Construction method and system of sports information intelligent service platform

Abstract

The invention relates to the technical field of information platform construction, in particular to a construction method and system of a sports information intelligent service platform. The method comprises the following steps: collecting user motion image data and user health basic data; key body node position information in the movement process of the user is extracted; generating a user motion track curve based on the key node position information, and calculating a deviation angle and a deviation distance with a standard track template to obtain motion correctness data; constructing a user fitness risk assessment model according to the action correctness data and the user health basic data, analyzing the difference degree with a preset safety threshold, and generating layered health guidance data; and decomposing the layered health guidance data into quantifiable progress units, and generating progress tracking data. According to the method, the user motion image data and the health basic data are combined, the evaluation model capable of identifying individualized risk factors is constructed, and more accurate safety guidance is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of information platform construction, and in particular to a method and system for constructing a sports information intelligent service platform. Background Art

[0002] The Sports Information Intelligent Service Platform is a comprehensive system based on big data, artificial intelligence, the Internet of Things, and cloud computing technologies. It aims to integrate and analyze sports-related data from multiple sources (such as athlete physiological data, event information, venue management data, and user behavior data) and provide personalized services, real-time analysis, and predictive support through intelligent algorithms and models. The platform collects athlete physical data and training performance through IoT devices, combines computer vision technology with tactical and movement analysis of game videos, and utilizes cloud computing to store and calculate massive amounts of data, enabling real-time live broadcasting of events, tactical optimization, audience services, and venue management.

[0003] However, traditional sports information intelligent service platforms often suffer from the following issues: Traditional systems rely primarily on subjective coaching judgment or simple sensor data for movement assessment, lacking the ability to accurately capture and analyze subtle deviations in user movements during exercise. Traditional fitness guidance systems often set vague, long-term goals for improvement, lacking quantifiable and trackable short-term progress indicators, making it difficult for users to perceive gradual improvement. Summary of the Invention

[0004] Based on this, it is necessary for the present invention to provide a method and system for constructing a sports information intelligent service platform to solve at least one of the above technical problems.

[0005] To achieve the above objectives, a method for constructing a sports information intelligent service platform includes the following steps:

[0006] Step S1: Collect user motion image data and user basic health data; extract key body node position information during the user's motion process; generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain movement correctness data;

[0007] Step S2: Build a user fitness risk assessment model based on the action correctness data and the user's basic health data, analyze the difference from the preset safety threshold, and generate layered health guidance data;

[0008] Step S3: Decomposing the layered health guidance data into quantifiable progress units to generate progress tracking data; performing anonymous user matching based on similar fitness characteristics and progress patterns on the progress tracking data to form interactive incentive data; and establishing an initial user movement pattern map based on the movement correctness data, the layered health guidance data, and the progress tracking data.

[0009] Step S4: Calculate the collection environment stability index and user feedback consistency index of each data point of the initial user motion pattern map and screen high-confidence data points to build a user motion evaluation model; analyze the operation fluency and usage frequency of each functional module based on the user motion evaluation model and interactive incentive data, and implement a low-threshold progressive function unlocking mechanism in combination with the interactive incentive data.

[0010] The present invention also includes a system for constructing a sports information intelligent service platform, which is used to execute the above-mentioned method for constructing a sports information intelligent service platform. The system for constructing a sports information intelligent service platform includes:

[0011] The motion trajectory analysis module is used to collect user motion image data and basic health data; extract the position information of key body nodes during the user's movement; generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain the movement correctness data;

[0012] The fitness risk assessment module is used to build a user fitness risk assessment model based on movement correctness data and user health basic data, analyze the difference from the preset safety threshold, and generate stratified health guidance data;

[0013] The progress tracking and interactive matching module is used to decompose the layered health guidance data into quantifiable progress units to generate progress tracking data; anonymously match the progress tracking data with users based on similar fitness characteristics and progress patterns to form interactive incentive data; and establish an initial user movement pattern map based on movement correctness data, layered health guidance data, and progress tracking data.

[0014] The user motion evaluation module is used to collect environmental stability index and user feedback consistency index of each data point of the initial user motion pattern map, and screen high-confidence data points to build a user motion evaluation model; based on the user motion evaluation model and interactive incentive data, the operation fluency and usage frequency of each functional module are analyzed, and a low-threshold progressive function unlocking mechanism is implemented in combination with the interactive incentive data.

[0015] Focusing on the core application requirements of a sports information intelligent service platform, this invention constructs a closed-loop mechanism for the entire process, from multi-source data collection to user behavior analysis, and then to capability assessment and functional guidance. This achieves a refined understanding and dynamic adaptation of users' movement status, behavioral habits, and learning abilities, significantly enhancing the platform's comprehensive capabilities in personalized services, data reliability assurance, and user experience optimization. In the data processing phase, by integrating movement correctness data, layered health guidance data, and progress tracking data, a unified, time-aligned multi-source data integration table is constructed. Collection quality is graded and managed using environmental parameters such as light intensity, background complexity, and device stability, ensuring the stability and reliability of the data foundation for subsequent model training and analysis tasks. Furthermore, the user technical feature profiles and learning characteristic data established by the platform not only reveal typical deviations in user movement execution and the speed at which new knowledge is absorbed, but also reflect the underlying dependency patterns between technical behaviors through a feature association network. This supports the establishment of a sports ability development model and generates short- to long-term ability growth predictions, providing a scientific basis for the development of personalized training paths. In terms of its evaluation mechanism, the platform assesses the authenticity and representativeness of data from two dimensions: environmental interference factors and user feedback performance. This generates a collection environment stability index and a user feedback consistency index. A weighted mechanism forms a comprehensive data confidence evaluation model, enabling the platform to automatically identify high-confidence data points during data screening and modeling, thereby building a more accurate and robust campaign evaluation model. After model training, the platform further mines behavioral trajectory information generated during user interaction with the system. By recording parameters such as operation duration, access frequency, and page paths, it identifies peak usage times and changing trends in user function preferences. Combined with the user's performance profile, it dynamically constructs a mapping between user function requirements and provides a behavioral basis for feature recommendations. Furthermore, by analyzing user operation paths, decision-making node preferences, and interactive response characteristics, the platform accurately identifies their exploration behavior patterns and learning paths, enabling proactive and personalized adjustments to feature adaptation and content push, effectively enhancing user engagement and platform stickiness. The platform also measures user proficiency and mastery of each module by calculating operation completion time, retry frequency, and help information access, allowing for dynamic adjustments to user function access permissions. In terms of function guidance, a low-threshold progressive unlocking strategy is introduced. After users gradually master basic functions, they will automatically receive prompts to unlock new functions based on their performance. This well-paced advancement mechanism lowers the learning threshold for function usage and prevents barriers to use caused by information overload or complex operations. Furthermore, the system dynamically adjusts the scoring weights and time limits in the unlocking mechanism based on user incentive response data, such as point achievement rate and duration of task completion. This makes the unlocking process more flexible and adaptable, maintaining a sense of challenge for users while enhancing their motivation for continuous learning.Through the systematic implementation of these methods, the platform not only achieves a multi-dimensional understanding of individual users' athletic behavior, learning rhythm, and operational preferences, but also precisely constructs training recommendations, function push paths, and an evaluation system tailored to their specific characteristics, enhancing the intelligence and usability of the platform's services. This data-driven approach, guided by user behavior, seamlessly integrates technical models with human-computer interaction, facilitating the platform's shift from passive recording to active guidance. This approach demonstrates significant practicality and potential for widespread adoption in the context of intelligent sports services. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0017] Figure 1 A schematic flow chart of the steps of a method for constructing a sports information intelligent service platform according to the present invention;

[0018] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0019] Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0020] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0021] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0022] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0023] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for constructing a sports information intelligent service platform, the method comprising the following steps:

[0024] Step S1: Collect user motion image data and user basic health data; extract key body node position information during the user's motion process; generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain movement correctness data;

[0025] In the sports information intelligent service platform, the embodiment of the present invention first collects the motion image data of the user when performing preset fitness movements (such as squats, push-ups, lunges, etc.) through a multi-angle camera system deployed in a gym or home scene, and simultaneously obtains the user's basic health data, including weight, height, BMI index, resting heart rate value, previous sports injury records, etc. These health data can be obtained through smart bracelets, smart body fat scales, and filling out historical health questionnaires. After collecting the motion image data, the platform extracts the three-dimensional coordinate information of the user's key body nodes through a posture recognition algorithm (such as OpenPose or MediaPipe), and the key nodes include the head, shoulders, elbows, wrists, hip joints, knees and ankles. Next, the user's complete motion trajectory curve is generated through a time series tracking algorithm, and then compared frame by frame with the standard motion trajectory template. The angle deviation between each key node and the standard node (for example, the difference between the user's knee joint bending angle and the standard angle) and spatial displacement (that is, the Euclidean distance of the user node relative to the standard trajectory node) are calculated to obtain the motion correctness data. The data is recorded with a sampling time of 0.1 seconds per frame, and the generated data fields include "node number, time point, angle deviation (degrees), displacement distance (cm)".

[0026] Step S2: Build a user fitness risk assessment model based on the action correctness data and the user's basic health data, analyze the difference from the preset safety threshold, and generate layered health guidance data;

[0027] After completing the extraction of the user's action correctness data, the embodiment of the present invention constructs a personalized fitness risk assessment model based on the user's basic health data (such as joint flexibility index, muscle strength level, etc.). The model uses a regression analysis method based on gradient boosting tree (GBDT) to map variables such as angle deviation and displacement distance with the risk level that may lead to sports injury or action error, and outputs a risk score for each type of action. For example, if a user's knee abduction angle deviation is continuously higher than 15 degrees and the corresponding BMI is greater than 28, the system assesses that the knee pressure exceeds the standard when performing squats, and the injury risk exceeds the safety threshold. The platform sets this score as "high risk level". Based on the risk assessment results and the platform's preset risk stratification standards (such as the safety threshold is set to level 3 out of 5), the platform generates stratified health guidance data, which includes corrective suggestions for user actions (such as "knee abduction control" and "core strength training"), auxiliary equipment suggestions (such as using elastic bands to control trajectory) and training limit intensity levels, ensuring that the training suggestions generated by the platform are personalized, scientific and executable.

[0028] Step S3: Decomposing the layered health guidance data into quantifiable progress units to generate progress tracking data; performing anonymous user matching based on similar fitness characteristics and progress patterns on the progress tracking data to form interactive incentive data; and establishing an initial user movement pattern map based on the movement correctness data, the layered health guidance data, and the progress tracking data.

[0029] After obtaining the layered health guidance data, the embodiment of the present invention will break down the guidance suggestions into quantifiable progress units, that is, quantify the fitness goals according to the key points of action improvement and the degree of achievability. For example, if the user's knee abduction control problem is more serious, the problem will be broken down into progress units such as "5 minutes of daily training of half squats with elastic bands" and "target abduction angle controlled within 10 degrees". Each unit has a time requirement, action standard and target indicator. The platform generates progress tracking data by periodically evaluating the user's execution effect. The data is composed of fields such as "progress unit number, completion status, trajectory improvement degree, training frequency" and so on. On this basis, the platform uses clustering algorithms (such as K-Means) to analyze the fitness characteristics (such as physical conditions, action defect types, improvement paths) and progress patterns (such as improvement speed, training frequency, etc.) of different users to achieve anonymous user matching and form interactive incentive data, such as pushing "user challenge list with similar progress speed as you" or forming a virtual training group of users with similar problems. At the same time, the platform semantically annotates and maps the user's movement correctness data, health guidance data and progress tracking data to a graph structure to construct an initial user movement pattern map. The graph nodes represent the execution status of various actions, and the edges represent the conversion paths between actions or the user state transfer logic, supporting subsequent personalized training route recommendations.

[0030] Step S4: Calculate the collection environment stability index and user feedback consistency index of each data point of the initial user motion pattern map and screen high-confidence data points to build a user motion evaluation model; analyze the operation fluency and usage frequency of each functional module based on the user motion evaluation model and interactive incentive data, and implement a low-threshold progressive function unlocking mechanism in combination with the interactive incentive data.

[0031] After completing the construction of the initial user motion pattern map, the embodiment of the present invention, the platform screens each data node in the map for the collection environment stability index and the user feedback consistency index. The collection environment stability index is used to evaluate whether the data node is collected under the same environmental lighting, shooting angle, background interference and other conditions. The calculation basis is the standard deviation of the brightness and background disturbance indicators in each video frame image. If the standard deviation exceeds the threshold set by the platform (such as the brightness standard deviation is greater than 15 pixel value units), the credibility of the node data is reduced; the user feedback consistency index is obtained by calculating the correlation between the training satisfaction score submitted by the user after each training and the actual improvement in the action. If the consistency is high (such as the Pearson correlation coefficient is greater than 0.7), the node feedback is considered credible. The platform only retains data nodes with high dual indicators and builds a high-confidence user motion evaluation model, which is used to accurately evaluate the user's current motion state and predict its future action execution risk or ability change trend. Based on this evaluation model and the aforementioned interactive incentive data, the platform further analyzes the operational fluency of each functional module (such as action recognition response time, feedback generation time) and user usage frequency (such as the number of weeks a certain training suggestion is called). If a functional module has low incentive participation, low user usage frequency, and the evaluation model predicts that the module has little impact on user progress, the advanced version will not be opened for the time being; on the contrary, for modules with good incentive effects and strong user stickiness, a low-threshold progressive function unlocking mechanism will be implemented. For example, after the user completes three consecutive action accuracy improvements and meets the standards, the advanced challenge action training video will be automatically unlocked to enhance the platform's user activity and positive training feedback.

[0032] Preferably, step S1 includes the following steps:

[0033] Step S11: Capturing multiple frames of continuous images of a user performing fitness movements through a camera of a smart terminal device to generate user motion image data;

[0034] In the sports information intelligent service platform of the embodiment of the present invention, users use smart terminal devices equipped with high-resolution cameras (such as smart fitness mirrors, camera TVs or mobile phones with AI recognition functions) to perform fitness training. After the user selects a specified action (such as bent-over rowing, squats, and lunges), the platform starts the camera module and continuously shoots the user's complete training process at a frame rate of 15 frames per second. The acquisition cycle is the time required to complete the action (for example, the complete process of completing a squat is about 5 seconds, so 75 frames of continuous image data are generated). These continuous image data constitute the original user motion image data. The platform numbers each frame of the image and records the timestamp to ensure that subsequent analysis can accurately locate the details of the action under time changes. In order to enhance the quality of the image data, the system automatically detects the adequacy of lighting and the integrity of the character subject (such as whether there is limb occlusion or overexposure in the image) to ensure that only high-quality action image sequences that meet the processing standards are collected as the input basis for subsequent posture recognition and trajectory analysis.

[0035] Step S12: Collecting the user's height, weight, age, and exercise habits to form basic health data of the user;

[0036] In an embodiment of the present invention, before capturing motion images, the platform prompts the user to fill in basic health information through the smart terminal interface and calls the wearable device to synchronize data. The user needs to manually or through health record authorization to fill in height (in centimeters), weight (in kilograms), age (in years) and exercise habit information (including weekly exercise frequency, exercise time, main training items, etc.). Exercise habit information is collected in the form of a structured questionnaire and converted into a quantifiable training level indicator. For example, "training 3 times a week, 30 minutes each time, preferring no equipment training" will be classified as "medium frequency, low intensity trainer". In addition, the platform supports automatic acquisition of user physiological parameters such as body fat percentage, resting heart rate, etc. through Bluetooth connection to smart body fat scales, heart rate bracelets and other devices to enrich the dimensions of health data. The system encapsulates this type of information into a "user health basic data" object, with fields including basic physical signs, exercise frequency level, and training preference tags, which serve as the basis for subsequent personalized action risk analysis and recommended training templates.

[0037] Step S13: performing image preprocessing on the user's motion image data and performing skeleton point recognition to locate the user's key body nodes of the shoulder, elbow, wrist, hip, knee, and ankle, and generate key body node position information;

[0038] After acquiring the user's motion image data, the embodiment of the present invention first performs image preprocessing on each frame of the image, which mainly includes image denoising, background cutout and human body area enhancement. Image denoising uses a Gaussian filter to smooth the image edges and eliminate the interference of image particles that may occur during the shooting process; background cutout uses a deep learning segmentation network (such as U^2-Net or BodyPix) to separate the user from the background, so that the recognition algorithm focuses on the human body structure; human body area enhancement enhances the clarity of bone lines through brightness normalization and contrast enhancement. Subsequently, the platform performs bone point recognition on the preprocessed image based on a lightweight posture estimation algorithm (such as BlazePose or HRNet), extracts and tracks the position of the user's key body nodes. Key nodes include the shoulder (connecting the upper arm and the torso), the elbow (the key point of forearm movement), the wrist (the control point of the hand), the hip (the core point affecting the overall stability), the knee (the main force support point), the ankle (the dynamic stability control point of the lower limb), etc. The position of each key node is expressed in three-dimensional spatial coordinates, including lateral displacement, longitudinal displacement and estimated depth value (in centimeters). The platform integrates the skeletal node position information at multiple times into an array of "key body node position information" to construct dynamic motion trajectories.

[0039] Step S14: Generate a user motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain action correctness data.

[0040] Based on the key node position information generated in step S13, the embodiment of the present invention utilizes a motion trajectory reconstruction algorithm to model the three-dimensional motion path of each key node of the user over time during the entire training process, forming a user motion trajectory curve. The specific processing flow is as follows: the system interpolates and smoothes the spatial position of each key node in consecutive frames in chronological order, connects and generates a continuous curved trajectory, and simultaneously comprehensively models the relative movement patterns between multiple nodes to ensure the accuracy of the overall coordination analysis of the movement. Next, the platform calls the corresponding action template in the standard trajectory template library. The standard trajectory data in the template is collected and modeled by professional athletes under ideal conditions and has high reference value. The platform compares the user trajectory with the standard trajectory point by point through frame synchronization and key point matching mechanism, calculates the deviation angle (such as the difference between the knee joint flexion angle and the standard angle) and offset distance (such as the perpendicular distance between the shoulder joint displacement direction and the standard curve) of the key nodes in each frame, and the deviation units are angle (degrees) and distance (centimeter). For example, if the user's hip joint deviation angle is greater than 15 degrees and the ankle offset distance exceeds 10 cm when performing a lunge squat, the system will determine that the posture is abnormal. Finally, the platform summarizes the above angle and distance data into "action correctness data" and marks each frame with an "action matching score" as the core basis for subsequent action risk analysis and guidance suggestions.

[0041] Preferably, step S14 includes the following steps:

[0042] Step S141: performing time series analysis on the position information of key body nodes, recording the position changes of each key node in continuous time frames, and generating raw data of node motion trajectory;

[0043] After obtaining the array of position information of the user's key body nodes, the embodiment of the present invention starts the timing analysis module to dynamically track the changes in the three-dimensional coordinates of each key node (such as the shoulder, elbow, hip, knee, etc.) in continuous time frames (for example, the aforementioned 15 frames per second, with a total of 75 frames sampled). The system constructs a time series trajectory sequence for each node with the frame number as the horizontal axis and the spatial coordinate as the vertical axis to record its movement trajectory over time. For example, for the left knee node, the platform records its three-dimensional position changes from the 1st frame to the 75th frame as a three-dimensional vector sequence to form the original node motion trajectory data. This process is calibrated by timestamps to ensure that all node data strictly corresponds to the same time step. At the same time, the system automatically fills in the missing position information caused by short-term occlusion to improve the continuity and integrity of the trajectory. Finally, the platform combines the time series position data of all nodes to generate "node motion trajectory raw data" as the basis for subsequent curve construction and smoothing.

[0044] Step S142: Smoothing the original data of the node motion trajectory to obtain user motion trajectory curve data;

[0045] The embodiment of the present invention applies a motion trajectory smoothing algorithm to the original data of the node motion trajectory to eliminate the trajectory noise caused by jitter, environmental interference or recognition error during the user's movement. The specific method is a moving average smoothing algorithm based on a sliding window, setting the window size to 5 frames, and performing weighted averaging processing on the X, Y, and Z three-dimensional coordinate values ​​of each node, so that the coordinate value of each frame takes the average value of its adjacent previous and next frames, so that the trajectory curve is more visually coherent and more mathematically stable. For joint turning points (such as the changing points of the hip and knee angles in a squat), the system introduces an edge-preserving filtering strategy (such as Savitzky-Golay filtering) to prevent the turning point features of the action from being weakened by smoothing. The processed user motion trajectory curve data is presented as a continuous, smooth three-dimensional curve set, which can more realistically reflect the user's movement trend and be used for subsequent dynamic analysis and trajectory comparison.

[0046] Step S143: Calculating the motion angle, velocity, and acceleration of the human motion segments formed between adjacent key nodes based on the user motion trajectory curve data, and constructing a dynamic parameter set of the user motion segments;

[0047] Based on smoothed user motion trajectory curve data, the present invention calculates the dynamic parameters of human motion segments formed between adjacent key nodes. A "motion segment" is a body structure formed by two adjacent key points (such as shoulder-elbow, hip-knee, and knee-ankle), which rotates and moves over time during movement. The platform first constructs a vector between each pair of nodes and records the change in its direction in consecutive frames. The "motion angle" is calculated by calculating the change in the angle between the vectors in adjacent frames. Simultaneously, the system calculates the change in coordinate displacement of each motion segment vector between time frames and divides it by the time interval to obtain the "linear velocity." The velocity change is then time-differentiated to obtain "acceleration" data. All these parameters are combined into a "user motion segment dynamic parameter set," which accurately describes the motion characteristics of each body structure during training. For example, during a bent-over row, whether the elbow-shoulder segment's motion angle reaches above 45 degrees or is accompanied by excessive swinging (excessive acceleration) can be reflected in this parameter set, enabling further movement recognition and precise feedback.

[0048] Step S144: matching the user's motion segment dynamic parameter set with the standard parameter set in the action library to determine the type of fitness action currently performed by the user and generate an action type recognition result;

[0049] In an embodiment of the present invention, the constructed dynamic parameter set of the user's motion segments is compared and matched with the standard parameter set in the platform's action standard library to identify the type of fitness action the user is performing. The platform pre-establishes standard parameter models for common fitness actions (such as squats, deadlifts, bent-over rows, etc.). Each model contains the angle range, speed rhythm, acceleration change characteristics, etc. of multiple typical motion segments. For example, the "standard squat" model defines that during the squat phase, the hip-knee segment must form an angle closure greater than 90 degrees within 1 second, and the ankle speed must not exceed 0.5 meters per second. The platform uses a feature vector similarity algorithm (such as cosine similarity or dynamic time warping) to match the user's dynamic parameter set with each standard model one by one to identify the closest action type. The system outputs an "action type recognition result", which includes the recognized action name, matching confidence score (between 0 and 1), and reference action number, providing a basis for subsequent standard trajectory extraction and alignment operations.

[0050] Step S145: extracting corresponding standard trajectory template data from the action standard library according to the action type recognition result, and performing spatiotemporal alignment between the user motion trajectory curve data and the standard trajectory template data to obtain normalized trajectory comparison data;

[0051] According to the embodiment of the present invention, based on the identified action type, the platform extracts the standard trajectory template data of the corresponding action from the action standard library. The data is composed of a three-dimensional trajectory curve generated by a professional athlete performing the action in a standard manner. The platform then performs a "time-space alignment" operation on the user trajectory curve data and the standard trajectory to resolve the differences between the two trajectories in terms of starting time, action rhythm, scale unit, etc. First, the user and the standard trajectory are automatically aligned on the time axis through the dynamic time warping (DTW) algorithm, so that the key nodes of the two actions from the start to the completion process are matched as much as possible; secondly, the platform normalizes the spatial scale by converting the two trajectories into a standard unit coordinate system with a percentage of height as the unit, eliminating the influence of the absolute coordinates of the trajectory caused by body shape differences, and obtaining "normalized trajectory comparison data" for standard deviation calculation and error analysis.

[0052] Step S146: Calculate the Euclidean distance deviation between the user trajectory and the standard trajectory at each key time point in the normalized trajectory comparison data to generate a point deviation sequence, and calculate the angular deviation between the user motion segment and the standard motion segment at each key time point to generate an angular deviation sequence;

[0053] The embodiment of the present invention is based on the normalized trajectory comparison data, and sequentially compares the user's motion trajectory with the standard motion trajectory at several "key time points". The so-called key time point refers to the moment in the action when the posture changes most significantly and has the greatest impact on the effect of the action (such as the lowest point of squatting, the turning point of standing up, etc.). The platform predefines the number and position of key time points for each action (for example, a complete squat contains 5 key points). The platform calculates the "Euclidean distance deviation" of the corresponding node at each key point, that is, the spatial difference distance between the user node position and the standard node position in the three-dimensional coordinate system (in centimeters); at the same time, it calculates the "angle deviation" of the same segment at this time point, that is, the angle difference between the user segment vector and the standard segment vector (in degrees). The system generates two sequences: one is a "point deviation sequence", which reflects the position error; the other is an "angle deviation sequence", which reflects the structural posture error. These two sequences are used to comprehensively evaluate the user's motion accuracy.

[0054] Step S147: Score the overall motion trajectory deviation based on the point deviation sequence and the angle deviation sequence to obtain motion correctness data.

[0055] The embodiment of the present invention inputs the point deviation sequence and the angle deviation sequence into the action scoring module, adopts the weighted scoring algorithm to score the deviation of the overall action, and generates "action correctness data". Among them, the platform sets the deviation sensitivity weight according to the action type. For example, it pays more attention to the hip and knee angle deviation for squat-type actions, and pays more attention to the shoulder and elbow angle and upper body posture deviation for rowing-type actions. The scoring algorithm assigns weights to the point deviation and angle deviation of each key time point respectively, and performs normalization calculations. For example, a deviation of less than 5 cm or 5 degrees can be recorded as full marks, less than 10 is passed, and more than 10 is unqualified. The platform finally outputs a "action correctness score" value (0 to 100 points), and marks the deviation concentration section and the corresponding joint to facilitate subsequent user guidance feedback and action correction suggestion generation. The whole process lays the foundation for the platform to provide intelligent training effect evaluation and safety monitoring capabilities.

[0056] Preferably, step S2 includes the following steps:

[0057] Step S21: extracting action deviation characteristic indicators based on the action correctness data to form action risk basic assessment data, wherein the action deviation characteristic indicators include the maximum deviation value, average deviation value and deviation duration of key nodes;

[0058] After generating movement correctness data, the platform system of this embodiment of the present invention further analyzes the data and extracts movement deviation characteristic indicators to form basic movement risk assessment data. This process first reads the point deviation sequence and angle deviation sequence in time series format, and statistically analyzes the deviation trajectory of each key body node within the complete movement cycle. Based on this, three core indicators are extracted: the maximum deviation value of the key node, which is used to measure the most significant deviation in the user's movement; the average deviation value, which reflects the prevalence of deviation throughout the entire movement process; and the deviation duration, which is the time period during which the deviation value exceeds the threshold, which is used to reflect the duration of the user's non-standard movement. For example, during a squat exercise, if the angle between the user's knee and hip joints consistently deviates from the standard trajectory by more than 10 degrees for more than 1.5 seconds, the system will record this deviation as a high-risk feature. These features are statistically analyzed by movement category and node type and stored in the movement risk basic database, providing basic input for subsequent risk modeling.

[0059] Step S22: extracting user physiological characteristic parameters from the user's basic health data to generate user physiological risk factor data;

[0060] The embodiment of the present invention obtains the user's basic health data through user-side upload or interface call. The data usually includes the user's age, gender, weight, height, previous sports injury history, bone health status (such as whether there is joint cartilage degeneration), heart rate change range, flexibility score and other information. The system extracts physiological characteristic parameters from it and generates user physiological risk factor data. During the processing, the platform uses a screening model based on the medical knowledge graph to extract features from health data, and automatically selects the corresponding high-correlation factor combination according to the type of action currently performed by the user. For example, for users who perform jumping actions, the system preferentially extracts knee joint stability parameters and ankle joint flexibility parameters as physiological risk factors. In addition, the platform introduces data standardization processing to normalize the risk factors of different users to between 0 and 1, which is convenient for weighted modeling in the subsequent risk calculation process.

[0061] Step S23: establishing a correlation model between the basic action risk assessment data and the user's physiological risk factor data, calculating the degree of injury risk to the user caused by various action deviations, and generating an individualized risk assessment value;

[0062] After obtaining the action deviation characteristic index and the user's physiological risk factor data, the embodiment of the present invention constructs a correlation model between the two to calculate the risk level of sports injuries caused by deviations when the user performs a certain type of action. The model adopts a risk inference model based on a combination of logistic regression and Bayesian networks. During the training phase, a large amount of historical sample data is used to mark the correspondence between action deviation and injury consequences; during the prediction phase, the current user's action deviation index and its physiological risk factor are used as input to output the individualized injury risk value of the action. For example, when the system detects that a user is doing a lunge squat, the deviation value of the inward angle of the knee joint is greater than 15 degrees and lasts for more than 2 seconds. At the same time, the user has a history of knee ligament injury. The platform uses model reasoning to deduce that the individualized risk value corresponding to the action is 0.82 (range of 0 to 1), indicating a high risk level.

[0063] Step S24: stratify the risk levels according to the individualized risk assessment value and the preset safety threshold to generate risk grading data;

[0064] The embodiment of the present invention performs risk level stratification based on the calculated individualized risk assessment value and the risk safety threshold set by the platform. Specifically, the system divides the risk assessment values ​​between 0 and 1 into several risk level segments, for example, 0.0 to 0.3 is low risk, 0.3 to 0.6 is medium risk, and 0.6 to 1.0 is high risk, and the assessment values ​​are assigned to the corresponding levels to form risk grading data. This data not only includes basic information such as action type, user ID, and action occurrence time, but also includes risk level labels, risk values, and main characteristic parameters that affect the risk assessment results (for example, which deviation indicator contributes the most to the risk). For example, the platform records the user's "squat" action assessment value as 0.72, and the system classifies it as "high risk" and marks the main source of the risk as "the average knee angle deviation value is high and lasts for a long time."

[0065] Step S25: performing differentiated processing on action deviations of different risk levels according to the risk classification data, and recording the action deviation patterns and guidance responses to obtain user historical feedback data;

[0066] After obtaining the risk grading data, the embodiment of the present invention performs differentiated processing on the movement deviation according to the risk level. The system can automatically ignore the deviation of the low-risk level or only prompt correction suggestions. The medium-risk level will prompt the user to pay attention and provide correction guidance. If it is a high-risk level, the system will immediately issue a forced pause training suggestion and start the virtual coach assisted teaching process. At the same time, it records the feedback processing strategy of each movement deviation of the user and the user's response behavior to the processing to form the user's historical feedback data. For example, when a user is identified as high-risk by the system for performing high-position leaning movements three times in a row, the system not only performs a voice prompt to pause, but also records whether the user has made adjustments after the prompt, whether the movement deviation has been alleviated after the adjustment, and other information, and stores the processing results in a structured manner.

[0067] Step S26: Visualize the health guidance content based on the risk classification data and the user's historical feedback data, thereby generating layered health guidance data.

[0068] The embodiment of the present invention integrates risk grading data and user historical feedback data to generate layered health guidance content and realize visual presentation. The platform adopts a visualization module based on a graphics engine to superimpose the deviation points in each action process on the user's human skeleton model in the form of a color heat map, and displays the corresponding risk level and recommended feedback action through the data panel. For example, if a user repeatedly encounters excessive hip swing problems during high-leg training, the platform will automatically generate a risk curve for the user under this action and superimpose text suggestions, such as "It is recommended to lower the leg height and strengthen the stability of the core muscles." At the same time, the platform classifies guidance strategies under different risk levels, provides daily training suggestions for low-risk users, pushes video correction training content to medium-risk users, and recommends professional rehabilitation courses to high-risk users, realizing personalized services for layered health guidance data and supporting the call of subsequent intelligent sports recommendation modules.

[0069] Preferably, step S25 includes the following steps:

[0070] Step S251: extracting motion segment angle deviation values, speed control deviation values, and motion stability indices from low-risk motion deviations in the risk classification data, performing quantitative comparisons with preset standard motion models, identifying optimizable parameters, and obtaining low-risk optimizable parameters;

[0071] The embodiment of the present invention further extracts optimizable motion parameters for movement deviations that have been categorized as low-risk. First, the system performs a structural deconstruction analysis on the raw motion data of this type of deviation, calculates the angular deviation values ​​between key joints segment by segment according to the motion chain (such as the shoulder-elbow-wrist chain, hip-knee-ankle chain), and derives the speed control deviation value in combination with time series analysis, that is, the degree of difference between the speed fluctuation of the user when completing a certain key movement segment and the speed curve of the standard model. At the same time, the platform also introduces movement stability index evaluation, including the smoothness of the movement trajectory (assessed by acceleration variance) and the center of gravity fluctuation range (measured by the displacement amplitude of the user's center point). The system compares the above three types of parameters with the ideal parameters in the standard movement model one by one, and uses threshold judgment and deviation sorting methods to identify movement chains and indicators that still have room for improvement within the low-risk range. For example, if the user's back contraction speed is fast during rowing training, deviating from the standard model mean by 8%, but does not cause obvious instability, the system will determine that the speed control index is a "low-risk optimizable parameter."

[0072] Step S252: Formulate a progressive adjustment plan for low-risk optimizable parameters to form general optimization suggestion data;

[0073] After identifying low-risk optimizable parameters, the embodiment of the present invention automatically generates a progressive optimization plan on the platform to form general optimization recommendation data. This process is based on the principle of "small incremental adjustments" in motor learning theory, and plans the adjustment range in combination with the user's existing ability level. For example, for users with large speed control deviations, the platform sets an optimization plan with a daily target value improvement of no more than 5%, to avoid new deviations caused by excessive corrections. At the same time, the system provides action demonstration videos and real-time voice prompts corresponding to the suggestions, such as the prompt "Slow down the back stretching speed by 0.2 seconds in the next set of training to keep the muscle group tension stable." The optimization suggestion data is stored as a structured record, which includes the optimization target parameters, adjustment range, operation frequency and adaptive training action list, and can be presented in the form of pictures and texts in the user interface.

[0074] Step S253: analyzing the proximity between the movement deviation point and the human biomechanical safety limit for the movement deviation of the medium risk level in the risk classification data, and calculating the potential stress concentration area;

[0075] In the embodiment of the present invention, for movement deviations of medium risk level, the platform further analyzes the degree of proximity between the movement deviation point and the human biomechanical safety limit. First, the system calls the built-in human safety limit model, which combines the range of motion (ROM), tendon load limit and maximum safety value of bone force defined in the biomechanical research literature to map the nodes of the user's deviation movement. Subsequently, the inverse dynamics algorithm is used to estimate the force path and load distribution of each key node in the deviation state, and identify possible "potential stress concentration areas". For example, in the bent-over press, if the shoulder flexion angle exceeds 90% of the physiological limit and the humeral internal rotation angle deviates significantly from the standard value, the system will determine that there is potential stress concentration in the acromion area and mark it as a medium-risk warning point.

[0076] Step S254: Based on the potential stress concentration areas and the user's physiological risk factor data, the risk of injury to the user's joints, muscles, and ligaments caused by the deviation is assessed, and basic safety guidance data is generated, including a description of prohibited actions, a definition of safe action ranges, and recommendations for alternative actions.

[0077] After identifying potential stress concentration areas, the embodiment of the present invention uses the platform to further evaluate the possibility of the movement deviation causing damage to the user's joints, muscles and ligaments in combination with the user's physiological risk factor data. The system uses mechanical simulation and individual parameter embedding models for cross-evaluation. If a user has a background of articular cartilage degeneration and the load on the shoulder rotation muscle group is continuously high in a certain deviation movement, the system will label the deviation as a "high stress-physiological fragility" combined risk and automatically generate basic safety guidance data containing three parts: the first is "determining prohibited action descriptions", which clearly prompts users to avoid certain specific actions; the second is "defining the range of safe actions", which clearly shows the user's acceptable joint movement range through diagrams; the third is "recommendation of alternative actions", which recommends to users actions with similar functional goals but lower biomechanical burdens, such as replacing high-load push-ups with elastic band lateral raises. The above information is integrated to generate personalized guidance documents and can be synchronized to the user's mobile terminal.

[0078] Step S255: extracting the degree and duration of the high-risk action deviations exceeding the safety threshold from the risk classification data, and performing error classification based on the high-risk action deviations to obtain error type data, wherein the error type data includes technical errors, physical limitation errors, or potential health problem errors;

[0079] In this embodiment of the present invention, for high-risk movement deviations, the platform needs to further extract two indicators: the "degree to which the deviation exceeds the safety threshold" and the "duration." The system first locates the deviation mutation point based on the time-series data of the user's entire movement process, and calculates the difference between the peak deviation value and the preset safety threshold. For example, if the hip adduction angle deviates from the standard posture by more than 25 degrees and lasts for more than 3 seconds, it exceeds the upper threshold by more than 50%. The system then calls a well-trained error recognition model (integrating expert rules and supervised learning methods) to classify and identify the high-risk movement deviation, and determines whether the error type belongs to one of the following three categories: one is "technical error", such as movement route control error and muscle group dyssynergy; the second is "physical limitation error", such as insufficient flexibility and muscle strength that make the movement unable to be completed; and the third is "potential health problem error", such as repeated deviations accompanied by abnormal physiological reactions (such as abnormal heart rate fluctuations), indicating that the user may have unidentified health problems. The system outputs error type data for subsequent personalized intervention.

[0080] Step S256: Generate professional consultation guidance data according to the error type data;

[0081] The embodiment of the present invention generates professional consultation guidance data based on the aforementioned error type data. For "technical errors", the system pushes a special corrective training plan designed by a sports coach, and guides the user to make an appointment for an online video guidance course; for "physical limitation errors", the platform pushes a low-intensity training plan based on physical fitness improvement, and provides training progress management tools; and for "potential health problem errors", the platform will forcibly trigger a health warning mechanism, recommending users to suspend related training, push a consultation appointment portal for sports rehabilitation experts or sports medicine physicians, and guide users to go to designated physical examination institutions for special assessments when necessary. The above-mentioned professional consultation guidance data is stored in the form of a structured data packet, which includes an error type tag, intervention suggestion type, recommended resource link and tracking feedback interface, and supports cross-module interactive use.

[0082] Step S257: Record the user's behavioral response after receiving general optimization suggestion data, basic safety guidance data or professional consultation guidance data, including the time delay in receiving guidance, the completion degree of executing guidance and the persistence degree of continuously applying guidance, and generate user historical feedback data.

[0083] The embodiment of the present invention records and models the entire process of the user's behavioral response after receiving various types of guidance data (including general optimization suggestion data, basic safety guidance data or professional consultation guidance data). The platform obtains the "time delay in receiving guidance" (i.e. the time it takes for the user to respond for the first time after receiving the prompt), "completion of execution of guidance" (through comparative analysis of subsequent action data and target guidance parameters to determine whether the adjustment is in place) and "persistence in continuous application of guidance" (statistics of whether the user continues to complete training according to the guidance within a certain period) through the user behavior monitoring interface. For example, a user starts to modify his movements within 3 hours after receiving shoulder joint safety guidance data. After the modification, the shoulder angle deviation drops by more than 15%, and the new movement posture is maintained for three consecutive days. Based on this, the platform scores the user as a "high response-high completion-high persistence" type, generates user historical feedback data and uploads it to the user behavior modeling module for subsequent personalized content recommendation and risk control strategy adjustment.

[0084] It is particularly important that step S26 includes the following steps:

[0085] Analyze historical user feedback data to determine the most effective information delivery channel by analyzing the user's response speed and execution completion rate to different presentation formats;

[0086] Evaluate users' understanding and preference for professional terminology based on historical user feedback data, and establish a user terminology acceptance model;

[0087] Based on the terminology acceptance model, the frequency of use and depth of explanation of professional terms in general optimization suggestion data, basic safety guidance data, or professional consultation guidance data are adjusted to generate language complexity adaptation data.

[0088] Generate personalized guidance form data based on language complexity adaptation data and information delivery channels;

[0089] Construct stratified health guidance data based on risk grading data and personalized guidance form data.

[0090] The embodiment of the present invention analyzes the user's response time and execution completion rate to different types of guidance content presentation forms (such as text description, dynamic graphics, voice broadcast, video demonstration, text and graphics combination, etc.) based on the behavior tracking records in the user's historical feedback data. The response speed is defined as the time interval from the user's first attempt to perform the relevant action after receiving the guidance content, and the execution completion rate is calculated by comparing the action parameter suggestions in the guidance content with the action data generated after the user actually executes it, and calculating the proportion of the suggestions achieved. For example, in analyzing the training data of a certain user for 30 consecutive days, it was found that the user's average response time to voice broadcast guidance content was 12 minutes, and the execution completion rate was 82%; while the response time to text-based guidance content was as high as 50 minutes, and the completion rate was only 41%. The platform adopts a multi-dimensional response scoring model, combining the three dimensions of information form, response delay, and completion to construct a scoring matrix, and finally outputs the most suitable information delivery channel for the user as a combination of "voice broadcast + dynamic graphics". This result will serve as the basis for the subsequent generation of personalized guidance forms. Data tags related to specialized terminology are extracted from user feedback interaction logs and training behaviors. For example, user responses to terms like "scapular stability," "isometric contraction," or "hip flexor activation" are recorded, including whether they consulted for explanations, whether they delayed execution due to confusion, and whether they proactively requested simplification. The platform also categorizes terms into three categories: basic, intermediate, and advanced. Using a decision tree-based term acceptance modeling approach, the platform assigns acceptance weights to each category based on user response time, frequency of consultation, and execution deviation. For example, a user who spent an average of 70 seconds consulting intermediate or advanced terms like "lower cross syndrome" and experienced a deviation in instructional content execution exceeding 20% ​​was classified as a "low-acceptance term" by the platform, receiving a corresponding weight of only 0.3 in the user's term acceptance model. The term acceptance model ultimately generates a personalized term acceptance threshold curve for each user, which is used to guide language complexity adaptation strategies. Based on the output of the term acceptance model, the platform automatically retrieves the frequency and level of terms in the current instructional content and compares them with the user's acceptance curve for selection and replacement. Specifically, the platform uses a semantic replacement template library to replace low-acceptance terms with synonymous but popular expressions, such as replacing "scapula winging" with "scapula valgus"; at the same time, when retaining the terms, annotations or pop-up explanation modules are added to clarify the meaning of the terms and the training-related meaning. The platform uses a content complexity calculation model to evaluate the Flesch reading difficulty index of the text before and after the adjustment, and saves the adjusted version together with the original version as language complexity adaptation data, giving it identification labels (such as "standard version", "simplified version", "graphic enhanced version", etc.) so that the platform can flexibly call it according to user needs when outputting content. For example, for users whose term acceptance threshold is less than 0.5, a "simplified version + semantic annotation type" content version is output. This data provides a language-level adaptation basis for the subsequent generation of personalized guidance content.Combining the user's optimal information delivery channel (e.g., voice broadcast + dynamic graphics) with the "simplified version + semantically annotated" content version from language complexity adaptation data, the system uses a content generation engine to dynamically assemble the final output format. The system generates a multimodal content package based on the combined dimensions of "carrier format + language style + interaction method." For example, the output might be "short voice instructions + interactive animation + simplified terminology." The voice content is kept to under 20 seconds, the animation annotates joint movement trajectories, and provides an interactive button for users to click to view terminology explanations. The platform assigns a unique ID to the personalized instruction format data and associates it with the user's ID. This data is then tracked for display, frequency of use, and effectiveness, allowing for continuous optimization of the output strategy. For example, the instructional output for a sports rehabilitation user during postoperative training might be: "Please slowly raise your arms outward to shoulder level while keeping your shoulders against the wall (this is simplified terminology for preventing scapular valgus). You can watch the animation on the right and listen to the voice prompts to complete the movement." This achieves highly adaptive personalized communication. By integrating movement deviation risk levels (low, medium, and high), personalized guidance format data, and language adaptation strategies, a guidance content template system tailored to users' tiered health intervention needs is constructed. For low-risk deviations, the system matches general optimization recommendation data, primarily using simplified terminology and illustrations. For medium-risk deviations, the system matches basic safety guidance data, employing a medium-complexity format of illustrations, audio, and term annotations. For high-risk deviations, the platform leverages professional consultation guidance data, including original medical terminology, explanatory documents, and a doctor's appointment portal. The platform dynamically adjusts the tiering strategy for matching content based on user history and terminology acceptance models. For example, if a user's acceptance of the term "rotator cuff tear risk" in the high-risk category is only 0.4, the platform will add an illustration and audio explanation of the term to the high-risk guidance data, and simultaneously send a link to physician guidance via SMS. This creates a tiered health guidance data structured as "high risk, low term acceptance, and multimodal re-guidance," enabling refined management. This tiered health guidance data serves as the core reference for the platform's intelligent intervention decisions and can be dynamically utilized during subsequent training, rehabilitation, or consultations.

[0091] Preferably, decomposing the hierarchical health guidance data into quantifiable progress units in step S3 includes:

[0092] Structurally disassemble the hierarchical health guidance data, identify the core improvement goals in the health guidance content, and obtain improvement goal extraction data;

[0093] Extracting the action angle control, muscle stability, movement rhythm and joint range of motion from the improved target extraction data to form improved target structure data;

[0094] Based on the improvement target structure data, the phased improvement standards are designed, and the long-term improvement goals are decomposed into three levels of improvement goals: short-term, medium-term, and long-term, to obtain multi-time domain improvement target mapping data;

[0095] Refine the multi-temporal progress target mapping data into micro-progress units that can be directly measured in daily training, including angle deviation reduction value, stability improvement percentage and movement fluency improvement index, to form a micro-progression assessment module;

[0096] Convert micro-progress assessment modules into specific numerical indicators and visual expressions, thereby generating quantifiable progress unit data;

[0097] Set the collection frequency and feedback cycle for progress unit data, and record the micro-progress achieved during the user's training process to obtain progress tracking data.

[0098] After constructing the layered health guidance data, the embodiment of the present invention uses a semantic understanding and contextual relationship analysis model (such as a nested intent recognition algorithm based on BERT) to structure the guidance content and convert the training objectives contained in the text into clear "improvement target" items. For example, for the high-level guidance content of "maintaining scapula stability while performing bent-over rowing", the platform automatically breaks it down into two core improvement goals: "enhancing scapula area muscle stability" and "controlling upper limb rowing angle stability". This process first performs entity recognition and verb target mapping on the guidance sentence, extracts action control words such as "maintain" and "perform", and then performs context matching with the training parts (such as "scapula" and "upper limb") to generate a preliminary target list. The platform then prioritizes and removes redundancy based on task word frequency, action result type and associated risk level, and finally forms structured "improvement target extraction data". This data contains fields such as target name, associated action, affected part, priority level, etc., which is the basic data for subsequent action-level analysis. From the extracted goals, four core variables involving quantifiable dimensions are selected: movement angle control (e.g., squat angle accuracy), muscle stability (e.g., scapular EMG signal stability), movement rhythm (e.g., movement execution time consistency), and joint range of motion (e.g., hip abduction amplitude). The system uses a mapping model between goals and sensor data to associate goal fields with measurable parameters within the platform. For example, for the goal "Improve shoulder abduction control," the platform extracts "shoulder abduction angle error less than 5 degrees" from the angle control category as the angle control sub-goal, and "average EMG fluctuation rate of the target muscle group less than 15%" from the muscle stability category as the stability sub-goal. A unified structured data entry is constructed, recording the calculation method, required equipment (e.g., inertial sensors, EMG devices), and units for each sub-goal. Finally, the platform generates "improved goal structure data," organized in a JSON format. Each entry includes the goal dimension, reference standard, measurement method, and the current user starting state, serving as input for phased goal division and micro-progress monitoring. Based on the difference between the initial value and the target expected value in the improvement target structure data, combined with the user's training frequency and risk level, an interval-based progressive strategy is used to divide the target into three phases: short-term (e.g., within 7 days), medium-term (e.g., within 21 days), and long-term (e.g., within 60 days). The system divides the overall improvement path into arithmetic or weighted incremental target intervals, and sets reasonable increments based on the average user progress in the platform's big data. For example, if a user's "shoulder abduction angle control error" starts at 12 degrees and the platform's target is to control it within 3 degrees, the system sets a short-term target of 9 degrees, a medium-term target of 6 degrees, and a long-term target of 3 degrees, reducing it by 3 degrees every 7 days. Similarly, the "muscle stability fluctuation rate" target is gradually reduced from the current 25% to 10%, with the platform adjusting the reduction rate to 5% and 10% for each phase based on training load.All phased goals are mapped to a unified timeline progress model, generating "multi-temporal progress goal mapping data" for dynamic feedback and user goal perception management. Each phased goal is further refined into measurable parameters for daily training units, such as a 0.5-degree reduction in angle error, a 1% decrease in EMG fluctuation rate, and a 2% reduction in execution time fluctuation. The platform quantifies these changes by defining "micro-progress units" (small improvements that can be achieved after a single or repeated training session) and establishing a micro-progress indicator library within the system. For example, in the angle control category, the "angle deviation reduction value" is defined as the difference in degree between the user's movement and the average posture of the target movement; muscle stability is expressed as the percentage decrease in the standard deviation of the EMG signal; and the movement smoothness improvement index is calculated by combining the smoothness of the acceleration curve and the frequency of movement interruptions. All these quantitative indicators are encapsulated into a "micro-progress evaluation module," which is automatically called by the platform's analysis module after each training session to output the evaluation results. The visualization template engine is called to convert micro-progress indicators into graphical data displays, such as using a line graph to show the trend of angle error changes, using a radar chart to show the improvement in muscle stability in various directions, and using a bar chart to compare the uniformity of movement rhythm in different time periods. Each indicator is given a clear numerical unit and stage label, and non-compliant items are marked with color (such as red warnings or yellow prompts). For example, a user's shoulder abduction angle deviation dropped from 5.2 degrees to 4.8 degrees on the same day. The platform displays "Today's Progress: Angle deviation reduced by 0.4 degrees (achievement rate 80%)", and at the same time, the "Fluency Index" chart shows an improvement of 12% compared to last week. All charts and values ​​are bound to the "Progress Unit Data", which records fields such as date, training project, indicator type, target value, actual value and achievement rate in a structured manner to support subsequent automatic aggregation and trend analysis. The data collection frequency (such as each training record for high-frequency training users and weekly summary for low-frequency users) and feedback cycle (such as daily summary push and weekly trend report push) are set for different user types. The system uses a rolling window statistical model to continuously sample the "progress unit data" and compare it with the target to identify trend changes in real time. For example, for the setting of one feedback cycle for every three training sessions, the platform automatically extracts the angle control and fluency indicators from the latest three training data to fit the trend line and determine whether it is currently in a state of progress, stagnation or regression. The platform will feed back the analysis results to the user in the form of a "micro-progress heat map" or a "weekly progress curve", and store them in the "progress tracking data". This data supports subsequent personalized training plan adjustments and re-optimization of the platform's health guidance strategy. The system will also automatically trigger the guidance content or terminology adjustment mechanism when the user's progress stagnates for a long time, forming a closed-loop optimization.

[0099] Preferably, performing anonymous user matching based on similar fitness characteristics and progress patterns on the progress tracking data in step S3 includes:

[0100] Conduct time series analysis on progress tracking data to identify characteristics of user progress rate, plateau period and breakthrough point, and generate progress pattern feature data;

[0101] Based on the progress pattern feature data, anonymous user groups with similar fitness goals, physical conditions and initial ability levels are extracted from a preset user database to form a candidate matching user pool;

[0102] Calculate the similarity between the progress pattern feature data and each user in the candidate matching user pool, identify anonymous users with high similarity in progress trajectories, and thus generate user matching results;

[0103] Based on the user matching results, the successful breakthrough strategies and motivational factors of the matching users are extracted, and the personalized motivational strategies are generated by combining the current user's progress pattern feature data.

[0104] Transform personalized motivation strategies into interactive motivational data that includes success story sharing, breakthrough point prompts, and virtual competition mechanisms.

[0105] This embodiment of the present invention receives progress unit data output by a user's micro-progress assessment module and constructs time series data based on timestamp information. It then analyzes the user's training data over multiple time periods using a sliding window approach (e.g., a 7-day window with one-day sliding intervals). The system identifies the user's progress rate within a specific time period by calculating the rate of change of progress within each window. When progress values ​​fluctuate minimally or stagnate over multiple consecutive windows, the system identifies a training stagnation period. When the system detects a significant increase in a progress indicator (e.g., movement fluency improvement index or angular deviation reduction value) exceeding a certain threshold (e.g., an increase of more than 20% compared to the previous stage average) within a short period of time, the system marks it as a breakthrough point. The platform encodes features such as progress rate, stagnation period, and breakthrough point in a structured manner to form progress pattern feature data, which serves as the basis for subsequent user clustering and matching. Based on the progress pattern feature data, the user feature database is searched. The database contains a large number of anonymous users' fitness goals (such as core strength enhancement, shoulder joint activity improvement, etc.), basic physical conditions (such as height, weight, gender, body fat percentage) and initial ability levels (such as initial controllable movement angle, basic stability score, etc.). The system first uses the feature standardization method to process the various indicators of the current user and database users, and uses the K-nearest neighbor algorithm in the standardized space to preliminarily screen out user groups that have similarities with the current user in goals and ability levels (Euclidean distance or Manhattan distance is less than a certain threshold, such as 0.3), forming a candidate matching user pool as the input basis for the next step of precise matching. For each user in the candidate matching user pool, we extract their complete historical progress pattern feature data (including the progress rate curve and breakthrough point distribution at each training stage). We then perform a dynamic time warping (DTW) similarity calculation between the current user's progress pattern features and each candidate user. The DTW method can capture the similarities between two progress trajectories at different time scales. The platform sets a similarity threshold (e.g., a similarity score greater than 0.85) to screen out users with similar progress trajectories. Finally, we output several anonymous users with high matching rankings (e.g., Top 5), forming the user matching results for subsequent successful strategy mining. Based on the anonymous user groups identified in the user matching results, their behavioral changes before and after the breakthrough point are analyzed, and their successful breakthrough strategies (such as adjusting training frequency, adding auxiliary training methods, and changing training time periods) and motivational factors (such as participating in virtual competitions, obtaining positive feedback, and setting phased rewards) are extracted. These contents are classified and labeled with their associated progress pattern features through the strategy encoding module. For example, a correspondence is established between "adding low-intensity stretching recovery training" and "slowing down stability improvement before the breakthrough point"; then, combined with the current user's progress pattern feature data, the most relevant breakthrough strategy set is matched to form a personalized incentive strategy, and the label structure (such as [phased stagnation + shoulder joint limitation] → [Strategy: adjust training angle + mild stretching assistance]) is output for the next step of conversion.The personalized incentive strategy is converted into multimodal interactive incentive data through the content generation module, which includes three main modules: 1) The success case sharing module generates personalized text and image content by matching the user's corpus materials and progress trajectory diagram, such as showing the story of "User A, who also has shoulder joint limitations, broke through the angle limit of 20 degrees through daily shoulder small-angle swing training within two weeks"; 2) The breakthrough point prompt module pushes prompt information such as "A breakthrough in movement stability is expected after the next training" in real time based on the current user data; 3) The virtual competition mechanism module sets challenge goals based on the user's current stage and the average level of matched users (such as "Increase the stability score by 5% within 3 days"), and enhances training motivation through rankings, point rewards, etc., and finally generates a terminal-oriented visual incentive interface.

[0106] Preferably, the step S3 of establishing an initial user motion pattern map based on the action correctness data, the layered health guidance data, and the progress tracking data includes:

[0107] Establish a user data fusion preprocessing mechanism based on action correctness data, layered health guidance data, and progress tracking data to generate a multi-source data integration table;

[0108] Perform time synchronization on multi-source data integration tables, establish a unified time axis reference system, and form a time series synchronization data set;

[0109] Extract users' common movement deviation patterns and technical defect features from movement correctness data to form a user technical feature profile;

[0110] Analyze users' acceptance and implementation of different types of guidance, as well as the rhythm of their ability development, based on stratified health guidance data and progress tracking data, to build user learning characteristic data.

[0111] Identify mutual influence relationships and dependency patterns based on user technical feature profiles and user learning feature data, and establish a feature association network;

[0112] Build a user athletic ability development prediction model based on the feature association network, and conduct short-term, medium-term, and long-term ability predictions to form a user potential assessment report;

[0113] A visualization scheme is generated for the user potential assessment report, including node definition, connection rules and graph hierarchy. Visual optimization and interaction design are performed based on highlighting important nodes, coloring key paths and interactive functions, thus forming an initial user movement pattern graph.

[0114] In an embodiment of the present invention, the system integrates various types of data generated by users during training. The system first obtains real-time scoring results of user training movements from the movement correctness recognition module, including data such as angle deviation, posture error, and disordered movement execution sequence. Simultaneously, it extracts personalized training recommendations, movement decomposition information, and control points recommended to users from the hierarchical health guidance system, and simultaneously reads the micro-progress change values ​​and feedback periodicity records recorded by the progress tracking module. Through a data fusion preprocessing mechanism, the platform establishes a unified data interface with user ID and timestamp as anchors. It relies on data cleaning functions to perform linear interpolation or historical data backfilling for missing items. For health guidance data with large structural differences, it adopts a structural mapping conversion strategy (such as mapping guidance suggestion text labels to structured control elements). Finally, it arranges and combines data record units in a unified format according to time, constructing a multi-source data integration table that can be used for analysis, and is then used for subsequent multi-dimensional modeling and analysis. After obtaining the multi-source data integration table, the records from different data sources need to be synchronized in time. Since motion capture data is usually collected at the second level during user training, and health guidance feedback and progress tracking data are mostly collected at the daily or weekly granularity, the platform uses a timeline alignment algorithm to map all data to a unified training day-level timeline. The specific approach is to set a unified training time slice (such as each time slice is one day), downsample the high-frequency data (such as taking the maximum value and average value within the day), and upsample and fill the low-frequency data (such as using forward filling or moving average) to ensure that each time slice contains complete information such as movement correctness, health guidance acceptance status, and progress feedback, forming a time-series synchronized data set, and laying a time-series consistency foundation for subsequent user behavior feature modeling. Based on the time-series synchronized data set, the platform uses statistical clustering and pattern recognition methods to analyze user movement correctness data, and extract users' common movement deviation patterns and technical defect characteristics. The platform first uses the Local Outlier Factor algorithm to identify posture anomalies that recur in the training sequence (such as knee valgus, shoulder compensation, etc.), and then uses the K-means clustering method to classify these anomaly types. At the same time, it combines the error value and the position of the action key frame to extract key technical defects (such as insufficient core strength leading to lower limb instability, upper limb movement delay, etc.). All these features are written into the user's technical feature file in a standardized description. Each record includes the name of the deviation action, frequency of occurrence, error amplitude, scope of impact and inferred cause, etc., which become the key input features for subsequent ability development prediction. Combined with the user's technical feature file, the user's acceptance of different types of guidance content and the training feedback effect are further analyzed from the stratified health guidance data and progress tracking data.The platform uses a weighted scoring mechanism to correlate the magnitude of micro-improvements achieved within a specific time window after users receive specific health coaching types (such as paced guidance, movement easing, and local activation). It also assesses the user's progress cycle (i.e., the number of days from receiving coaching to achieving their goals) within each training mode, extracting characteristics of the pace of ability development, such as "quick response to muscle activation suggestions and delayed response to pace control suggestions." Finally, the platform uses a combination of labels to construct user learning characteristic data, including dimensions such as guidance preference category, feedback sensitivity, and training adaptation pace, providing a basis for ability development prediction and personalized content delivery. Based on user technical feature profiles and learning feature data, a feature association network is constructed using graph relational modeling. Specifically, each type of technical feature and learning feature is used as a node, and edge-weighted connections are established between nodes based on co-occurrence frequency, causal order, and correlation metrics. For example, "movement delay" and "slow pace adaptation" form a strong edge connection due to their frequent co-occurrence during the same training phase. The platform further uses PageRank or graph embedding algorithms to identify key nodes at the center of the network and their influencing paths. This feature association network reflects the dynamic interdependence between the root causes of a user's ability limitations and their training adaptation methods, serving as a crucial structural foundation for deducing subsequent ability development trends. Based on this feature association network, the platform designed a user athletic ability development prediction model. A graph convolutional neural network (GCN) was used to represent and learn the user's node features. The user's historical progress time series was used as the temporal input channel. By integrating the graph structure with the time series data, the model outputs predictions for the user's key ability indicators (such as stability score, movement accuracy, and reaction speed) over the short (7 days), medium (30 days), and long (90 days) periods. The platform uses historical matching of the user's evolutionary trajectory to adjust the confidence level of the prediction results and outputs a user potential assessment report, which identifies the user's current ability position, the predicted upper limit of achievable values, and recommended development paths and breakthrough points. The core ability nodes and their evolutionary paths in the user potential assessment report are visualized and mapped to form an initial user athletic pattern map. Specifically, it includes three aspects: 1) The node definition module converts the user's ability components (such as movement control, rhythm control, and joint flexibility) into graph nodes; 2) The connection rules establish directed edges between nodes based on the feature association network, and set edge weights according to the training impact path and time dependency; 3) The graph hierarchy is divided into basic ability layer, comprehensive ability layer and target ability layer.The platform further implements visual optimization on the front-end interface, including highlighting the size and color of important nodes, using gradient lines on key paths to deepen user attention, and supporting users to view training suggestions and predict change trends by clicking on nodes, realizing interactive functions, and finally outputting visual results in the form of a map to guide users' long-term training path planning and dynamic target adjustment.

[0115] Preferably, step S4 includes the following steps:

[0116] Step S41: Extracting environmental characteristic parameters from each data point in the initial user motion pattern map during collection, and combining the image clarity score, skeleton point recognition confidence, and tracking stability score to form a data collection environment quality assessment table, where the environmental characteristic parameters include light intensity, background complexity, and device stability;

[0117] After completing the construction of the initial user motion pattern map, the embodiment of the present invention further extracts the environmental characteristic parameters of the video clips or image frames corresponding to each data point in the map at the time of collection, which are used to evaluate the degree of external interference in the data quality. Specifically, the platform uses built-in or external sensor modules to collect light intensity (in Lux), background complexity of the camera device (calculated by the number of edges or color histogram dispersion in the image background area), and device stability (measured by IMU camera shake frequency and amplitude). The platform then combines the image clarity score extracted from the image processing module (calculated by image gradient, texture richness or Laplace variance), skeleton point recognition confidence (derived from the confidence interval score of each skeleton point by the posture recognition model), and tracking stability score (measurement of the position fluctuation degree of the same skeleton point in consecutive frames) and other internal processing indicators to conduct a comprehensive analysis of each data point and summarize it into a data collection environment quality evaluation table. This table records the external environmental interference factors and internal image processing quality corresponding to each motion data point, providing a basis for subsequent judgment of data reliability.

[0118] Step S42: Evaluate the degree to which each data point is affected by the environment based on the consistency performance of each data point according to the data collection environment quality evaluation table, and generate a collection environment stability index;

[0119] After obtaining the data acquisition environment quality evaluation table, the embodiment of the present invention evaluates the degree to which each data point is affected by the environment based on the consistency between the environmental interference index and the image processing stability index of each data point. The platform first normalizes the degree of fluctuation of each data point in three external parameters (light intensity, background complexity, and device stability), and sets weight coefficients (such as background complexity weight 0.4, light intensity weight 0.3, and device stability weight 0.3), compares the corresponding image quality degradation, and constructs an impact mapping relationship. Then, through correlation analysis, the strength of the linear relationship between the image quality of the data point and the environmental fluctuation is judged to determine its environmental sensitivity. For data points whose image quality is significantly affected by environmental changes, their acquisition environment stability index is low; otherwise, it is high. The platform generates an acquisition environment stability index for each data point by comprehensively calculating the stability scores of all data points, which is used to mark whether the data is generated under ideal acquisition conditions.

[0120] Step S43: Record and analyze the user's feedback behavior on the health guidance data provided by the system, including acceptance, execution completion and subjective evaluation, evaluate the temporal stability and content consistency of the user's feedback, and establish a user feedback consistency index;

[0121] The embodiment of the present invention comprehensively records and analyzes the user's feedback behavior on the provided health guidance data during the use of the system. The behavioral data includes three aspects: first, whether the user accepts the health guidance task recommended by the system, that is, "acceptance"; second, whether the user completes the task according to the guidance content and uploads the training record, as "execution completion status"; third, whether the user provides subjective evaluation, such as marking the effectiveness, giving text feedback, etc., as "subjective evaluation". The platform records the timestamp and content of these feedback data, and performs time series analysis to extract the time stability of the feedback (that is, whether the user feedback occurs regularly and whether it is consistent with the training cycle) and content consistency (such as whether the same type of guidance is consistently evaluated and whether there are changes in opinions during repeated execution). Through the fusion of these dimensions, the platform constructs a user feedback consistency index for each user. Users with a high index represent that their feedback has stability, availability and a high credible reference value.

[0122] Step S44: performing weighted fusion of the acquisition environment stability index and the user feedback consistency index, and combining the importance of each data point in the motion assessment to generate a data reliability weight coefficient;

[0123] After obtaining the acquisition environment stability index and user feedback consistency index of each data point, the embodiment of the present invention further weights and fuses these two indexes to form a comprehensive data reliability evaluation index. The platform first sets a weighting factor, for example, the acquisition environment stability accounts for 0.6, and the user feedback consistency accounts for 0.4, reflecting the platform's comprehensive trust in the external acquisition environment and the user's subjective response; then, for each data point in the initial user motion pattern map, combined with its position and role in the map (for example, whether it is a key node or path node, whether it participates in the ability assessment model training), the importance of the data point in the motion assessment is determined, and it is introduced as an adjustment factor in the weighted calculation, and finally a data reliability weight coefficient is assigned to each data point. This coefficient is used to quantify whether the data point is qualified to be trusted and retained in the analysis model, and is a key mediating variable for the platform to build a high-quality evaluation system.

[0124] Step S45: setting a screening threshold according to the data reliability weight coefficient to identify and extract high-confidence data points, forming a high-quality data subset, and constructing a user movement evaluation model;

[0125] The embodiment of the present invention uses the data reliability weight coefficient as the screening basis, sets the reliability threshold (such as 0.75), identifies and extracts high-confidence data points, and forms a high-quality data subset. The high-quality data subset excludes data points generated in abnormal collection environments or by users with inconsistent feedback, and retains data sources with high platform credibility, good training environment, and positive user feedback. On this basis, the platform uses the subset to retrain the user motion evaluation model. The evaluation model adopts a multi-input and multi-output neural network structure, and takes factors such as user technical characteristics, learning characteristics, behavioral performance, and external interference as input nodes. Through training, it extracts the user's current sports ability level score and outputs the predicted trend. The high-quality data subset ensures that the training results of the model are more accurate and more robust, thereby improving the overall reliability of the platform's user ability analysis and personalized recommendation capabilities.

[0126] Step S46: Analyze the operation fluency and usage frequency of each functional module based on the user motion evaluation model and interactive incentive data to generate functional mastery evaluation data; implement a low-threshold progressive function unlocking mechanism based on the functional mastery evaluation data and interactive incentive data.

[0127] In the embodiment of the present invention, based on the updated user motion evaluation model, the platform further analyzes the user's usage trajectory and interaction records in each functional module (such as action correction, health plan recommendation, feedback record, interactive incentive module, etc.), evaluates the operation fluency (such as the number of clicks required to complete an operation, the average stay time, the error frequency) and the usage frequency (such as the number of daily visits, the single usage time, the operation coverage), and generates the function mastery evaluation data of each functional module. The platform associates these evaluation data with the user's interactive incentive data (such as the number of incentive points obtained, the number of task achievements achieved, the frequency of participating in challenges, etc.) to determine the user's current stage level of mastery of the platform functions. On this basis, the platform sets a low-threshold progressive function unlocking mechanism, that is, when the user's mastery reaches the basic requirements, more training modules or analysis reports are automatically unlocked. For example, when the mastery score of the "action evaluation function" exceeds 60 points and meets the standard for three consecutive days, the "target prediction function" is automatically unlocked; this mechanism not only reduces the user's initial cognitive burden, but also enhances the sense of accomplishment of later function exploration, thereby improving the overall platform stickiness and user satisfaction.

[0128] It is particularly important that step S46 includes the following steps:

[0129] Based on the user movement evaluation model, the user's usage time, access frequency and operation sequence in each functional module of the platform are recorded, and time distribution analysis is performed to identify user usage peaks and changes in functional preferences. In addition, a user function demand mapping relationship is constructed based on user capability characteristics.

[0130] Extract operation path features and decision node selection tendencies based on user function requirement mapping relationships. Combined with user response features in interactive incentive data, identify user function exploration methods and learning paths, and generate user behavior pattern analysis results.

[0131] Based on the results of user behavior pattern analysis, statistics are collected on the user's operation completion time, number of retries, and auxiliary information query frequency in each functional module. Combined with the frequency and duration of function use, the operation fluency index and function familiarity score of each function are calculated to form function mastery assessment data;

[0132] Implement a low-threshold progressive function unlocking mechanism based on function mastery assessment data and interactive incentive data.

[0133] After constructing a high-quality athletic ability profile based on the user's athletic assessment model, the platform, in this embodiment of the present invention, further records detailed user behavior data across each platform module in real time, including the start and end time of each function usage, access frequency, click sequence, and page jump path. By performing timestamp parsing and sequence reconstruction on this data, the platform extracts the usage duration and access frequency of each user in specific functions (such as "Movement Guidance," "Feedback Submission," and "Training Task Browsing") and establishes a timeline of operational events. The platform uses temporal distribution analysis methods (such as hourly-granular heatmap generation or normal distribution fitting) to identify peak user function usage periods (e.g., 7:00 PM–9:00 PM daily) and their temporal trends. Furthermore, the platform correlates the most frequently used functions with the ability profile reflected in the user's athletic assessment model (e.g., users with higher core stability scores prefer dynamic balance training modules). This allows for a mapping of user function needs, clarifying the function preferences and usage habits of each user group at different ability levels. Based on the mapping of user function requirements, the platform further extracts the characteristics of each user's common operation paths within the platform. This is the logic of the user's transitions between modules when using the platform, such as the sequence "training plan → action guidance → feedback record → personal summary." The platform utilizes path sequence mining algorithms (such as PrefixSpan sequence pattern recognition) to identify frequently occurring operation paths and analyzes the selection trends at key nodes (i.e., "decision nodes") within each path. For example, in the "Training Recommendations" module, users tend to click "Basic Difficulty" rather than "Advanced Challenge." Based on this, combined with user response characteristics from interactive incentive data (such as response time, completion rate, and skip rate to incentive tasks), the platform can identify users' function exploration methods (e.g., linear progression, skip attempts, repeated review) and their overall learning path preferences (e.g., preference for understanding content first and then practicing, or learning while doing). Through this analysis, the platform generates user behavior pattern analysis results, forming a structured profile of users' function usage logic, selection habits, and exploration depth within the platform, providing behavioral evidence for subsequent function proficiency assessment. Leveraging the results of user behavior pattern analysis, we quantitatively evaluate each user's operational performance in each functional module, primarily across three key dimensions: operation completion time (the average time it takes for a user to successfully complete a task after entering a functional module), number of retries (the average number of failed operations, rollbacks, and resubmissions during the completion of the same functional task), and frequency of auxiliary information queries (such as the number of times the "Help" button is clicked, "Operation Examples," or "Frequently Asked Questions" are viewed). The platform aggregates these behavioral data, combines them with the frequency and duration of use of the functional modules, and uses a weighted calculation to derive the "Operation Fluency Index" (a measure of the user's proficiency and ease of operation in using the function) and "Function Familiarity Score" (a measure of whether the user has the ability to independently complete functional operations) for each functional module.For example, in the "Personalized Training Recommendations" module, if a user has a short average completion time, low number of retries, rarely consults the help documentation, and uses the module frequently, the system will assign higher fluency and familiarity scores; otherwise, lower scores will be assigned. These scores will be used to further develop an assessment framework for users' mastery of platform functionality. Based on each user's fluency index and familiarity score across different modules, combined with their interactive incentive data (such as points earned, incentive task completion cycle, and consecutive days of use), a "low-threshold progressive feature unlocking mechanism" is dynamically implemented. The core of this mechanism is to automatically release more advanced features based on the user's current level of mastery, thereby increasing user exploration enthusiasm and lowering the threshold for functional understanding. For example, the platform sets the unlocking threshold for the "Training Load Adjustment" module as follows: a user's familiarity score ≥70 in the "Action Assessment" module and a completion rate of ≥60% for incentive tasks. This triggers the automatic unlocking process and simultaneously prompts the user to "unlock new features" through a notification. Furthermore, to encourage users to continuously improve their usage, the platform adopts a "low threshold + time-driven" strategy. This means that even if a user's rating is slightly low, by continuously using a basic functional module for more than five days and accumulating a certain amount of behavioral records, the next level of functionality, such as "Training Prediction" and "Action Comparison and Review," will be gradually unlocked. This progressive unlocking strategy ensures a controllable user experience while encouraging users to gradually master the full functional chain, thereby increasing overall platform usage depth and stickiness.

[0134] The present invention also includes a system for constructing a sports information intelligent service platform, which is used to execute the above-mentioned method for constructing a sports information intelligent service platform. The system for constructing a sports information intelligent service platform includes:

[0135] The motion trajectory analysis module is used to collect user motion image data and basic health data; extract the position information of key body nodes during the user's movement; generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain the movement correctness data;

[0136] The fitness risk assessment module is used to build a user fitness risk assessment model based on movement correctness data and user health basic data, analyze the difference from the preset safety threshold, and generate stratified health guidance data;

[0137] The progress tracking and interactive matching module is used to decompose the layered health guidance data into quantifiable progress units to generate progress tracking data; anonymously match the progress tracking data with users based on similar fitness characteristics and progress patterns to form interactive incentive data; and establish an initial user movement pattern map based on movement correctness data, layered health guidance data, and progress tracking data.

[0138] The user motion evaluation module is used to collect environmental stability index and user feedback consistency index of each data point of the initial user motion pattern map, and screen high-confidence data points to build a user motion evaluation model; based on the user motion evaluation model and interactive incentive data, the operation fluency and usage frequency of each functional module are analyzed, and a low-threshold progressive function unlocking mechanism is implemented in combination with the interactive incentive data.

[0139] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0140] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a sports information intelligent service platform, characterized in that: The following steps are involved: Step S1: Collect user motion image data and user basic health data; extract key body node position information during the user's motion process; Generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain the action correctness data; Step S2: Build a user fitness risk assessment model based on the action correctness data and the user's basic health data, analyze the difference from the preset safety threshold, and generate layered health guidance data; Step S3: Decomposing the hierarchical health guidance data into quantifiable progress units to generate progress tracking data; performing anonymous user matching based on similar fitness characteristics and progress patterns on the progress tracking data to form interactive incentive data; Build an initial user movement pattern map based on movement correctness data, layered health guidance data, and progress tracking data; Step S4: Calculate the collection environment stability index and user feedback consistency index of each data point of the initial user motion pattern map and screen high-confidence data points to build a user motion evaluation model; analyze the operation fluency and usage frequency of each functional module based on the user motion evaluation model and interactive incentive data, and implement a low-threshold progressive function unlocking mechanism in combination with the interactive incentive data.

2. The method for constructing a sports information intelligent service platform according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Capturing multiple frames of continuous images of a user performing fitness movements through a camera of a smart terminal device to generate user motion image data; Step S12: Collecting the user's height, weight, age, and exercise habits to form basic health data of the user; Step S13: performing image preprocessing on the user's motion image data and performing skeleton point recognition to locate the user's key body nodes of the shoulder, elbow, wrist, hip, knee, and ankle, and generate key body node position information; Step S14: Generate a user motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain action correctness data.

3. The method for constructing a sports information intelligent service platform according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: performing time series analysis on the position information of key body nodes, recording the position changes of each key node in continuous time frames, and generating raw data of node motion trajectory; Step S142: Smoothing the original data of the node motion trajectory to obtain user motion trajectory curve data; Step S143: Calculating the motion angle, velocity, and acceleration of the human motion segments formed between adjacent key nodes based on the user motion trajectory curve data, and constructing a dynamic parameter set of the user motion segments; Step S144: matching the user's motion segment dynamic parameter set with the standard parameter set in the action library to determine the type of fitness action currently performed by the user and generate an action type recognition result; Step S145: extracting corresponding standard trajectory template data from the action standard library according to the action type recognition result, and performing spatiotemporal alignment between the user motion trajectory curve data and the standard trajectory template data to obtain normalized trajectory comparison data; Step S146: Calculate the Euclidean distance deviation between the user trajectory and the standard trajectory at each key time point in the normalized trajectory comparison data to generate a point deviation sequence, and calculate the angular deviation between the user motion segment and the standard motion segment at each key time point to generate an angular deviation sequence; Step S147: Score the overall motion trajectory deviation based on the point deviation sequence and the angle deviation sequence to obtain motion correctness data.

4. The method for constructing a sports information intelligent service platform according to claim 3, characterized in that: Step S2 includes the following steps: Step S21: extracting action deviation characteristic indicators based on the action correctness data to form action risk basic assessment data, wherein the action deviation characteristic indicators include the maximum deviation value, average deviation value and deviation duration of key nodes; Step S22: extracting user physiological characteristic parameters from the user's basic health data to generate user physiological risk factor data; Step S23: establishing a correlation model between the basic action risk assessment data and the user's physiological risk factor data, calculating the degree of injury risk to the user caused by various action deviations, and generating an individualized risk assessment value; Step S24: stratify the risk levels according to the individualized risk assessment value and the preset safety threshold to generate risk grading data; Step S25: performing differentiated processing on action deviations of different risk levels according to the risk classification data, and recording the action deviation patterns and guidance responses to obtain user historical feedback data; Step S26: Visualize the health guidance content based on the risk classification data and the user's historical feedback data, thereby generating layered health guidance data.

5. The method for constructing a sports information intelligent service platform according to claim 4, characterized in that: Step S25 includes the following steps: Step S251: extracting motion segment angle deviation values, speed control deviation values, and motion stability indices from low-risk motion deviations in the risk classification data, performing quantitative comparisons with preset standard motion models, identifying optimizable parameters, and obtaining low-risk optimizable parameters; Step S252: Formulate a progressive adjustment plan for low-risk optimizable parameters to form general optimization suggestion data; Step S253: analyzing the proximity between the movement deviation point and the human biomechanical safety limit for the movement deviation of the medium risk level in the risk classification data, and calculating the potential stress concentration area; Step S254: Based on the potential stress concentration areas and the user's physiological risk factor data, the risk of injury to the user's joints, muscles, and ligaments caused by the deviation is assessed, and basic safety guidance data is generated, including a description of prohibited actions, a definition of safe action ranges, and recommendations for alternative actions. Step S255: extracting the degree and duration of the high-risk action deviations exceeding the safety threshold from the risk classification data, and performing error classification based on the high-risk action deviations to obtain error type data, wherein the error type data includes technical errors, physical limitation errors, or potential health problem errors; Step S256: Generate professional consultation guidance data according to the error type data; Step S257: Record the user's behavioral response after receiving general optimization suggestion data, basic safety guidance data or professional consultation guidance data, including the time delay in receiving guidance, the completion degree of executing guidance and the persistence degree of continuously applying guidance, and generate user historical feedback data.

6. The method for constructing a sports information intelligent service platform according to claim 5, characterized in that: Decomposing the stratified health coaching data into quantifiable progress units in step S3 includes: Structurally disassemble the hierarchical health guidance data, identify the core improvement goals in the health guidance content, and obtain improvement goal extraction data; Extracting the action angle control, muscle stability, movement rhythm and joint range of motion from the improved target extraction data to form improved target structure data; Based on the improvement target structure data, the phased improvement standards are designed, and the long-term improvement goals are decomposed into three levels of improvement goals: short-term, medium-term, and long-term, to obtain multi-time domain improvement target mapping data; Refine the multi-temporal progress target mapping data into micro-progress units that can be directly measured in daily training, including angle deviation reduction value, stability improvement percentage and movement fluency improvement index, to form a micro-progression assessment module; Convert micro-progress assessment modules into specific numerical indicators and visual expressions, thereby generating quantifiable progress unit data; Set the collection frequency and feedback cycle for progress unit data, and record the micro-progress achieved during the user's training process to obtain progress tracking data.

7. The method for constructing a sports information intelligent service platform according to claim 6, characterized in that: The anonymous user matching based on similar fitness characteristics and progress patterns of the progress tracking data in step S3 includes: Conduct time series analysis on progress tracking data to identify characteristics of user progress rate, plateau period and breakthrough point, and generate progress pattern feature data; Based on the progress pattern feature data, anonymous user groups with similar fitness goals, physical conditions and initial ability levels are extracted from a preset user database to form a candidate matching user pool; Calculate the similarity between the progress pattern feature data and each user in the candidate matching user pool, identify anonymous users with high similarity in progress trajectories, and thus generate user matching results; Based on the user matching results, the successful breakthrough strategies and motivational factors of the matching users are extracted, and the personalized motivational strategies are generated by combining the current user's progress pattern feature data. Transform personalized motivation strategies into interactive motivational data that includes success story sharing, breakthrough point prompts, and virtual competition mechanisms.

8. The method for constructing a sports information intelligent service platform according to claim 7, characterized in that: The step S3 of establishing an initial user motion pattern map based on the movement correctness data, the layered health guidance data, and the progress tracking data includes: Establish a user data fusion preprocessing mechanism based on action correctness data, layered health guidance data, and progress tracking data to generate a multi-source data integration table; Perform time synchronization on multi-source data integration tables, establish a unified time axis reference system, and form a time series synchronization data set; Extract users' common movement deviation patterns and technical defect features from movement correctness data to form a user technical feature profile; Analyze users' acceptance and implementation of different types of guidance, as well as the rhythm of their ability development, based on stratified health guidance data and progress tracking data, to build user learning characteristic data. Identify mutual influence relationships and dependency patterns based on user technical feature profiles and user learning feature data, and establish a feature association network; Build a user athletic ability development prediction model based on the feature association network, and conduct short-term, medium-term, and long-term ability predictions to form a user potential assessment report; A visualization scheme is generated for the user potential assessment report, including node definition, connection rules and graph hierarchy. Visual optimization and interaction design are performed based on highlighting important nodes, coloring key paths and interactive functions, thus forming an initial user movement pattern graph.

9. The method for constructing a sports information intelligent service platform according to claim 8, characterized in that: Step S4 includes the following steps: Step S41: Extracting environmental characteristic parameters from each data point in the initial user motion pattern map during collection, and combining the image clarity score, skeleton point recognition confidence, and tracking stability score to form a data collection environment quality assessment table, where the environmental characteristic parameters include light intensity, background complexity, and device stability; Step S42: Evaluate the degree to which each data point is affected by the environment based on the consistency performance of each data point according to the data collection environment quality evaluation table, and generate a collection environment stability index; Step S43: Record and analyze the user's feedback behavior on the health guidance data provided by the system, including acceptance, execution completion and subjective evaluation, evaluate the temporal stability and content consistency of the user's feedback, and establish a user feedback consistency index; Step S44: performing weighted fusion of the acquisition environment stability index and the user feedback consistency index, and combining the importance of each data point in the motion assessment to generate a data reliability weight coefficient; Step S45: setting a screening threshold according to the data reliability weight coefficient to identify and extract high-confidence data points, forming a high-quality data subset, and constructing a user movement evaluation model; Step S46: Analyze the operation fluency and usage frequency of each functional module based on the user motion evaluation model and interactive incentive data to generate functional mastery evaluation data; implement a low-threshold progressive function unlocking mechanism based on the functional mastery evaluation data and interactive incentive data.

10. A system for constructing a sports information intelligent service platform, characterized in that: The method for constructing a sports information intelligent service platform according to claim 1 is used to execute the construction system of the sports information intelligent service platform, wherein the construction system comprises: The motion trajectory analysis module is used to collect user motion image data and basic health data; extract the position information of key body nodes during the user's movement; generate the user's motion trajectory curve based on the key node position information, and calculate the deviation angle and distance from the standard trajectory template to obtain the movement correctness data; The fitness risk assessment module is used to build a user fitness risk assessment model based on movement correctness data and user health basic data, analyze the difference from the preset safety threshold, and generate stratified health guidance data; The progress tracking and interactive matching module is used to decompose the layered health guidance data into quantifiable progress units to generate progress tracking data; anonymously match the progress tracking data with users based on similar fitness characteristics and progress patterns to form interactive incentive data; and establish an initial user movement pattern map based on movement correctness data, layered health guidance data, and progress tracking data. The user motion evaluation module is used to collect environmental stability index and user feedback consistency index of each data point of the initial user motion pattern map, and screen high-confidence data points to build a user motion evaluation model; based on the user motion evaluation model and interactive incentive data, the operation fluency and usage frequency of each functional module are analyzed, and a low-threshold progressive function unlocking mechanism is implemented in combination with the interactive incentive data.

Images