A method and device for generating an exercise prescription, an electronic device and a storage medium

Abstract

This invention relates to the field of sports and health management technology, and discloses a method, device, electronic device, and storage medium for generating exercise prescriptions. The method includes: acquiring user data; generating an exercise prescription and a predicted adaptation index based on this data using a large language model; instructing the user to exercise according to the exercise prescription and acquiring relevant data for the current exercise stage; if the preset exercise goal is not achieved, performing a three-dimensional ability assessment to obtain a comprehensive muscle strength index; quantifying the user's actual adaptation index at the current exercise stage; determining exercise intensity constraint parameters based on the actual adaptation index; optimizing the large language model; regenerating the exercise prescription and predicted adaptation index using the optimized large language model; and returning to the step of instructing the user to exercise according to the exercise prescription until the preset exercise goal is achieved. This invention, through a closed-loop feedback mechanism, enables the model to continuously absorb individual user exercise data, making the exercise prescription conform to individual exercise patterns and improving the accuracy and scientific nature of the exercise prescription.

Description

Technical Field

[0001] This invention relates to the field of sports and health management technology, specifically to a method, apparatus, electronic device, and storage medium for generating exercise prescriptions. Background Technology

[0002] Exercise prescriptions are commonly used in the medical field, fitness industry, elderly care, and corporate health management. In the medical field, exercise prescriptions can help doctors and patients manage chronic diseases and recover from surgery; in the fitness industry, exercise prescriptions can provide customers with personalized training programs; in elderly care, exercise prescriptions can reduce the risk of falls among the elderly; in corporate health management, exercise prescriptions can improve the physical fitness of employees; in addition, exercise prescriptions can be used for the scientific training of athletes to achieve health promotion for the entire population in multiple scenarios.

[0003] Currently, personalized exercise prescriptions are typically generated based on user health data (such as heart rate, body fat percentage, and exercise capacity) and medical examinations, combined with the FITT-VP principle (type, intensity, time, frequency, total amount, and progress). For example, by integrating data from wearable devices and medical examination reports, the prescription intensity can be dynamically adjusted, and features such as video guidance and progress tracking can be provided, covering scenarios such as chronic disease management, post-operative rehabilitation, and muscle building, achieving closed-loop management of assessment-prescription-execution-feedback.

[0004] However, the above-mentioned methods for generating exercise prescriptions are greatly affected by equipment errors and subjective user input, resulting in large errors and poor scientific validity in the exercise prescriptions. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and storage medium for generating exercise prescriptions, in order to solve the problem that existing exercise prescription generation methods are greatly affected by equipment errors and subjective user input, resulting in large errors and poor scientific accuracy in exercise prescriptions.

[0006] In a first aspect, the present invention provides a method for generating an exercise prescription, the method comprising: The system acquires user data and uses a large language model to generate exercise prescriptions and predictive adaptation indices based on the user data, which includes preset exercise goals. The system allows users to exercise according to a prescription and obtains relevant data about the user's current exercise stage. When a user completes the current exercise phase but fails to reach the preset exercise goal, a three-dimensional ability assessment is performed on the user based on user data and related data to obtain a comprehensive muscle strength index. Based on relevant data and a comprehensive muscle strength index, the user's degree of adaptation at the current exercise stage is quantified to obtain the actual adaptation index. Determine the exercise intensity constraint parameters based on the actual adaptation index; Based on the exercise prescription, predicted adaptation index, actual adaptation index, and exercise intensity constraint parameters, the large language model is optimized. The optimized large language model is then used to regenerate the user's exercise prescription and predicted adaptation index, returning to the steps of enabling the user to exercise according to the exercise prescription and obtaining relevant data of the user at the current exercise stage, until the user reaches the preset exercise goal.

[0007] This invention acquires user data and uses a large language model to generate an initial exercise prescription and a predicted adaptation index. Users then follow this prescription, and relevant data on the user's current exercise stage is obtained. If the user fails to reach the preset exercise goal, a three-dimensional ability assessment is performed to obtain a comprehensive muscle strength index, quantifying the user's exercise ability at the current stage and further quantifying their adaptation level to obtain an actual adaptation index. Based on this, exercise intensity constraint parameters are determined, providing a precise basis for subsequent model optimization and prescription adjustment, ensuring that the exercise intensity always matches the user's real-time adaptation level, balancing ability improvement with the risk of overtraining. The large language model is optimized based on the exercise prescription, the aforementioned indices, and parameters. The optimized large language model is then used to generate another exercise prescription and predicted adaptation index, ensuring that the regenerated exercise prescription better matches the user's real-time ability and adaptation status, solving the problems of rigid prescriptions and lagging adjustments in traditional methods. Users then exercise according to the new exercise prescription and execute subsequent processes until the preset exercise goal is reached. A closed-loop feedback mechanism allows the model to continuously absorb individual user exercise data, ensuring that the exercise prescription aligns with individual exercise patterns and improving the accuracy and scientific validity of the exercise prescription.

[0008] In one alternative implementation, the user data includes peak concentric and peak eccentric forces from a lower limb motor ability test, and related data includes the user's concentric and eccentric forces during the current movement phase. Based on user data and related data, a three-dimensional ability assessment is conducted on the user to obtain a comprehensive muscle strength index, including: Force capacity is calculated based on the peak force of centripetal motion and the peak force of centrifugal motion. Based on the centripetal and centrifugal forces of the user in the current phase of motion, calculate stability and balance. A comprehensive muscle strength index is determined based on strength, stability, and balance.

[0009] This embodiment assesses a user's strength, stability, and balance from three dimensions to obtain a comprehensive muscle strength index, which quantifies the user's ability to move at the current stage of exercise and provides a core basis for decision-making in subsequent processes.

[0010] In one alternative implementation, the relevant data also includes motion accuracy score, heart rate variability score, subjective fatigue level, mental state score, number of completed exercises, and target number of exercises. Based on relevant data and a comprehensive muscle strength index, the user's adaptation level at the current exercise stage is quantified to obtain an actual adaptation index, including: Based on the comprehensive muscle strength index and movement accuracy score, the score of the current exercise performance is calculated. Get the user's performance score in the previous exercise phase, and calculate the performance improvement rate based on the user's performance scores in the current exercise phase and the previous exercise phase. The recovery quality index is calculated based on heart rate variability score, subjective fatigue level, and psychological state score. The ratio of the number of completed exercises to the target number of exercises is defined as the exercise consistency index. The actual adaptation index is determined based on the rate of improvement in athletic performance, the recovery quality index, and the athletic consistency index.

[0011] This embodiment determines the actual adaptation index based on the improvement rate of athletic performance, the recovery quality index, and the exercise consistency index. This index is used to measure the user's physiological and performance adaptation to a given load during the current training phase. This helps to dynamically adjust the exercise prescription, ensuring that the user can gradually improve their abilities while avoiding overtraining.

[0012] In one optional implementation, the large language model is optimized based on the exercise prescription, the predicted adaptation index, the actual adaptation index, and the exercise intensity constraint parameters, including: The actual exercise intensity is determined based on the exercise intensity and exercise intensity constraint parameters of the exercise prescription. Calculate intensity matching loss based on exercise intensity prescribed by exercise prescription and actual exercise intensity. The adaptive loss is calculated based on the predicted and actual adaptation indices. Calculate safety loss based on exercise intensity and preset safety thresholds according to exercise prescriptions; The training loss is calculated based on the strength matching loss, the adaptive loss, and the safety loss. Optimize large language models based on training loss.

[0013] This embodiment calculates training loss from multiple dimensions and optimizes the model accordingly, specifically enhancing its adaptability to exercise prescription generation scenarios. This includes more accurate exercise intensity estimation, more realistic adaptation index prediction, and stricter safety constraints, gradually reducing the deviation between the exercise prescription generated by the model and the individual's actual situation. This makes the exercise prescription more in line with the user's actual situation and improves the scientificity and accuracy of the exercise prescription.

[0014] In an alternative implementation, after regenerating the user's exercise prescription and predicted adaptation index using the optimized large language model, the method further includes: Determine the intensity deviation between the actual exercise intensity and the exercise intensity of the regenerated exercise prescription; When the intensity deviation exceeds the first preset threshold, an expert system is used to conduct a safety review of the regenerated exercise prescription.

[0015] This embodiment performs a safety review when the deviation between the exercise intensity of the regenerated exercise prescription and the actual exercise intensity exceeds a first preset threshold, ensuring that the exercise prescription not only meets individual needs but also complies with medical-grade safety standards.

[0016] In one alternative implementation, the method further includes: Obtain the three-dimensional motion trajectory of multiple key joints of the user in the current motion phase; Based on the standard motion library, the deviation of the three-dimensional motion trajectory of each key joint is determined; When the deviation exceeds the second preset threshold, a posture correction suggestion is generated.

[0017] This embodiment acquires the three-dimensional motion trajectory of multiple key joints of the user, compares it with a standard motion library, and generates posture adjustment suggestions when the deviation is greater than a second preset threshold, thereby reducing the risk of sports injuries caused by non-standard posture.

[0018] In a second aspect, the present invention provides an exercise prescription generation device, the device comprising: The first generation module is used to acquire user data and use a large language model to generate user exercise prescriptions and predictive adaptation indices based on user data. User data includes preset exercise goals. The first acquisition module is used to enable users to exercise according to the exercise prescription and acquire relevant data of the user at the current exercise stage. The assessment module is used to conduct a three-dimensional ability assessment of the user based on user data and related data when the user completes the current exercise phase but does not reach the preset exercise goal, and obtains a comprehensive muscle strength index. The quantification module is used to quantify the user's adaptation level in the current exercise phase based on relevant data and comprehensive muscle strength index, and obtain the actual adaptation index. The first determining module is used to determine the motion intensity constraint parameters based on the actual adaptation index; The second generation module is used to optimize the large language model based on the exercise prescription, predicted adaptation index, actual adaptation index and exercise intensity constraint parameters. The optimized large language model is then used to regenerate the user's exercise prescription and predicted adaptation index, and return to the steps of enabling the user to exercise according to the exercise prescription and obtaining relevant data of the user in the current exercise stage, until the user reaches the preset exercise goal.

[0019] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the exercise prescription generation method described in the first aspect or any corresponding embodiment thereof.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the exercise prescription generation method described in the first aspect or any of its corresponding embodiments.

[0021] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the exercise prescription generation method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart of an exercise prescription generation method according to an embodiment of the present invention; Figure 2 This is a flowchart of another method for generating exercise prescriptions according to an embodiment of the present invention; Figure 3 This is a structural block diagram of an exercise prescription generation device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] Traditional methods for generating exercise prescriptions are significantly affected by equipment errors and subjective user input, resulting in large errors and poor scientific accuracy. This invention utilizes a closed-loop feedback mechanism to enable the model to continuously absorb individual user exercise data, ensuring that the exercise prescription aligns with individual exercise patterns and improving its accuracy and scientific validity.

[0028] According to an embodiment of the present invention, an embodiment of a method for generating exercise prescriptions is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] This embodiment provides a method for generating exercise prescriptions. Figure 1 This is a flowchart of an exercise prescription generation method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain user data, and use a large language model to generate the user's exercise prescription and predicted adaptation index based on the user data. The user data includes preset exercise goals.

[0030] Specifically, comprehensive multi-dimensional user data is collected to provide the foundation for generating exercise prescriptions. This user data encompasses three core dimensions: basic information, exercise information, and testing information, as detailed below: (1) Basic information: Individual basic attributes such as gender, age, height, weight, and BMI (Body Mass Index), as well as health risk-related data such as historical diseases, sports injury sites, sports injury types, and sports injury medical history, while also covering basic physiological indicators such as blood pressure, resting heart rate, and daily heart rate. The above data is mainly collected by users filling out health questionnaires and by equipment testing (such as blood pressure monitors and heart rate monitors); (2) Exercise information: Past exercise habits data such as weekly exercise frequency, duration of each exercise session, and weekly strength training frequency; subjective needs data such as exercise preferences (e.g., type of movement, training scenario preference); and preset exercise goals (e.g., increasing the peak strength of lower limb eccentric movements from 500N to 650N). This type of data is used to match the user's exercise habits and needs, avoiding a disconnect between the prescription and the user's actual situation; (3) Test information: including peak concentric and eccentric strength measured in the lower limb motor ability test, subjective fatigue level self-assessed by the user (using Borg RPE 1-6 level scoring), and psychological state assessment score, etc.

[0031] After obtaining the aforementioned user data, a safety screening can be conducted using a disease characteristic database and a sports risk questionnaire to determine whether the user is suitable for exercise, thereby mitigating the risk of sports injuries from the outset. The disease characteristic database contains a contraindication map for various diseases, which can quickly match the user's historical medical history, sports injury history, and other data to identify potential health risks. The sports risk questionnaire focuses on key dimensions such as abnormal physiological indicators, undisclosed medical history, and exercise tolerance, supplementing the collection of subjective health information and improving the risk assessment dimensions. If both screenings are passed (i.e., no clear exercise contraindications and physiological indicators meet safety thresholds), the process proceeds to the next step; if either screening fails (e.g., the database matches a high-risk disease, or the questionnaire indicates uncontrolled blood pressure abnormalities), all subsequent operations are terminated.

[0032] The large language model uses the Qwen3-8B large language model as its core decision engine. It has undergone supervised fine-tuning using knowledge from the sports and health domain (such as the *ACSM Guidelines for Exercise Testing and Prescription*, *ACSM Introduction to Physical Training*, and basic medical and sports health knowledge). The underlying architecture retains a Decoder-only Transformer architecture and incorporates optimized designs such as a grouped query attention mechanism, rotational position encoding, SwiGLU activation structure, and RMSnorm pre-normalization. While maintaining language understanding and reasoning capabilities, it also possesses the ability to accurately apply professional sports science logic, generating a complete exercise prescription in one go using an autoregressive approach. This large language model is used to perform reasoning processing based on the aforementioned multi-dimensional user data to generate exercise prescriptions and predict adaptation indices. Optionally, the processing steps of this large language model are existing technologies and will not be elaborated upon here.

[0033] The exercise prescriptions generated by the large language model are targeted training programs. For example, if a 40-year-old adult male's preset exercise goal is to improve lower limb muscle strength, the model generates a two-month exercise prescription, specifying multiple exercise stages and the content of each stage (such as seated leg raises, standing calf raises, etc.), the number of repetitions, the intensity (muscle strength training load) of each exercise, and the duration of each exercise (e.g., 30-40 minutes). The predictive adaptation index is the large language model's prediction of the user's degree of adaptation to the exercise prescription.

[0034] Step S102 involves having the user exercise according to the exercise prescription and obtaining relevant data about the user at the current exercise stage.

[0035] Specifically, after wearing the monitoring device, the user exercises according to the prescribed exercise formula, and relevant data is collected during the current exercise phase. This data includes centripetal and eccentric forces, movement accuracy score, heart rate variability score, subjective fatigue level, psychological state score, number of completed exercises, and target number of exercises. The movement accuracy score is an expert's assessment of the user's proper form during the current exercise phase; the heart rate variability score is a quantified score of the heart rate variability (HRV) monitored by the device, converted according to preset standards; the subjective fatigue level is the user's subjective score after the current exercise phase ends; the psychological state score is obtained by the user completing a standardized scale reflecting their physical and mental adaptation after the current exercise phase ends; the number of completed exercises is the actual number of exercises completed; and the target number of exercises is the number of exercises the user plans to complete. By collecting relevant data from the user during the current exercise phase, the system comprehensively covers both objective performance and subjective feelings during exercise, ensuring that subsequent calculations more closely reflect the user's actual state and providing a foundation for model optimization and dynamic prescription adjustments.

[0036] Step S103: When the user completes the current exercise phase but fails to reach the preset exercise goal, a three-dimensional ability assessment is performed on the user based on user data and related data to obtain a comprehensive muscle strength index.

[0037] Specifically, after a user completes the current exercise phase, the system determines whether the preset exercise goal has been achieved based on the relevant data from that phase. If not, the user needs to continue exercising; if the goal has been achieved, the exercise stops. When the user fails to reach the preset exercise goal, the system evaluates the user based on user data and relevant data collected during the current exercise phase, assessing strength, stability, and balance to obtain a comprehensive muscle strength index, thus quantifying the user's exercise ability during the current phase.

[0038] Step S104: Based on relevant data and comprehensive muscle strength index, quantify the user's adaptation level in the current exercise stage to obtain the actual adaptation index.

[0039] Specifically, based on relevant data and a comprehensive muscle strength index, the user's level of adaptation at the current exercise stage is quantified to obtain the Stage Adaptation Index (SAI).

[0040] Step S105: Determine the motion intensity constraint parameters based on the actual adaptation index.

[0041] Specifically, when the SAI is high (e.g., SAI ≥ 0.75), it indicates that the user has strong adaptability to exercise load, and the exercise intensity constraint parameter can be set to 5%~15% to promote ability improvement by increasing exercise intensity. When the SAI is in the medium range (e.g., 0.45 ≤ SAI < 0.75), it is considered that the user is well adapted to the current load, and the current exercise intensity is maintained, with the exercise intensity constraint parameter set to 0. When the SAI is low (e.g., SAI < 0.45), it indicates that there may be fatigue accumulation, insufficient recovery, or the current load exceeding the tolerance range, and the exercise intensity constraint parameter can be set to -10%~-5% to ensure training safety and sustainability by appropriately reducing exercise intensity. Optionally, when the SAI is low, the rest interval between exercise phases can be appropriately extended to optimize the user's recovery effect. By mapping the actual adaptation index to a specific exercise intensity constraint parameter, a precise basis for intensity adjustment is provided for subsequent model optimization and prescription adjustment, ensuring that the exercise intensity always matches the user's real-time adaptation level and balances ability improvement with the risk of overtraining. Optionally, the above-mentioned SAI range can be dynamically adjusted according to different preset exercise goals, and this embodiment of the invention does not limit this.

[0042] Step S106: Based on the exercise prescription, predicted adaptation index, actual adaptation index, and exercise intensity constraint parameters, optimize the large language model, regenerate the user's exercise prescription and predicted adaptation index using the optimized large language model, return to the step of having the user exercise according to the exercise prescription, and obtain relevant data of the user in the current exercise stage, until the user reaches the preset exercise goal.

[0043] Specifically, based on the exercise prescription, predicted adaptation index, actual adaptation index, and exercise intensity constraint parameters, the large language model is optimized, allowing the model to continuously learn the user's individual exercise patterns. The optimized large language model is then used to regenerate the exercise prescription and predicted adaptation index, ensuring that the regenerated exercise prescription better matches the user's real-time ability and adaptation status, solving the problems of rigidity and lagging adjustments in traditional prescriptions. Based on the regenerated exercise prescription, the process returns to step S102, repeating the above process until the user reaches the preset exercise goal. This closed-loop feedback mechanism allows the model to continuously absorb the user's individual exercise data, ensuring that the exercise prescription aligns with individual exercise patterns and improving the accuracy and scientific validity of the exercise prescription.

[0044] It's important to note that the large language model generates a complete exercise prescription each time. Users will continue exercising according to the new prescription, rather than restarting the exercise process. For example, if a user is currently performing the second stage of the original exercise prescription, after the exercise prescription is regenerated, the user will directly transition to the third stage, ensuring the continuity and gradual progression of the exercise.

[0045] This invention acquires user data and uses a large language model to generate an initial exercise prescription and a predicted adaptation index. Users then follow this prescription, and relevant data on the user's current exercise stage is obtained. If the user fails to reach the preset exercise goal, a three-dimensional ability assessment is performed to obtain a comprehensive muscle strength index, quantifying the user's exercise ability at the current stage and further quantifying their adaptation level to obtain an actual adaptation index. Based on this, exercise intensity constraint parameters are determined, providing a precise basis for subsequent model optimization and prescription adjustment, ensuring that the exercise intensity always matches the user's real-time adaptation level, balancing ability improvement with the risk of overtraining. The large language model is optimized based on the exercise prescription, the aforementioned indices, and parameters. The optimized large language model is then used to generate another exercise prescription and predicted adaptation index, ensuring that the regenerated exercise prescription better matches the user's real-time ability and adaptation status, solving the problems of rigid prescriptions and lagging adjustments in traditional methods. Users then exercise according to the new exercise prescription and execute subsequent processes until the preset exercise goal is reached. A closed-loop feedback mechanism allows the model to continuously absorb individual user exercise data, ensuring that the exercise prescription aligns with individual exercise patterns and improving the accuracy and scientific validity of the exercise prescription.

[0046] This embodiment provides a method for generating exercise prescriptions, which specifically includes the following steps: Step S201: Obtain user data, and use a large language model to generate the user's exercise prescription and predicted adaptation index based on the user data. The user data includes preset exercise goals. For details, please refer to [link to details]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0047] Step S202 involves instructing the user to exercise according to the exercise prescription and obtaining relevant data on the user's current exercise stage. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0048] Step S203: When the user completes the current exercise phase but does not reach the preset exercise goal, a three-dimensional ability assessment is performed on the user based on user data and related data to obtain a comprehensive muscle strength index. The user data includes the peak concentric and eccentric strength of the lower limb motor ability test, and the related data includes the user's concentric and eccentric strength in the current exercise phase.

[0049] Specifically, step S203 includes: Step S2031: Calculate the force capability based on the peak force of centripetal motion and the peak force of centrifugal motion.

[0050] Specifically, the user's strength ability can be calculated by the following formula (1), which can reflect the user's maximum muscle strength level during muscle contraction and extension, and provide data for determining the range of motion resistance.

[0051] (1) In the formula, Indicates strength and ability; Indicates the peak force during centripetal motion; This indicates the peak force of the centrifugal motion; This represents the standard reference value, which is the average peak strength of a group of people of the same age, gender, and BMI as this user.

[0052] Step S2032: Calculate stability and balance based on the user's centripetal and centrifugal forces during the current motion phase.

[0053] Specifically, the user's stability ability is calculated by the following formula (2) to quantify the consistency and stability of the user's force output in the current exercise phase. The higher the value, the more stable and controllable the user's force output is, and the more suitable it is for training with higher loads. The lower the value, the greater the force fluctuation, and the more necessary it is to reduce sudden loads or adjust the rhythm of movements to avoid injury caused by unstable force exertion.

[0054] (2) In the formula, Indicates stability capability; F This represents the overall force time series consisting of centripetal and centrifugal forces in the current phase of motion. Indicates standard deviation; This represents the average value.

[0055] The user's balance ability is calculated by the following formula (3) to quantify the symmetry of the ratio of centripetal and centrifugal force, which helps to ensure a balance in load distribution and avoid injury caused by unilateral overload leading to muscle imbalance.

[0056] (3) In the formula, Indicates balance ability; This represents the average value of the centripetal force during the current phase of motion. This represents the average value of the centrifugal force during the current phase of motion.

[0057] Step S2033: Determine the comprehensive muscle strength index based on strength, stability, and balance.

[0058] Specifically, the weights of the three abilities mentioned above are dynamically assigned based on the user's preset exercise goals, and the sum of the three weights is 1. For example, if the preset exercise goal is a muscle-building quantitative indicator (such as a 30% increase in peak concentric strength of the lower limbs), then the weight of strength ability is assigned higher; if the preset exercise goal is a rehabilitation quantitative indicator, then the weights of stability ability and balance ability are assigned higher. The comprehensive muscle strength ability index is obtained by weighting and summing the three abilities and their weights to quantify the user's exercise ability at the current stage of exercise, providing a core decision-making basis for subsequent processes.

[0059] Step S204: Based on relevant data and comprehensive muscle strength index, quantify the user's adaptation level in the current exercise stage to obtain the actual adaptation index. Relevant data also include movement accuracy score, heart rate variability score, subjective fatigue level, psychological state score, number of completed exercises, and target number of exercises.

[0060] Specifically, step S204 includes: Step S2041: Calculate the performance score of the current movement phase based on the comprehensive muscle strength index and movement accuracy score.

[0061] Specifically, weights are dynamically assigned to the comprehensive muscle strength index and movement accuracy score based on actual needs, and the sum of the two weights is 1. The comprehensive muscle strength index and movement accuracy score, along with their respective weights, are weighted and summed to obtain the user's training performance score for the current exercise phase.

[0062] Step S2042: Obtain the user's performance score in the previous exercise phase, and calculate the performance improvement rate based on the user's performance scores in the current exercise phase and the previous exercise phase.

[0063] Specifically, obtain the user's performance score in the previous exercise stage. If the current exercise stage is the first exercise stage, then the performance score in the previous exercise stage is 0. Calculate the performance improvement rate using the following formula (4) to reflect the user's progress compared to the previous exercise stage.

[0064] (4) In the formula, Indicates the rate of improvement in athletic performance; t Index indicating the phase of motion; This indicates the score for training performance at the current stage of exercise; This indicates the score for performance in the previous stage of the exercise.

[0065] Step S2043: Calculate the recovery quality index based on heart rate variability score, subjective fatigue level, and psychological state score.

[0066] Specifically, weights are dynamically assigned to heart rate variability score, subjective fatigue score, and mental state score based on actual needs, with the sum of the three weights being 1. The heart rate variability score, subjective fatigue score, and mental state score, along with their respective weights, are weighted and summed to obtain a recovery quality index. This index quantifies the user's recovery level after the current exercise phase. A higher value indicates a better recovery state, suitable for higher intensity or more intensive exercise; a lower value suggests the need to extend the recovery interval or reduce the intensity of subsequent exercise.

[0067] Step S2044: The ratio of the number of completed exercises to the target number of exercises is determined as the exercise consistency index.

[0068] Specifically, the ratio of completed exercise sessions to target exercise sessions is defined as the exercise consistency index to quantify whether users are completing their exercise prescriptions as planned. A higher value indicates better user compliance and better assurance of the continuity of exercise effects; a lower value suggests that the feasibility of the exercise prescription needs to be optimized.

[0069] Step S2045: Determine the actual adaptation index based on the improvement rate of athletic performance, the recovery quality index, and the athletic consistency index.

[0070] Specifically, weights are dynamically assigned to the three indicators mentioned above based on actual needs, and the sum of the three weights is 1. The three indicators and their corresponding weights are weighted and summed to obtain the actual adaptation index, which is used to measure the user's physiological and performance adaptation to a given load in the current training phase. Its value is directly proportional to the exercise intensity, which is conducive to dynamically adjusting the exercise prescription and ensuring that the user can gradually improve their ability while avoiding overtraining.

[0071] Step S205: Determine the motion intensity constraint parameters based on the actual adaptation index. For details, please refer to [link to relevant documentation]. Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0072] Step S206: Based on the exercise prescription, predicted adaptation index, actual adaptation index, and exercise intensity constraint parameters, optimize the large language model, regenerate the user's exercise prescription and predicted adaptation index using the optimized large language model, return to the step of having the user exercise according to the exercise prescription, and obtain relevant data of the user in the current exercise stage, until the user reaches the preset exercise goal.

[0073] Specifically, step S206 above optimizes the large language model based on the exercise prescription, predicted adaptation index, actual adaptation index, and exercise intensity constraint parameters, including: Step S2061: Determine the actual exercise intensity based on the exercise intensity and exercise intensity constraint parameters of the exercise prescription.

[0074] Specifically, the exercise intensity constraint parameter is determined by the user's actual adaptation index at the current exercise stage. This index is then summed with the exercise intensity predicted by the large language model for the exercise prescription to obtain the actual exercise intensity. This ensures that the exercise intensity matches the user's actual adaptation level, fully aligns with the user's current muscle strength level, recovery status, and load tolerance, and avoids excessive intensity leading to injury or insufficient intensity affecting the exercise effect.

[0075] For example, assuming that the large language model predicts an exercise intensity of 55% of maximum muscle strength for a certain exercise prescription, and the exercise intensity constraint parameter is -6%, then the actual exercise intensity = 55% + (-6%) = 49% of maximum muscle strength.

[0076] It should be noted that the number of repetitions for each phase of exercise is usually more than one. The exercise intensity here refers to the intensity of each repetition. The actual exercise intensity for each repetition can be obtained through the steps described above.

[0077] Step S2062: Calculate intensity matching loss based on the exercise intensity of the exercise prescription and the actual exercise intensity.

[0078] Specifically, the intensity matching loss is calculated using the following formula (5) to measure the deviation between the training intensity predicted by the model and the actual exercise intensity, which helps to adjust the exercise prescription so that the exercise intensity matches the user's actual ability. The smaller the loss value, the higher the degree of fit between the predicted exercise intensity and the actual exercise intensity, and the stronger the scientific validity of the prescription intensity.

[0079] (5) In the formula, Indicates the intensity matching loss; This indicates the total number of repetitions in the current phase of the exercise. An index representing the number of movements; Indicates the first The intensity of the exercise; Indicates the first The actual intensity of the exercise.

[0080] Step S2063: Calculate the adaptation loss based on the predicted adaptation index and the actual adaptation index.

[0081] Specifically, the adaptation loss is calculated using the following formula (6) to measure the difference between the user's actual adaptation performance in the current exercise phase and the predicted adaptation index output by the model, which helps to adjust the exercise prescription and dynamically track the individual's exercise progress.

[0082] (6) In the formula, Indicates loss of adaptation; S Indicates the actual adaptability index; This indicates the predictive adaptation index.

[0083] Step S2064: Calculate safety loss based on the exercise intensity of the exercise prescription and the preset safety threshold.

[0084] Specifically, the preset safety threshold is a preset reasonable or safe upper limit for exercise intensity. The safety loss is calculated using the following formula (7).

[0085] (7) In the formula, Indicates a loss of security; This indicates a preset safety threshold.

[0086] If the exercise intensity of the exercise prescription does not exceed the preset safety threshold, the safety loss is 0; if it does exceed the threshold, a penalty is applied by squaring the intensity, forcing the model to reduce the intensity during subsequent optimization.

[0087] Step S2065: Calculate the training loss based on the strength matching loss, the adaptive loss, and the safety loss.

[0088] Specifically, weights are dynamically assigned to the three losses according to actual needs, and the sum of the three weights is 1. The training loss is obtained by weighted summing of the three losses and their corresponding weights.

[0089] Step S2066: Optimize the large language model based on the training loss.

[0090] Specifically, with minimizing the training loss as the optimization objective, the backpropagation algorithm is used to feed the training loss back to each layer of the model network, achieving precise fine-tuning of parameters. During the optimization process, the LoRA parameter efficient fine-tuning mechanism is employed to specifically enhance the model's adaptability to the exercise prescription generation scenario while maintaining its original inference capabilities. This includes more accurate exercise intensity estimation, more realistic adaptation index prediction, and stricter safety constraint control. By optimizing the model based on the training loss, the deviation between the model-generated exercise prescription and the individual's actual situation is gradually reduced, making the exercise prescription more consistent with the user's actual situation and improving the scientific validity and accuracy of the exercise prescription.

[0091] Step S2067: Determine the intensity deviation between the actual exercise intensity and the exercise intensity of the regenerated exercise prescription.

[0092] Specifically, the optimized model is used to regenerate the exercise prescription, and the absolute value of the difference between the exercise intensity at the current exercise stage and the actual exercise intensity is calculated as the intensity deviation, which reflects the difference between the individual's adapted intensity and the intensity predicted by the model.

[0093] Step S2068: When the intensity deviation is greater than the first preset threshold, the exercise prescription is reviewed for safety using an expert system.

[0094] Specifically, when the intensity deviation exceeds a first preset threshold (e.g., 20%), it indicates a significant difference between the actual exercise intensity and the model's predicted intensity, posing a risk of sports injury due to excessive intensity adjustment. In this case, the expert system is automatically triggered to conduct a medical-grade safety review. The expert system employs a collaborative working mode combining an expert rule base (containing multiple clinical sports medicine rules) and an AI self-learning module. It combines user data, comprehensive muscle strength index, and real-time physiological status to focus on reviewing the scientific validity and safety of the intensity adjustment, determining whether there are risks such as muscle overload or excessive joint pressure. Once the review is passed, the newly generated exercise prescription takes effect. If the review identifies potential risks, the expert system outputs intensity adjustment suggestions, which are fed back to the model for secondary optimization, ensuring that the exercise prescription both meets individual needs and complies with medical-grade safety standards.

[0095] Step S207: Obtain the three-dimensional motion trajectory of multiple key joints of the user in the current motion phase.

[0096] Specifically, the key joint points are bony landmarks of human joints, including four hip joints, four knee joints, four ankle joints, and two toe joints. The joint angle recognition model uses three high-speed cameras to monitor the user in real time at fixed angles. After collecting raw joint movement data, it is processed by professional sports biomechanics analysis software to generate three-dimensional motion trajectories for each key joint point.

[0097] Step S208: Based on the standard motion library, determine the deviation of the three-dimensional motion trajectory of each key joint.

[0098] Specifically, the standard motion library includes standardized motion trajectory templates for different movements. The three-dimensional motion trajectory of each key joint is compared with the corresponding standardized motion trajectory template in the standard motion library. The Euclidean distance deviation of the trajectory coordinates is calculated using a dynamic time warping algorithm, and the trajectory deviation of each joint is finally output, which intuitively reflects the degree of fit between the user's actual movement and the standard movement.

[0099] Step S209: When the deviation is greater than the second preset threshold, generate a posture correction suggestion.

[0100] Specifically, when the trajectory deviation of any key joint point exceeds a second preset threshold (e.g., 5°), it is determined that the user's current movement has a problem with standardization, and targeted posture correction suggestions are generated. For example, if the knee joint deviates inward by 6°, it is recommended to adjust the distance between the feet to 1.2 times the shoulder width, keep the knee and toes aligned, and at the same time trigger real-time voice reminders to help the user correct the movement in time and reduce the risk of sports injuries caused by non-standard posture.

[0101] Optionally, user data is continuously acquired and monitored in real time. An emergency response is initiated when three consecutive heart rate recovery rates are found to be less than 12 bpm / min or the correlation coefficient between the RPE index and HRmax is less than 0.7. The RPE (Rating of Perceived Exertion) scale is a subjective assessment of exercise intensity or fatigue levels. The RPE index is the quantified version of this scale. HRmax refers to the peak heart rate during exercise, and the HRmax correlation coefficient is a statistical measure of the strength / direction of the quantitative relationship between other variables (such as age and blood pressure) and HRmax.

[0102] In some alternative implementations, Figure 2 This is a flowchart of another exercise prescription generation method according to an embodiment of the present invention, such as... Figure 2As shown, user data is acquired and subjected to a security screening. If the screening fails, the process ends; otherwise, a large language model is used to generate an exercise prescription and a predicted adaptation index. The user then exercises according to the prescription, and relevant data on the user at the current exercise stage is collected. If the user reaches the preset exercise goal, the process ends; otherwise, a three-dimensional ability assessment is performed to obtain a comprehensive muscle strength index. The user's actual adaptation index at the current exercise stage is further calculated, and exercise intensity constraint parameters are determined. Based on the exercise prescription, the aforementioned indices, and parameters, the large language model is optimized. The optimized large language model is used to regenerate the exercise prescription and predicted adaptation index, and the intensity deviation between the actual exercise intensity and the intensity of the regenerated exercise prescription is calculated. If the intensity deviation exceeds a first preset threshold, an expert system is used for security review, and the large language model is optimized a second time. If the intensity deviation does not exceed the first preset threshold, the user exercises according to the regenerated exercise prescription, and the subsequent process continues until the user reaches the preset exercise goal.

[0103] This invention acquires user data and uses a large language model to generate an initial exercise prescription and a predicted adaptation index. Users then follow this prescription, and relevant data on the user's current exercise stage is obtained. If the user fails to reach the preset exercise goal, a three-dimensional ability assessment is performed to obtain a comprehensive muscle strength index, quantifying the user's exercise ability at the current stage and further quantifying their adaptation level to obtain an actual adaptation index. Based on this, exercise intensity constraint parameters are determined, providing a precise basis for subsequent model optimization and prescription adjustment, ensuring that the exercise intensity always matches the user's real-time adaptation level, balancing ability improvement with the risk of overtraining. The large language model is optimized based on the exercise prescription, the aforementioned indices, and parameters. The optimized large language model is then used to generate another exercise prescription and predicted adaptation index, ensuring that the regenerated exercise prescription better matches the user's real-time ability and adaptation status, solving the problems of rigid prescriptions and lagging adjustments in traditional methods. Users then exercise according to the new exercise prescription and execute subsequent processes until the preset exercise goal is reached. A closed-loop feedback mechanism allows the model to continuously absorb individual user exercise data, ensuring that the exercise prescription aligns with individual exercise patterns and improving the accuracy and scientific validity of the exercise prescription.

[0104] This embodiment also provides an exercise prescription generation device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0105] This embodiment provides a device for generating exercise prescriptions, such as... Figure 3 As shown, it includes: The first generation module 301 is used to acquire user data and generate user exercise prescriptions and predictive adaptation indices based on user data using a large language model. The user data includes preset exercise goals.

[0106] The first acquisition module 302 is used to enable the user to exercise according to the exercise prescription and acquire relevant data of the user in the current exercise stage.

[0107] The assessment module 303 is used to conduct a three-dimensional ability assessment of the user based on user data and related data when the user completes the current exercise phase but does not reach the preset exercise goal, and obtain a comprehensive muscle strength index.

[0108] The quantification module 304 is used to quantify the user's degree of adaptation in the current exercise stage based on relevant data and comprehensive muscle strength index, and obtain the actual adaptation index.

[0109] The first determining module 305 is used to determine the motion intensity constraint parameters based on the actual adaptation index.

[0110] The second generation module 306 is used to optimize the large language model based on the exercise prescription, predicted adaptation index, actual adaptation index and exercise intensity constraint parameters, regenerate the user's exercise prescription and predicted adaptation index using the optimized large language model, return to the step of making the user exercise according to the exercise prescription, obtain the user's relevant data at the current exercise stage, until the user reaches the preset exercise goal.

[0111] In some optional implementations, user data includes peak concentric and peak eccentric forces from lower limb motor ability tests, with related data including the user's concentric and eccentric forces during the current phase of the exercise. Evaluation module 303 includes: The first calculation unit is used to calculate force capability based on the peak force of centripetal motion and the peak force of centrifugal motion.

[0112] The second calculation unit is used to calculate stability and balance based on the centripetal and centrifugal forces of the user during the current phase of motion.

[0113] The first determining unit is used to determine the comprehensive muscle strength index based on strength, stability, and balance.

[0114] In some alternative implementations, the relevant data may also include motion accuracy scores, heart rate variability scores, subjective fatigue scores, mental state scores, number of completed exercises, and target number of exercises. Quantization module 304 includes: The third calculation unit is used to calculate the score of the current exercise performance based on the comprehensive muscle strength index and the movement accuracy score.

[0115] The fourth calculation unit is used to obtain the user's performance score in the previous exercise phase, and calculate the performance improvement rate based on the user's performance scores in the current exercise phase and the previous exercise phase.

[0116] The fifth calculation unit is used to calculate the recovery quality index based on heart rate variability score, subjective fatigue level, and psychological state score.

[0117] The second determining unit is used to determine the ratio of the number of completed movements to the target number of movements as the movement consistency index.

[0118] The third determining unit is used to determine the actual adaptation index based on the improvement rate of athletic performance, the recovery quality index, and the athletic consistency index.

[0119] In some alternative implementations, the second generation module 306 includes: The fourth determining unit is used to determine the actual exercise intensity based on the exercise intensity and exercise intensity constraint parameters of the exercise prescription.

[0120] The sixth calculation unit is used to calculate the intensity matching loss based on the exercise intensity of the exercise prescription and the actual exercise intensity.

[0121] The seventh calculation unit is used to calculate the adaptation loss based on the predicted adaptation index and the actual adaptation index.

[0122] The eighth calculation unit is used to calculate safety loss based on the exercise intensity of the exercise prescription and the preset safety threshold.

[0123] The ninth computational unit is used to calculate the training loss based on the strength matching loss, the adaptive loss, and the safety loss.

[0124] The optimization unit is used to optimize large language models based on training loss.

[0125] In some alternative embodiments, after the second generation module 306, the apparatus further includes: The fifth determining unit is used to determine the intensity deviation between the actual exercise intensity and the exercise intensity of the regenerated exercise prescription.

[0126] The review unit is used to conduct a safety review of the regenerated exercise prescription using an expert system when the intensity deviation exceeds a first preset threshold.

[0127] In some alternative embodiments, the device further includes: The second acquisition module is used to acquire the three-dimensional motion trajectory of multiple key joints of the user in the current motion phase.

[0128] The second determination module is used to determine the deviation of the three-dimensional motion trajectory of each key joint based on the standard motion library.

[0129] The third generation module is used to generate posture correction suggestions when the deviation is greater than the second preset threshold.

[0130] The exercise prescription generation device provided in this embodiment of the invention can execute the exercise prescription generation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0131] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0132] The following is a detailed reference. Figure 4 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0133] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0134] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the exercise prescription generation method of the embodiments of the present invention.

[0135] Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0136] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the exercise prescription generation method shown in the above embodiments is implemented.

[0137] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0138] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method of exercise prescription generation, characterized by, The method includes: Acquire user data, and use a large language model to generate exercise prescriptions and predictive adaptation indices for the user based on the user data, wherein the user data includes preset exercise goals; The user is instructed to exercise according to the exercise prescription, and relevant data of the user at the current exercise stage is obtained. When the user completes the current exercise phase but fails to reach the preset exercise goal, a three-dimensional ability assessment is performed on the user based on the user data and the relevant data to obtain a comprehensive muscle strength index. Based on the relevant data and the comprehensive muscle strength index, the user's degree of adaptation in the current exercise phase is quantified to obtain the actual adaptation index; The motion intensity constraint parameters are determined based on the actual adaptation index. Based on the exercise prescription, the predicted adaptation index, the actual adaptation index, and the exercise intensity constraint parameters, the large language model is optimized. The optimized large language model is then used to regenerate the user's exercise prescription and predicted adaptation index. The process returns to the steps of having the user exercise according to the exercise prescription and obtaining relevant data of the user at the current exercise stage, until the user reaches the preset exercise goal.

2. The method of claim 1, wherein, The user data includes the peak concentric and eccentric strength of the lower limb motor ability test, and the relevant data includes the user's concentric and eccentric strength in the current exercise phase. The process of performing a three-dimensional ability assessment on the user based on the user data and the relevant data to obtain a comprehensive muscle strength index includes: Calculate the force capability based on the peak force of the centripetal motion and the peak force of the centrifugal motion; Based on the centripetal and centrifugal forces of the user in the current phase of motion, calculate the stability and balance capabilities. The comprehensive muscle strength index is determined based on the strength, stability, and balance abilities.

3. The method of claim 1, wherein, The relevant data also includes motion accuracy score, heart rate variability score, subjective fatigue level, psychological state score, number of completed exercises, and target number of exercises; The process of quantifying the user's adaptation level in the current exercise phase based on the relevant data and the comprehensive muscle strength index to obtain the actual adaptation index includes: Based on the comprehensive muscle strength index and the movement accuracy score, calculate the score of the movement performance in the current movement phase; Obtain the user's performance score in the previous exercise phase, and calculate the performance improvement rate based on the user's performance scores in the current exercise phase and the previous exercise phase. Based on the heart rate variability score, the subjective fatigue level, and the psychological state score, the recovery quality index is calculated. The ratio of the number of completed movements to the target number of movements is determined as the movement consistency index; The actual adaptation index is determined based on the improvement rate of athletic performance, the recovery quality index, and the athletic consistency index.

4. The method of claim 1, wherein, The optimization of the large language model based on the exercise prescription, the predicted adaptation index, the actual adaptation index, and the exercise intensity constraint parameters includes: The actual exercise intensity is determined based on the exercise intensity of the exercise prescription and the exercise intensity constraint parameters. Calculate the intensity matching loss based on the exercise intensity of the exercise prescription and the actual exercise intensity; The adaptation loss is calculated based on the predicted adaptation index and the actual adaptation index. Based on the exercise intensity and preset safety threshold of the exercise prescription, calculate the safety loss; The training loss is calculated based on the strength matching loss, the adaptive loss, and the security loss. The large language model is optimized based on the training loss.

5. The method of claim 4, wherein, After regenerating the user's exercise prescription and predicted adaptation index using the optimized large language model, the method further includes: Determine the intensity deviation between the actual exercise intensity and the exercise intensity of the regenerated exercise prescription; When the intensity deviation exceeds a first preset threshold, an expert system is used to conduct a safety review of the regenerated exercise prescription.

6. The method of claim 1, wherein, The method further includes: Obtain the three-dimensional motion trajectory of multiple key joints of the user in the current motion phase; Based on the standard motion library, the deviation of the three-dimensional motion trajectory of each key joint is determined; When the deviation exceeds a second preset threshold, a posture correction suggestion is generated.

7. A device for generating exercise prescriptions, characterized in that, The device includes: The first generation module is used to acquire user data and generate the user's exercise prescription and predictive adaptation index based on the user data using a large language model. The user data includes preset exercise goals. The first acquisition module is used to enable the user to exercise according to the exercise prescription and acquire relevant data of the user in the current exercise stage; The assessment module is used to perform a three-dimensional ability assessment on the user based on the user data and the relevant data when the user completes the current exercise phase but does not reach the preset exercise goal, and obtain a comprehensive muscle strength index. The quantification module is used to quantify the user's degree of adaptation in the current exercise phase based on the relevant data and the comprehensive muscle strength index, and obtain the actual adaptation index. The first determining module is used to determine the motion intensity constraint parameters based on the actual adaptation index; The second generation module is used to optimize the large language model based on the exercise prescription, the predicted adaptation index, the actual adaptation index, and the exercise intensity constraint parameters, regenerate the user's exercise prescription and predicted adaptation index using the optimized large language model, return to the steps of making the user exercise according to the exercise prescription, and obtain relevant data of the user in the current exercise stage, until the user reaches the preset exercise goal.

8. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the exercise prescription generation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the exercise prescription generation method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to perform the exercise prescription generation method according to any one of claims 1 to 6.

Images