System and method for automatically generating training scheme to carry out anti-resistance training
Through the system and method of automatically generating training plans, using operating terminals, servers and fitness equipment, the training plans problem in the absence of coaching guidance is solved, and the remote generation and download of personalized training plans is realized, improving the quality and experience of fitness.
Patent Information
- Application Number
- CN202510261159.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-06
- Publication Date
- 2025-06-20
AI Technical Summary
In the absence of coaching guidance, fitness athletes cannot obtain personalized training plans and plans, which will affect the fitness effect and experience.
Through a system and method that automatically generates training schemes, using operating terminals, servers and fitness equipment to collect training data and generate personalized training schemes, users can remotely view and download the schemes to guide the actual training process.
It realizes remote automatic generation of personalized training plans, improves fitness quality and experience, and solves the problem of training plans when there is a lack of coaching guidance.
Smart Images
Figure CN120168934A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fitness management, and in particular to a system and method for automatically generating a training plan for resistance training. Background Art
[0002] In traditional fitness resistance training exercises, heavy objects such as barbells and dumbbells are used to provide resistance, and people resist the resistance of the heavy objects to carry out fitness exercises.
[0003] With the progress of the times and the development of technology, people now start to use motors to provide resistance to replace heavy objects such as dumbbells and barbells for resistance training, such as squats, deadlifts, bench presses, presses, etc., to achieve the purpose of exercising the body, enhancing health, and improving sports performance.
[0004] The user manually sets their own training plan parameters (such as training actions, resistance weight, number of training sets, number of repetitions, etc.) on the fitness equipment, and the equipment controls the motor to provide or stop resistance according to the set parameters, and people carry out resistance training by resisting the resistance.
[0005] However, fitness exercises require a local coach to give scientific training plans and schedules; if there is no local coach for mass fitness, targeted training plans and schedules cannot be given, which will affect the fitness experience and fitness effect of athletes. Summary of the Invention
[0006] Object of the Invention: The present invention provides a system and method for automatically generating a training plan for resistance training, which automatically generates a personalized training plan for users remotely and effectively improves the fitness quality and experience.
[0007] Technical Solution: A system for automatically generating a training plan for resistance training according to the present invention includes: an operation terminal, a server, and fitness equipment; the fitness equipment records the user's training process, collects training data, and transmits it to the server by wire or wirelessly. The server generates a user training plan and stores it in the user directory. The user connects to the server through the operation terminal, views the training plan, and guides the actual training process.
[0008] Further, the fitness equipment includes a fitness equipment network transmission module, a fitness equipment user login module, a fitness equipment configuration controller, a fitness equipment motor controller, and a fitness equipment motor; the user logs in to any fitness equipment through the fitness equipment user login module. The fitness equipment network transmission module connects to the server and downloads the training plan of the user stored on the server to the fitness equipment. The fitness equipment configuration controller loads the training plan of the user and parses the plan, and sends corresponding control commands to the fitness equipment motor controller. The fitness equipment motor starts, changes, or stops the resistance according to the instructions of the fitness equipment motor controller.
[0009] Correspondingly, a method for automatically generating a training plan for resistance training includes the following steps:
[0010] Step 1: The user logs in to any fitness device. The fitness device records the user's training process, collects training data, and transmits it to the server via wired or wireless means.
[0011] Step 2: The server generates the user's training plan and stores it in the user's directory.
[0012] Step 3: The user connects to the server through an operating terminal, views the automatically generated training plan, edits and modifies the training plan, and shares the training plan with other users. During training, the training plan is directly downloaded to the fitness device. The fitness device configures the fitness device motor according to the training plan, provides resistance for training, views the training plan, and guides the actual training process.
[0013] Further, in Step 2, the server generates the user's training plan and stores it in the user's directory. Initially, an initial training plan is automatically generated according to the classification, and later, according to the feedback of the user's training records, a targeted user training plan is automatically generated and stored in the user's directory.
[0014] Further, the initial training plan is automatically generated according to the classification, specifically including the following steps:
[0015] Step a: Conduct preliminary manual annotation and classification based on the user's basic information to construct an initial training template for different categories of users.
[0016] Step b: Consider each user as a data point, and each data point contains user basic information features in multiple dimensions.
[0017] Step c: For each subsequent new user, conduct classification learning and calculate the Euclidean distance between this new user and other existing users.
[0018] Step d: The new user starts training using the initial training template of the same category, records the user's training process, and uploads it to form a corresponding training record.
[0019] Further, in Step c, for two points P(x1, y1, …, z1) and Q(x2, y2, …, z2) in a multi-dimensional space, their Euclidean distance d is calculated using the following formula:
[0020]
[0021] Find the n users closest in distance. Check which category the majority of these n users belong to, and classify the new user into that category. The value of n is taken as the integer part of the square root of the number of samples in the dataset. Generally, n is taken as an odd number to facilitate calculating the majority in n. If the square root is an even number, add 1.
[0022] Furthermore, based on the feedback of the user training records in the later stage, a targeted user training plan is automatically generated, which specifically includes the following steps:
[0023] Step e: Use the training records of the users uploaded in the server user directory, and regard each training record of the user as a state set.
[0024] Step f: When the user executes the action recommended by the model, give a reward according to the feedback and actual effect.
[0025] Step g: The model updates the action value function using the reinforcement learning algorithm according to the obtained reward signal.
[0026] Step h: By continuously iterating the above process, the model learns which actions to take in different states to obtain the maximum cumulative reward, thereby optimizing the training plan, and finally obtaining the optimal training plan recommendation strategy for different user states.
[0027] Furthermore, in step e, the state set S includes the following parameters: user heart rate, body fat percentage, muscle percentage, group interval time, completion degree of training groups, completion degree of training duration, completion degree of weekly training, maximum strength; the action set A includes various training actions and corresponding training parameter adjustments.
[0028] Furthermore, in step f, the rewards include completion reward, user feedback reward, maximum strength feedback reward, and safety penalty feedback reward. Combine all the rewards to construct the reward function formula as follows:
[0029] R′ = completion reward + (w1r1 + w2r2) physiological feedback reward + λ·rfeedback manual feedback reward + rstrength strength feedback reward - I·5 safety penalty reward, where λ = 0.3 (manual feedback influence coefficient), and I is an indicator function (takes 1 when there is a violation, otherwise 0).
[0030] Furthermore, in step g, the model updates the action value function using the reinforcement learning algorithm according to the obtained reward signal; the model updates the action value function Q(s,a) using the reinforcement learning algorithm according to the obtained reward signal, and the formula is as follows:
[0031] Q(s,a) ← Q(s,a) + α[R′ + γmaxa′Q(s′,a′) - Q(s,a)]
[0032] Where: Q(s,a) is the action value of taking action a in state s; α is the learning rate, which is 0.2, and a strategy of dynamically adjusting the learning rate is adopted. As the training progresses, the learning rate is gradually reduced to improve the stability of the model; R′ is the immediate reward obtained after taking action a, which is set according to the reward mechanism; γ is the discount factor, which determines the influence of future rewards on the current action value, and is 0.95; s′ is the new state reached after taking action a; maxa′Q(s′,a′) is the maximum action value of all possible actions in the new state s′.
[0033] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the present invention remotely and automatically generates personalized training plans for users, and automatically downloads training plans at the same time, controls the motor on any fitness equipment, matches the logged-in user to perform resistance training, and effectively improves the quality and experience of fitness. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 It is a schematic diagram of the system structure of the present invention.
[0035] Figure 2 It is a schematic diagram of the method flow of the present invention.
[0036] Figure 3 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION
[0037] like Figure 1 As shown, a system for automatically generating training plans for resistance training includes: an operation terminal, a server and fitness equipment; a user logs in to any fitness equipment, the fitness equipment records the user's training process, collects training data, and transmits it to the server via wired or wireless means, the server generates a user training plan and stores it in the user directory, the user connects to the server via the operation terminal, views the training plan, and guides the actual training process.
[0038] Training data includes training movements, resistance weights, number of training sets, number of movements, group intervals, training duration, movement rate, user heart rate, body fat percentage, etc. The heart rate is collected by the heart rate sensor in the handle of the fitness equipment, the body fat percentage and muscle percentage are collected by the body fat scale that comes with the fitness equipment, and the remaining parameters are collected by the motor controller of the fitness equipment.
[0039] The fitness equipment includes a fitness equipment network transmission module, a fitness equipment user login module, a fitness equipment configuration controller, a fitness equipment motor controller, and a fitness equipment motor; the user logs in to any fitness equipment through the fitness equipment user login module, the fitness equipment network transmission module is connected to the server, downloads the training plan of this user stored on the server to the fitness equipment, the fitness equipment configuration controller loads the training plan of this user and parses the plan, and sends corresponding control commands to the fitness equipment motor controller, and the fitness equipment motor starts, changes, or stops the resistance according to the instructions of the fitness equipment motor controller.
[0040] As Figure 2 and Figure 3 shown, a method for automatically generating a training plan for resistance training includes the following steps:
[0041] Step 1: The user logs in to any fitness equipment, the fitness equipment records the user's training process, collects training data, and transmits it to the server by wired or wireless means;
[0042] Step 2: The server generates the user's training plan and stores it in the user's directory.
[0043] Step 3: The user connects to the server through the operation terminal, can view the automatically generated training plan, and can edit, modify the training plan, or share the training plan with other users; during training, the training plan is directly downloaded to the device, and the device configures the motor according to the training plan to provide resistance for training; view the training plan to guide the actual training process.
[0044] In Step 2, the server generates the user's training plan and stores it in the user's directory. Initially, an initial training plan is automatically generated according to the classification, and later, according to the feedback of the user's training records, a targeted user training plan is automatically generated and stored in the user's directory.
[0045] Initially, an initial training plan is automatically generated according to the classification, which specifically includes the following steps:
[0046] Step a: Conduct preliminary manual annotation classification according to the user's basic information. The user submits basic information through the operation terminal: fitness goals, age, gender, height, weight, weekly training time, etc., and stores it in the server user directory; construct initial training templates for different categories of users. The fitness coach, according to professional knowledge, manually classifies the first batch of users according to the basic information stored by the users, and creates initial training templates for each category of users.
[0047] Step b: Regard each user as a data point, and each data point contains user basic information features in multiple dimensions.
[0048] Step c: For each subsequent new user, conduct classification learning and calculate the Euclidean distance between this new user and other existing users. For two points P(x1, y1, …, z1) and Q(x2, y2, …, z2) in a multi-dimensional space, the Euclidean distance d between them is calculated using the following formula:
[0049]
[0050] Find the n users with the closest distances and see which category the majority of these n users belong to, then assign the new user to that category. The value of n is taken as the integer part of the square root of the number of samples in the dataset. Generally, n is taken as an odd number to facilitate calculating the majority in n. If the square root is an even number, then add 1.
[0051] Step d: The new user starts training using the initial training template of the same category and records the user's training process, uploading it to form the corresponding training record.
[0052] In the later stage, according to the feedback of the user training record, a targeted user training plan is automatically generated, which specifically includes the following steps:
[0053] Step e: Use the training records of the users uploaded in the server user directory, and regard each training record of the user as a state set. The state set S contains the following parameters: user heart rate, body fat percentage, muscle percentage, group interval time, completion degree of training groups, completion degree of training duration, completion degree of weekly training, maximum strength. The action set A includes various training actions of fitness equipment and the corresponding training parameter adjustments.
[0054] Step f: When the user executes the actions recommended by the model, give rewards according to the feedback and actual effects. Specifically:
[0055] (1) Completion reward: Reward according to the completion degree of the training plan given by the system on the fitness equipment by the user, as follows: Positive reward (+8): Training plan completion degree ≥ 80%; Negative reward (-6): Training plan completion degree < 60%.
[0056] (2) User feedback reward, including user physiological parameter feedback and user manual feedback: User physiological parameter feedback; Use the user's physiological parameters for feedback, and in order to balance multiple fitness goals (such as fat loss, muscle gain), adopt a multi-objective optimization strategy - by assigning dynamic weights to each goal, the training plan can be flexibly adjusted according to the user's real-time physiological state and goal progress. Set two goals, fat loss and muscle gain, and dynamically allocate their weights: Goal 1 fat loss: f_fat_loss = (1 - current body fat percentage / initial body fat percentage) × 100%; Goal 2 muscle gain: f_muscle_gain = (current muscle mass - initial muscle mass) / initial muscle mass × 100%; The weight adjustment algorithm, the formula is as follows:
[0057]
[0058] The parameter β = 0.5 (temperature coefficient, controlling the adjustment range), and r is the reward score set according to the objective functions f for fat loss and f for muscle gain;
[0059] User manual feedback; When the user executes the training plan, they may manually adjust the training plan generated by the system according to their own feelings or needs. These manual adjustments are part of the user feedback and are recorded by the system and fed back to the model. For example: The user may reduce the weight or number of sets of a certain exercise; The user may increase the weight or number of sets of a certain exercise; The user may replace a certain exercise or adjust the training order; Set the manual feedback reward rule:
[0060] rfeedback =
[0061] {+2 if the user increases the weight or number of sets of a certain exercise
[0062] -2 if the user reduces the weight or number of sets of a certain exercise
[0063] -1 if the user adjusts the training order
[0064] -3 if the user replaces a certain exercise}
[0065] At the same time, according to the user's manual feedback, maintain an action blacklist: Disabling probability = number of negative feedbacks / (total number of executions + 1) × 100%. For example: The squat exercise receives 3 negative feedbacks of cancellation / 10 executions → 30% probability of being replaced.
[0066] (3) Maximum strength feedback reward: Maximum strength is an important indicator in fitness. As feedback, it will be recorded by the system and fed back to the model, such as:
[0067] Positive reward (+10): Maximum strength increase > 10%
[0068] Positive reward (+5): Maximum strength increase > 5%
[0069] Positive reward (+3): Maximum strength maintained within ±2%
[0070] Negative reward (-5): Maximum strength decrease > 5%
[0071] Negative reward (-10): Maximum strength decrease > 10%
[0072] (4) Safety penalty feedback reward: When the user's heart rate exceeds the threshold ((220 - age) × 80%) during the training process, a negative reward of -5 will be given.
[0073] Calculate the comprehensive reward and construct the reward function; combine all rewards to construct the reward function formula as follows: R′ = completion reward + (w1r1 + w2r2) physiological feedback reward + λ·rfeedback manual feedback reward + rstrength strength feedback reward - I·5 safety penalty reward, where λ = 0.3 (manual feedback influence coefficient) and I is an indicator function (taking 1 when there is a violation and 0 otherwise).
[0074] Step g: The model updates the action value function using the reinforcement learning algorithm based on the obtained reward signal; the model updates the action value function Q(s,a) using the reinforcement learning algorithm according to the obtained reward signal, and the formula is as follows:
[0075] Q(s,a)←Q(s,a)+α[R′+γmaxa′Q(s′,a′)-Q(s,a)]
[0076] Where: Q(s,a) is the action value of taking action a in state s; α is the learning rate, taking 0.2, and a strategy of dynamically adjusting the learning rate is adopted. As the training progresses, the learning rate is gradually decreased to improve the stability of the model; R′ is the immediate reward obtained after taking action a, set according to the reward mechanism; γ is the discount factor, which determines the influence degree of future rewards on the current action value, taking 0.95; s′ is the new state reached after taking action a; maxa′Q(s′,a′) is the maximum action value of all possible actions in the new state s′.
[0077] Step h: By continuously iterating the above process, the model learns which actions to take in different states to obtain the maximum cumulative reward, thereby optimizing the training plan and finally obtaining the optimal training plan recommendation strategy for different user states.
[0078] When a user logs in to any fitness device, the fitness device connects to the server through the network transmission module and downloads the training plan of this user stored on the server to the fitness device. The fitness device configures the controller module to load the user's training plan, parse the plan, and send corresponding control commands to the fitness device motor controller. The fitness device motor controller controls the fitness device motor according to the commands to start, change or stop the resistance; the user can carry out the training in sequence according to the plan.
[0079] The present invention combines the model with the actual application: combining the usage process of the fitness device, generating a personalized training plan according to the user's basic information and training records, and dynamically adjusting the training plan according to the user's actual completion degree, action feedback and fitness effect on the device. And it ensures that when the user logs in to any fitness device, they can match and obtain their own training plan, and feedback to the server in a timely manner, making the training plan have the characteristics of persistence and professionalism.
[0080] The model is comprehensively practical: Multi-objective optimization: By assigning dynamic weights to each fitness goal, the system can flexibly adjust the training plan during the training process according to the user's real-time physiological state and goal progress, balancing multiple fitness goals. User feedback mechanism: The system can dynamically adjust the training plan based on the user's physiological parameter feedback and manual feedback, improving the user's training experience and satisfaction. Safety penalty: By monitoring physiological indicators such as the user's heart rate, the system can give negative rewards in a timely manner when the user has safety problems, ensuring the safety of training.
[0081] Through the above solutions, the system can dynamically adjust the training plan according to the user's real-time state and feedback during each training, improving the training effect; for each user, providing a unique user experience, solving the dilemma of no coach guidance around the user, and achieving long-term fitness effects.
Claims
1. A system for automatically generating training plans for resistance training, characterized in that: include: operating terminals, servers, and fitness equipment; The fitness equipment records the user's training process, collects training data, and transmits it to the server via wired or wireless means. The server generates a user training plan and stores it in the user's directory. The user connects to the server through an operation terminal to view the training plan and guide the actual training process.
2. The system for automatically generating a training program for performing resistance training according to claim 1, characterized in that: The fitness equipment includes a fitness equipment network transmission module, a fitness equipment user login module, a fitness equipment configuration controller, a fitness equipment motor controller and a fitness equipment motor; the user logs in to any fitness equipment through the fitness equipment user login module, the fitness equipment network transmission module is connected to the server, and the training plan of the user stored on the server is downloaded to the fitness equipment, the fitness equipment configuration controller loads the training plan of the user and parses the plan, and sends a corresponding control command to the fitness equipment motor controller, and the fitness equipment motor starts, changes or stops resistance according to the instructions of the fitness equipment motor controller.
3. A method for applying the system for automatically generating training programs for resistance training as claimed in claim 1, characterized in that: The steps include: Step 1: The user logs in to any fitness device, and the fitness device records the user's training process, collects training data, and transmits it to the server via wired or wireless means; Step 2: The server generates a user training plan and saves it in the user directory. Step 3: The user connects to the server through the operation terminal, views the automatically generated training plan, edits and modifies the training plan, and shares the training plan with other users; During training, the training plan is directly downloaded to the fitness equipment. The fitness equipment configures the fitness equipment motor according to the training plan, provides resistance for training, and checks the training plan to guide the actual training process.
4. The method for automatically generating a training program for resistance training according to claim 3, characterized in that: In step 2, the server generates a user training plan and stores it in the user directory. In the early stage, an initial training plan is automatically generated according to the classification. In the later stage, a targeted user training plan is automatically generated according to the feedback of the user training record and stored in the user directory.
5. The method for automatically generating a training program for resistance training according to claim 4, characterized in that: In the early stage, the initial training plan is automatically generated according to the classification, which includes the following steps: Step a: Perform preliminary manual labeling and classification based on the basic information of users to construct initial training templates for different categories of users; Step b: Consider each user as a data point, and each data point contains multiple dimensions of basic user information features; Step c: For each subsequent new user, classification learning is performed to calculate the Euclidean distance between the new user and other existing users; Step d: The new user starts training using the initial training template of the same category, and records the user's training process and uploads it to form the corresponding training record.
6. The method for automatically generating a training program for resistance training according to claim 5, characterized in that: In step c, for two points P(x1, y1, ..., z1) and Q(x2, y2, ..., z2) in multidimensional space, the Euclidean distance d between them is calculated using the following formula: Find the n nearest users, see which category the majority of these n users belong to, and assign the new user to that category; the value of n is taken as the integer part of the square root of the number of samples in the data set. n is generally an odd number to facilitate the calculation of the largest number in n. If the square root is an even number, add 1.
7. The method for automatically generating a training program for resistance training according to claim 4, characterized in that: Later, according to the feedback from user training records, a targeted user training plan is automatically generated, which includes the following steps: Step e: using the training records of the user uploaded in the server user directory, and treating each training record of the user as a state set; Step f: After the user performs the action recommended by the model, rewards are given based on feedback and actual results; Step g: The model updates the action value function using a reinforcement learning algorithm based on the reward signal obtained; Step h: By continuously iterating the above process, the model learns which actions to take in different states to obtain the maximum cumulative reward, thereby optimizing the training plan and ultimately obtaining the optimal training plan recommendation strategy under different user states.
8. The method for automatically generating a training program for resistance training according to claim 7, characterized in that: In step e, the state set S includes the following parameters: user heart rate, body fat percentage, muscle percentage, group interval time, training group number completion, training duration completion, weekly training completion, and maximum strength; the action set A includes various training actions and corresponding training parameter adjustments.
9. The method for automatically generating a training program for resistance training according to claim 7, characterized in that: In step e and step f, the rewards include completion reward, user feedback reward, maximum strength feedback reward and safety penalty feedback reward. All rewards are combined to construct the reward function formula as follows: R′=completion reward+(w1r1+w2r2) physiological feedback reward+λ·rfeedback manual feedback reward+rstrength strength feedback reward-I·5 safety penalty reward, λ=0.3 (manual feedback influence coefficient), I is the indicative function, which takes 1 when there is a violation and 0 otherwise.
10. The method for automatically generating a training program for resistance training according to claim 7, characterized in that: Step g: The model updates the action value function Q(s,a) using the reinforcement learning algorithm based on the reward signal obtained. The model updates the action value function Q(s,a) using the reinforcement learning algorithm based on the reward signal obtained. The formula is as follows: Q(s,a)←Q(s,a)+α[R′+γmaxa′Q(s′,a′)-Q(s,a)] Where: Q(s,a) is the action value of taking action a in state s; α is the learning rate, which is 0.2, and a strategy of dynamically adjusting the learning rate is adopted. As the training progresses, the learning rate is gradually reduced to improve the stability of the model; R′ is the immediate reward obtained after taking action a, which is set according to the reward mechanism; γ is the discount factor, which determines the influence of future rewards on the current action value, which is 0.95; s′ is the new state reached after taking action a; maxa′Q(s′,a′) is the maximum action value of all possible actions in the new state s′.