitude decreases by more than 30% for 15 seconds, a protective mechanism is triggered: the pull resistance is reduced by 25%-30%, and the user is prompted to adjust their movements.

[0090] Electromyography (EMG) signal processing: Sensors: Surface electrode pads worn on the ab wheel handles or abdomen to collect 8-15kHz EMG signals; CNN model: Includes convolutional layers (for extracting temporal features), pooling layers (for dimensionality reduction), and fully connected layers (for classifying fatigue levels).

[0091] Fatigue assessment: Characteristic indicators: Root mean square value (RMS), mean absolute value (MAV); Triggering condition: RMS decreases by more than 30% and lasts for more than 15 seconds, indicating "moderate fatigue"; Protection mechanism: Immediately send a command to the elastic device to reduce resistance, and at the same time, the APP pops up a prompt "Please reduce the range of motion and pay attention to core strength".

[0092] Targeted monitoring of core muscle groups provides a more direct reflection of target muscle load than heart rate monitoring alone; proactive intervention before muscle fatigue leads to movement distortion reduces the risk of overuse injury, making it especially suitable for high-intensity training scenarios; combining electromyographic signals with movement data (such as displacement decrease) achieves dual protection of "muscle state-movement performance," for example, reducing resistance and prompting posture adjustment when fatigued.

[0093] In some embodiments, the method further includes: after each training session, performing multi-dimensional analysis on the motion state data, physiological state data, and resistance adjustment data of that training session to generate a personalized training report; the personalized training report includes: an assessment of the balance of force exertion of the target muscle group, an effectiveness score of the resistance adjustment strategy, and a suggested range of training intensity for the next stage; wherein, the effectiveness score of the resistance adjustment strategy is obtained by calculating the degree of agreement between the actual resistance adjustment timing and the theoretical optimal adjustment timing based on a time series prediction model.

[0094] After each training session, a personalized training report is generated through comprehensive analysis of multi-dimensional data. This report includes three core assessment modules: Target muscle group force balance assessment: Based on the differences in electromyographic signal amplitudes of core muscle groups such as the rectus abdominis and external oblique muscles, the report quantifies the force balance between the left and right sides or different muscle groups, identifying potential force compensation or muscle development imbalances; Resistance adjustment strategy effectiveness scoring: A time-series prediction model calculates the theoretically optimal resistance adjustment timing, compares it with the actual adjustment timing, and uses the degree of agreement as the scoring criterion to assess the timeliness and rationality of resistance adjustment; Next stage training intensity recommendation range: Combining historical training data, current physical condition (such as heart rate recovery rate and electromyographic fatigue threshold), and target training parameters, a personalized intensity recommendation is generated, including resistance adjustment range, exercise duration, and rest intervals.

[0095] Data acquisition and preprocessing automatically synchronize motion state data (displacement, velocity, acceleration), physiological data (heart rate, electromyography), and resistance adjustment data (adjustment time point, gear change) to the cloud or local database after training. The electromyography signals are bandpass filtered (5-500Hz) and rectified to extract the root mean square (RMS) values ​​of each muscle group as an indicator of force intensity.

[0096] Force balance is assessed by the difference rate of RMS between the left and right external oblique muscles (e.g., left and right external oblique muscles): Difference rate = |RMS left - RMS right| / 0.5 * (RMS left + RMS right) × 100%; a force balance radar chart is generated by muscle group dimension, and a difference rate > 20% is marked as "uneven", indicating that the user may have unilateral muscle compensation or weakness. A theoretical optimal adjustment timing prediction model is constructed: a Hidden Markov Model (HMM) or Long Short-Term Memory Network (LSTM) is used as input, historical training data (e.g., displacement before fatigue, heart rate rise trend) is input, and the "adjustment necessity probability" at each time point is output. When the probability > 80%, it is defined as the theoretical adjustment timing; the fit between the actual adjustment timing and the theoretical timing is calculated: the Dynamic Time Warping (DTW) algorithm is used to match the two time series, and the fit = number of matching points / total number of theoretical adjustment points × 100%, with a score range of 0-100 points. A score < 70 points is marked as "inefficient adjustment".

[0097] The training intensity recommendation generation includes: establishing an intensity recommendation regression model: input parameters include the current training fit, historical resistance adjustment frequency, and changes in the physical fitness baseline model (such as the increase in heart rate recovery rate); output parameters are the next stage resistance adjustment benchmark value (such as the recommended range of elastic element deformation), the increment of training time per set (±5 minutes), and the rest interval adjustment coefficient (such as increasing the rest interval by 10% for fatigued users); example of the recommended range: "The recommended range of pullback resistance torque for the next stage is 20-25 N*m (currently 22 N*m), the recommended training time per set is extended to 8 minutes, and the rest interval is shortened to 90 seconds (based on the improvement of heart rate recovery efficiency)."

[0098] Report presentation format: The visual interface includes charts (force balance radar chart, resistance adjustment timing comparison curve), scoring tables (effectiveness score, fitness progress index), and text suggestions (movement correction direction, intensity progression strategy); it supports users to export PDF reports or synchronize them to the sports and health APP to form a training history archive.

[0099] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of an intelligent matching system 200 for ab wheel training intensity based on adaptive adjustment of elastic resistance, provided in an embodiment of this application. This intelligent matching system 200 is used to execute the steps of the intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance shown in the above embodiments. The intelligent matching system 200 can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0100] like Figure 3 As shown, the ab wheel training intensity intelligent matching system 200 based on elastic resistance adaptive adjustment includes:

[0101] The data acquisition unit 201 is used to acquire the user's motion state data during ab wheel training in real time through the sensor module. The motion state data includes at least the rolling displacement, rolling speed, pull-back acceleration, and force frequency of the ab wheel.

[0102] The data input unit 202 is used to input motion state data and physiological state data into the training intensity matching model. The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user-input basic body parameters. The training target parameters include preset training duration, preset standard range of motion, and preset energy consumption threshold. The basic body parameters include height, weight, body fat percentage, and exercise ability level.

[0103] The dynamic adjustment unit 203 is used to generate a resistance adjustment command according to the training intensity matching model when the matching degree is lower than the preset matching threshold. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel body rolls, so that the resistance of the ab wheel increases nonlinearly with the increase of the rolling displacement when the user pushes forward, and the resistance during the pull-back is dynamically and adaptively adjusted as the pull-back speed increases.

[0104] In some embodiments, the training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and user-input basic body parameters. This includes: normalizing real-time collected motion state data and physiological state data; weighting the ab wheel rolling displacement, rolling speed, pull-back acceleration, force frequency, heart rate signal, and electromyography signal according to preset weights to obtain the current training intensity feature value; establishing a user fitness baseline model based on the basic body parameters; calculating the difference between the current training intensity feature value and the target intensity feature value corresponding to the preset training target parameters; and using the difference as the matching degree between the current training intensity and the target training intensity.

[0105] In some embodiments, the dynamic generation of a phased training plan based on user motion data, personalized parameters, historical training fitness, and training goals, including movement types, training duration per set, rest intervals, and weight adjustment strategies, includes: determining the initial training difficulty level based on the user's input basic body parameters and historical training records; combining preset training goal parameters to generate a basic training plan including standard movement demonstrations, training duration per set, and rest intervals; during training, evaluating the user's movement standardization and physical exertion based on real-time collected motion and physiological data; after each training cycle, iteratively optimizing the movement types and resistance adjustment strategies in the basic training plan based on historical training fitness parameters to form a phased training plan adapted to the user's current physical fitness; the historical training fitness parameters include resistance adjustment frequency, pullback acceleration achievement rate, and heart rate recovery rate.

[0106] In some embodiments, when the matching degree is lower than a preset matching threshold, generating a resistance adjustment instruction based on the training intensity matching model includes: if the matching degree is lower than a first preset threshold, increasing the deformation degree of the elastic element by a preset ratio according to the current rolling displacement and pull-back speed, so that the forward resistance increases by 5%-10% in the current displacement range; if the matching degree is lower than a second preset threshold, determining whether the user is in an overload state by combining real-time electromyography signals and heart rate signals, and after eliminating the risk of overload, switching the resistance level by controlling the number of meshing teeth of the elastic resistance adjustment device, so that the elastic resistance torque during pull-back increases exponentially with the increase of pull-back speed, while limiting the peak value of the forward resistance to not exceed the safe load threshold corresponding to the user's basic body parameters.

[0107] In some embodiments, the training intensity matching model establishes a multi-dimensional state-space model that includes motion state parameters, physiological state parameters, and resistance adjustment parameters. It then uses fuzzy logic control or neural network algorithms to dynamically optimize the real-time resistance output of the elastic resistance adjustment device. This allows the adjusted elastic resistance to match the user's current physical condition and target training intensity, forming a closed-loop feedback adjustment. This enables the adaptive adjustment of elastic resistance and intelligent matching of training intensity during ab wheel training.

[0108] For example, the neural network algorithm uses a long short-term memory network to construct a training intensity prediction model. The training intensity prediction model is trained offline using the user's historical training data. The input parameters include historical motion state data, physiological state data, resistance adjustment parameters, and training effect feedback. The output parameter is the resistance adjustment benchmark value for the next training cycle. During real-time training, the training intensity prediction model dynamically corrects the resistance adjustment benchmark value based on the current training data. The correction strategy includes: when the pull-back acceleration is lower than the benchmark value for three consecutive times, an automatic resistance reduction mechanism is triggered to reduce the elastic resistance torque by 10%-15% to avoid the user's movement being deformed due to excessive resistance. The training effect feedback includes the degree of muscle soreness and the movement completion score.

[0109] For example, the fuzzy logic control includes multiple sets of fuzzy rules. The fuzzy rules use the user's heart rate change rate, force frequency fluctuation amplitude, and rolling displacement achievement rate as input variables, and resistance adjustment level as output variables. The fuzzy rules include: if the heart rate change rate exceeds a preset safety threshold and the force frequency decreases significantly, it is determined that the user is in a fatigued state, and a resistance reduction command is output to reduce the deformation of the elastic element by 20%-30%; if the rolling displacement achievement rate exceeds 90% for three consecutive times and the electromyographic signal intensity is stable, it is determined that the user's physical fitness has improved, and a resistance increase command is output to increase the pull-back resistance torque by 15%-20% by increasing the number of meshing teeth.

[0110] For example, the training intensity matching model uses the number of times the user completes the standard movement, energy consumption efficiency, and movement standardization as reward functions, and resistance adjustment parameters as the movement space. It optimizes the elastic resistance adjustment strategy by constructing a Markov decision process. Specifically, in the early stage of training, an ε-greedy strategy is used to explore the optimal resistance adjustment range. After the user's cumulative training time exceeds 5 hours, the model switches to a strategy gradient-based optimization method. The response sensitivity of the resistance adjustment device is adjusted according to real-time reward feedback so that the nonlinear increasing curve of the forward resistance matches the user's muscle exertion characteristics.

[0111] In some embodiments, the electromyographic signals acquired by the sensor module include the surface electromyographic amplitudes of the rectus abdominis and external oblique muscles; it also includes: extracting features from the electromyographic signals using a convolutional neural network to assess the fatigue state of the target muscle group in real time; when the amplitude of the electromyographic signals of the target muscle group decreases by more than 30% and lasts for more than 15 seconds, the training intensity matching model automatically triggers a protective resistance adjustment mechanism, so that the elastic resistance torque during pullback is immediately reduced by 25%-30%, and prompts the user to adjust the range of motion through an interactive interface to avoid excessive muscle strain.

[0112] In some embodiments, the method further includes: after each training session, performing multi-dimensional analysis on the motion state data, physiological state data, and resistance adjustment data of that training session to generate a personalized training report; the personalized training report includes: an assessment of the balance of force exertion of the target muscle group, an effectiveness score of the resistance adjustment strategy, and a suggested range of training intensity for the next stage; wherein, the effectiveness score of the resistance adjustment strategy is obtained by calculating the degree of agreement between the actual resistance adjustment timing and the theoretical optimal adjustment timing based on a time series prediction model.

[0113] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the above-described intelligent matching system for ab wheel training intensity based on adaptive adjustment of elastic resistance and each module can be referred to the corresponding content in the various embodiments of the above-described intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance, and will not be repeated here.

[0114] The aforementioned intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance can be implemented as a computer program, which can, for example... Figure 3 It runs on the device shown.

[0115] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of the smart ab wheel provided in an embodiment of this application. The smart ab wheel includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0116] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance.

[0117] The processor provides computing and control capabilities to support the operation of the entire smart ab wheel.

[0118] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance.

[0119] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. A specific smart ab wheel may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0120] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0121] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0122] The sensor module acquires real-time motion data of the user during ab wheel training. The motion data includes at least the rolling displacement, rolling speed, pull-back acceleration, and force frequency of the ab wheel.

[0123] The exercise status data and physiological status data are input into the training intensity matching model. The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user's input basic body parameters. The training target parameters include preset training duration, preset standard range of motion, and preset energy consumption threshold. The basic body parameters include height, weight, body fat percentage, and exercise ability level.

[0124] When the matching degree is lower than the preset matching threshold, a resistance adjustment command is generated according to the training intensity matching model. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel body rolls, so that the resistance of the ab wheel increases nonlinearly with the increase of the rolling displacement when the user pushes forward, and the resistance during the pull-back is dynamically and adaptively adjusted as the pull-back speed increases.

[0125] In some embodiments, the training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and user-input basic body parameters. This includes: normalizing real-time collected motion state data and physiological state data; weighting the ab wheel rolling displacement, rolling speed, pull-back acceleration, force frequency, heart rate signal, and electromyography signal according to preset weights to obtain the current training intensity feature value; establishing a user fitness baseline model based on the basic body parameters; calculating the difference between the current training intensity feature value and the target intensity feature value corresponding to the preset training target parameters; and using the difference as the matching degree between the current training intensity and the target training intensity.

[0126] In some embodiments, the dynamic generation of a phased training plan based on user motion data, personalized parameters, historical training fitness, and training goals, including movement types, training duration per set, rest intervals, and weight adjustment strategies, includes: determining the initial training difficulty level based on the user's input basic body parameters and historical training records; combining preset training goal parameters to generate a basic training plan including standard movement demonstrations, training duration per set, and rest intervals; during training, evaluating the user's movement standardization and physical exertion based on real-time collected motion and physiological data; after each training cycle, iteratively optimizing the movement types and resistance adjustment strategies in the basic training plan based on historical training fitness parameters to form a phased training plan adapted to the user's current physical fitness; the historical training fitness parameters include resistance adjustment frequency, pullback acceleration achievement rate, and heart rate recovery rate.

[0127] In some embodiments, when the matching degree is lower than a preset matching threshold, generating a resistance adjustment instruction based on the training intensity matching model includes: if the matching degree is lower than a first preset threshold, increasing the deformation degree of the elastic element by a preset ratio according to the current rolling displacement and pull-back speed, so that the forward resistance increases by 5%-10% in the current displacement range; if the matching degree is lower than a second preset threshold, determining whether the user is in an overload state by combining real-time electromyography signals and heart rate signals, and after eliminating the risk of overload, switching the resistance level by controlling the number of meshing teeth of the elastic resistance adjustment device, so that the elastic resistance torque during pull-back increases exponentially with the increase of pull-back speed, while limiting the peak value of the forward resistance to not exceed the safe load threshold corresponding to the user's basic body parameters.

[0128] In some embodiments, the training intensity matching model establishes a multi-dimensional state-space model that includes motion state parameters, physiological state parameters, and resistance adjustment parameters. It then uses fuzzy logic control or neural network algorithms to dynamically optimize the real-time resistance output of the elastic resistance adjustment device. This allows the adjusted elastic resistance to match the user's current physical condition and target training intensity, forming a closed-loop feedback adjustment. This enables the adaptive adjustment of elastic resistance and intelligent matching of training intensity during ab wheel training.

[0129] For example, the neural network algorithm uses a long short-term memory network to construct a training intensity prediction model. The training intensity prediction model is trained offline using the user's historical training data. The input parameters include historical motion state data, physiological state data, resistance adjustment parameters, and training effect feedback. The output parameter is the resistance adjustment benchmark value for the next training cycle. During real-time training, the training intensity prediction model dynamically corrects the resistance adjustment benchmark value based on the current training data. The correction strategy includes: when the pull-back acceleration is lower than the benchmark value for three consecutive times, an automatic resistance reduction mechanism is triggered to reduce the elastic resistance torque by 10%-15% to avoid the user's movement being deformed due to excessive resistance. The training effect feedback includes the degree of muscle soreness and the movement completion score.

[0130] For example, the fuzzy logic control includes multiple sets of fuzzy rules. The fuzzy rules use the user's heart rate change rate, force frequency fluctuation amplitude, and rolling displacement achievement rate as input variables, and resistance adjustment level as output variables. The fuzzy rules include: if the heart rate change rate exceeds a preset safety threshold and the force frequency decreases significantly, it is determined that the user is in a fatigued state, and a resistance reduction command is output to reduce the deformation of the elastic element by 20%-30%; if the rolling displacement achievement rate exceeds 90% for three consecutive times and the electromyographic signal intensity is stable, it is determined that the user's physical fitness has improved, and a resistance increase command is output to increase the pull-back resistance torque by 15%-20% by increasing the number of meshing teeth.

[0131] For example, the training intensity matching model uses the number of times the user completes the standard movement, energy consumption efficiency, and movement standardization as reward functions, and resistance adjustment parameters as the movement space. It optimizes the elastic resistance adjustment strategy by constructing a Markov decision process. Specifically, in the early stage of training, an ε-greedy strategy is used to explore the optimal resistance adjustment range. After the user's cumulative training time exceeds 5 hours, the model switches to a strategy gradient-based optimization method. The response sensitivity of the resistance adjustment device is adjusted according to real-time reward feedback so that the nonlinear increasing curve of the forward resistance matches the user's muscle exertion characteristics.

[0132] In some embodiments, the electromyographic signals acquired by the sensor module include the surface electromyographic amplitudes of the rectus abdominis and external oblique muscles; the method further includes: extracting features from the electromyographic signals using a convolutional neural network to assess the fatigue state of the target muscle group in real time; when the amplitude of the electromyographic signals of the target muscle group decreases by more than 30% and lasts for more than 15 seconds, the training intensity matching model automatically triggers a protective resistance adjustment mechanism, causing the elastic resistance torque during pullback to immediately decrease by 25%-30%, and prompting the user to adjust the range of motion through an interactive interface to avoid excessive muscle strain.

[0133] In some embodiments, the method further includes: after each training session, performing multi-dimensional analysis on the motion state data, physiological state data, and resistance adjustment data of that training session to generate a personalized training report; the personalized training report includes: an assessment of the balance of force exertion of the target muscle group, an effectiveness score of the resistance adjustment strategy, and a suggested range of training intensity for the next stage; wherein, the effectiveness score of the resistance adjustment strategy is obtained by calculating the degree of agreement between the actual resistance adjustment timing and the theoretical optimal adjustment timing based on a time series prediction model.

[0134] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the intelligent matching method for ab wheel training intensity based on adaptive adjustment of elastic resistance as provided in any embodiment of this application.

[0135] The computer-readable storage medium can be the internal storage unit of the smart ab wheel described in the foregoing embodiments, such as the hard drive or memory of the smart ab wheel. Alternatively, the computer-readable storage medium can be an external storage device of the smart ab wheel, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the smart ab wheel.

[0136] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligently matching the training intensity of an ab wheel based on adaptive adjustment of elastic resistance, characterized in that, An application is made to a smart ab wheel, which includes a wheel body, an elastic resistance adjustment device, a sensor module, and a control module. The elastic resistance adjustment device is connected to the wheel axle of the wheel body to provide variable elastic resistance. The sensor module is used to collect motion and physiological data of the user during training. The control module has a pre-stored training intensity matching model. The method includes: The sensor module acquires real-time motion data of the user during ab wheel training. The motion data includes at least the rolling displacement, rolling speed, pull-back acceleration, and force frequency of the ab wheel. The exercise status data and physiological status data are input into the training intensity matching model. The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user's input basic body parameters. The training target parameters include preset training duration, preset standard range of motion, and preset energy consumption threshold. The basic body parameters include height, weight, body fat percentage, and exercise ability level. When the matching degree is lower than the preset matching threshold, a resistance adjustment command is generated according to the training intensity matching model. The resistance adjustment command controls the deformation degree or the number of meshing teeth of the elastic element in the elastic resistance adjustment device to change the elastic resistance torque when the ab wheel body rolls, so that the resistance of the ab wheel increases nonlinearly with the increase of the rolling displacement when the user pushes forward, and the resistance during the pull-back is dynamically and adaptively adjusted as the pull-back speed increases.

2. The method according to claim 1, characterized in that, The training intensity matching model calculates the matching degree between the current training intensity and the target training intensity based on preset training target parameters and the user's input basic body parameters, including: The real-time collected motion and physiological data are normalized, and the rolling displacement, rolling speed, pull-back acceleration, force frequency, heart rate signal, and electromyography signal of the ab wheel are weighted according to preset weights to obtain the current training intensity feature value. A user physical fitness baseline model is established based on the basic body parameters. The difference between the current training intensity feature value and the target intensity feature value corresponding to the preset training target parameter is calculated, and the difference is used as the matching degree between the current training intensity and the target training intensity.

3. The method according to claim 1, characterized in that, Based on user motion data, personalized parameters, historical training fitness, and training goals, a phased training plan is dynamically generated, including movement type, training duration for each set, rest intervals, and weight adjustment strategies. The initial training difficulty level is determined based on the user's input of basic body parameters and historical training records. Combined with preset training target parameters, a basic training plan is generated, which includes standard movement demonstrations, training duration for each set, and rest intervals. During training, the user's movement standardization and physical exertion are evaluated based on real-time collected motion and physiological data. After each training cycle, the movement types and resistance adjustment strategies in the basic training program are iteratively optimized based on historical training fitness parameters to form a phased training plan that is adapted to the user's current physical fitness. The historical training fitness parameters include resistance adjustment frequency, pullback acceleration achievement rate, and heart rate recovery rate.

4. The method according to claim 1, characterized in that, When the matching degree is lower than a preset matching threshold, generating a resistance adjustment instruction based on the training intensity matching model includes: If the matching degree is lower than the first preset threshold, the deformation degree of the elastic element is increased according to the current rolling displacement and pullback speed, so that the forward resistance increases by 5%-10% in the current displacement range. If the matching degree is lower than the second preset threshold, the real-time electromyography signal and heart rate signal are combined to determine whether the user is in an overload state. After eliminating the risk of overload, the resistance level is switched by controlling the number of meshing teeth of the elastic resistance adjustment device, so that the elastic resistance torque during pull-back increases exponentially with the speed of pull-back, while limiting the peak value of forward resistance to not exceed the safe load threshold corresponding to the user's basic body parameters.

5. The method according to claim 1, characterized in that, The training intensity matching model establishes a multi-dimensional state-space model that includes motion state parameters, physiological state parameters, and resistance adjustment parameters. It uses fuzzy logic control or neural network algorithms to dynamically optimize the real-time resistance output of the elastic resistance adjustment device, so that the adjusted elastic resistance can match the user's current physical condition and target training intensity, forming a closed-loop feedback adjustment. This enables the adaptive adjustment of elastic resistance and intelligent matching of training intensity during ab wheel training.

6. The method according to claim 5, characterized in that, The neural network algorithm uses a long short-term memory network to construct a training intensity prediction model. The training intensity prediction model is trained offline using the user's historical training data. The input parameters include historical motion state data, physiological state data, resistance adjustment parameters, and training effect feedback. The output parameter is the resistance adjustment benchmark value for the next training cycle. During real-time training, the training intensity prediction model dynamically corrects the resistance adjustment benchmark value based on the current training data. The correction strategy includes: when the pull-back acceleration is lower than the benchmark value for three consecutive times, the resistance reduction mechanism is automatically triggered to reduce the elastic resistance torque by 10%-15% to avoid the user's movement deformation due to excessive resistance. The training effect feedback includes the degree of muscle soreness and the score of movement completion.

7. The method according to claim 5, characterized in that, The fuzzy logic control includes multiple sets of fuzzy rules, which use the user's heart rate change rate, force frequency fluctuation amplitude, and rolling displacement achievement rate as input variables and resistance adjustment level as output variables. The fuzzy rules include: if the heart rate change rate exceeds a preset safety threshold and the force exertion frequency decreases significantly, the user is determined to be in a state of fatigue, and a resistance reduction command is output to reduce the deformation of the elastic element by 20%-30%; if the rolling displacement compliance rate exceeds 90% for three consecutive times and the electromyographic signal intensity is stable, the user's physical fitness is determined to be improved, and a resistance increase command is output to increase the pullback resistance torque by 15%-20% by increasing the number of meshing teeth.

8. The method according to claim 5, characterized in that, The training intensity matching model optimizes the elastic resistance adjustment strategy by constructing a Markov decision process, using the number of times the user completes the standard action, energy consumption efficiency, and action standardization as reward functions, and the resistance adjustment parameter as the action space. In this process, an ε-greedy strategy is used in the early stages of training to explore the optimal resistance adjustment range. When the user's training time exceeds 5 hours, the system switches to a strategy gradient-based optimization method. The response sensitivity of the resistance adjustment device is adjusted according to real-time reward feedback so that the non-linear increasing curve of the forward resistance matches the user's muscle exertion characteristics.

9. The method according to claim 1, characterized in that, The electromyographic signals acquired by the sensor module include the surface electromyographic amplitudes of the rectus abdominis and external oblique muscles; the method further includes: By extracting features from electromyographic signals using a convolutional neural network, the fatigue state of the target muscle group can be assessed in real time. When the amplitude of the electromyographic signal of the target muscle group decreases by more than 30% and lasts for more than 15 seconds, the training intensity matching model automatically triggers the protective resistance adjustment mechanism, which immediately reduces the elastic resistance torque during pullback by 25%-30%, and prompts the user to adjust the range of motion through the interactive interface to avoid excessive muscle strain.

10. The method according to claim 1, characterized in that, The method further includes: After each training session, the exercise status data, physiological status data, and resistance adjustment data of that training session are analyzed from multiple dimensions to generate a personalized training report. The personalized training report includes: an assessment of the balance of force exertion of the target muscle groups, an effectiveness score of the resistance adjustment strategy, and a suggested range of training intensity for the next stage. The effectiveness score of the resistance adjustment strategy is obtained by calculating the degree of agreement between the actual resistance adjustment timing and the theoretical optimal adjustment timing based on the time series prediction model.

Images