Intelligent sensing and feedback personalized sports training guidance system

Abstract

The application relates to the technical field of sports training and discloses an intelligent sensing and feedback personalized sports training guidance system, which comprises a dynamic interactive field sensing module, a data processing and analysis module, an immersive feedback module and a personalized model evolution module; the sensing module collects real-time interactive data through an intelligent training pad integrated with pressure and vibration sensors; the data processing and analysis module is internally provided with a predictive intention engine, which can evaluate a posture before a user action and trigger a feedforward early warning, and calculate an actual motion track of the user after the action, namely a fact path; the personalized model evolution module maintains a baseline model based on historical best performance of the user, and the data processing module calls the model to generate an idealized counterfactual path; the immersive feedback module presents an early warning signal on the training pad and synchronously visualizes playback of the fact path and the counterfactual path. The application realizes high self-adaptation and personalization of training guidance and significantly improves training efficiency.

Description

Technical Field

[0001] This invention relates to the field of sports training technology, specifically to a personalized sports training guidance system with intelligent sensing and feedback. Background Technology

[0002] Modern sports training increasingly relies on technology to quantify athlete performance and provide improvement suggestions. Traditional training guidance methods, such as verbal instructions from coaches or video playback analysis, while effective, have limitations such as strong subjectivity and untimely feedback. To overcome these shortcomings, various sensor-based motion analysis systems have emerged, such as wearable devices using inertial measurement units (IMUs), computer vision-based motion capture systems, or pressure pads used to analyze plantar pressure distribution.

[0003] However, existing technological solutions still have significant shortcomings in providing efficient and personalized guidance. Most of these systems focus on post-exercise analysis of completed movements, and their feedback is inherently delayed, failing to identify and intervene in potential errors during the preparation phase. This results in missing the optimal opportunity to prevent injury and correct incorrect postures. Furthermore, the guidance or ideal trajectories they provide are often based on a universal "expert model." This one-size-fits-all standard ignores the vast differences in physical condition, skill level, and exercise habits among users, making the goals potentially unattainable for beginners and lacking challenge for advanced athletes.

[0004] Furthermore, current feedback presentation methods typically rely on remote screen displays, forcing users to divert their attention between training actions and data interpretation. This spatiotemporal separation between feedback and action execution weakens the intuitiveness and effectiveness of guidance. Simultaneously, these systems fail to establish a dynamic model that adapts and evolves with individual user progress, causing the guidance's relevance to diminish over time. Therefore, there is an urgent need for a novel training guidance solution that can provide feedforward alerts, generate personalized and achievable optimization goals, present feedback in an immersive and intuitive manner, and possess adaptive learning capabilities. Summary of the Invention

[0005] The technical problem this invention aims to solve is that existing sports training guidance technologies, when sensing a user's movement state, either rely on visual devices that are easily obstructed and affected by ambient light, or on invasive sensors that require cumbersome wear, and generally lack the ability to deeply analyze the dynamic interaction information between the user and the environment. Furthermore, their feedback mechanisms are usually lagging and non-intuitive, making it difficult to achieve real-time and effective guidance for user actions.

[0006] To address the aforementioned technical problems, this invention provides a personalized sports training guidance system and method that enables users to intuitively understand their own athletic performance and make effective adjustments through intelligent sensing and feedback.

[0007] The first aspect of this invention provides a personalized sports training guidance system with intelligent sensing and feedback, the system comprising: The dynamic interactive field sensing module is used to collect interactive data generated when a user moves on the interactive surface. The interactive data is a signal that reflects the dynamic force information between the user and the interactive surface. Specifically, it may include a pressure distribution matrix to characterize the pressure distribution and a multi-channel vibration time-domain signal to characterize the impact vibration.

[0008] The data processing and analysis module is communicatively connected to the dynamic interactive field perception module. Its core function is to generate two motion paths for comparative analysis based on the interactive data. First, the module processes real-time interactive data to determine the factual path on the interactive surface that represents the user's current actual motion performance. Second, the module invokes a preset personalized baseline model that stores the user's historical best or ideal motion performance patterns, and generates a counterfactual path on the interactive surface corresponding to the current action context. This counterfactual path represents the motion path the user should ideally complete.

[0009] An immersive feedback module is communicatively connected to the data processing and analysis module. This module receives the factual path and the counterfactual path and drives a visual feedback device to synchronously and visually differentiate the playback of the factual path and the counterfactual path on the interactive surface. By presenting the user's actual performance juxtaposed with their achievable ideal performance, it provides the user with an intuitive and comparable basis for correction.

[0010] In a preferred technical solution, in order to more accurately quantify the factual path, the factual path is specifically the user's center of gravity pressure point trajectory. The data processing and analysis module analyzes the pressure distribution matrix. A weighted average calculation is performed to determine the position of the trajectory in real time. ; in: ; ; In this formula, It is the first Each sensor at a time point Pressure reading, It is the fixed position of the sensor in the coordinate system of the interactive surface.

[0011] In a further technical solution, the innovation of this invention is also reflected in its predictive capability. The data processing and analysis module further includes a predictive intent engine. This engine does not analyze completed actions, but rather predicts the user's next action intent by analyzing an interaction field data sequence composed of the interaction data within a preset time window. The interactive field data sequence may specifically include the time rate of change of the pressure distribution matrix and the micro-vibration modes of the multi-channel vibration time-domain signal. Its prediction principle can be based on a Bayesian inference framework, achieved by calculating the intention to maximize the posterior probability. ; in, It is the observed interaction field data sequence. It is the first Candidate movement intentions, It is its prior probability, the likelihood function It is learned from the user's historical data stored in the personalized baseline model.

[0012] Based on the aforementioned predictive capabilities, this system can implement a feedforward feedback control. When the predictive intent engine predicts the user's movement intention and determines that there is a significant deviation between the user's preparation mode and the health mode stored in the personalized baseline model, the immersive feedback module can drive the visual feedback device to provide early warning feedback before the user actually completes the action, so as to guide the user to make real-time adjustments before the action is executed, thereby avoiding potential action execution deviations.

[0013] To realize the function of the immersive feedback module, the visual feedback device may be a ground projector or a controllable light-emitting array integrated into the interactive surface.

[0014] In one specific embodiment, the interactive surface may be implemented as a smart training pad. The smart training pad may include a flexible substrate material and a sensor array embedded in the substrate material in an array form, consisting of a plurality of pressure sensors (e.g., piezoelectric sensors or piezoresistive sensors) to form the pressure distribution matrix.

[0015] Furthermore, the smart training mat can integrate one or more highly sensitive vibration sensors (e.g., accelerometers) to acquire the multi-channel vibration time-domain signals. This structure allows the interactive surface to not only provide motion support for the user but also function as a large-area, non-invasive sensor to accurately capture dynamic interaction information between the user and the ground.

[0016] A second aspect of this invention provides a personalized sports training guidance method based on intelligent sensing and feedback. This method is implemented through the aforementioned system and includes the following steps: 1) The dynamic interactive field sensing module collects interactive data generated when the user moves on the interactive surface. The interactive data includes a pressure distribution matrix and multi-channel vibration time-domain signals. 2) In one embodiment, before a macroscopic action occurs, the interaction field data sequence composed of the interaction data within a preset time window can be analyzed to predict the user's next movement intention and selectively provide feedforward early warning feedback. 3) Process the interaction data to determine the factual path on the interaction surface that represents the user's actual motion performance; 4) Invoke a preset personalized baseline model that stores the user's ideal motion performance pattern to generate a counterfactual path on the interaction surface that corresponds to the factual path; 5) Drive a visual feedback device to synchronize and visually distinguish the factual path and the counterfactual path on the interactive surface for the user to observe, compare and learn.

[0017] This invention provides a personalized sports training guidance system with intelligent sensing and feedback. It has the following beneficial effects: 1. This invention significantly enhances the intuitiveness of training feedback and improves user learning efficiency. The data processing and analysis module generates factual and counterfactual paths, which are then synchronously visualized and replayed on the interactive surface by the immersive feedback module. Users no longer receive abstract instructions or numerical values, but can directly and instantly observe the dynamic difference between their actual action trajectory and the ideal trajectory derived from their personalized baseline model. This intuitive and comparable feedback allows users to understand the key points of the action more quickly and make targeted corrections. 2. This invention provides an active and forward-looking training guidance method, which helps to optimize movement patterns and avoid potential injury risks. The predictive intent engine analyzes the interaction field data sequence before the macro-action occurs, and can identify the user's intent in advance when the user performs a biased preparatory action. It also drives the visual feedback device to provide feedforward early warning feedback, guiding the user to make adjustments before the action is executed, thereby avoiding the occurrence of incorrect actions and realizing the transformation from post-correction to pre-prevention. 3. This invention provides a non-invasive and more comprehensive motion dynamics analysis solution, balancing user comfort with in-depth data analysis. By employing a dynamic interactive field sensing module integrating pressure and vibration sensors, it eliminates reliance on wearable devices and avoids interference with the user's natural movements. Simultaneously, by fusing and analyzing the pressure distribution matrix and multi-channel vibration time-domain signals, the system can not only assess the user's balance stability but also quantify dynamic texture information such as motion impact force and cushioning techniques, which are difficult to perceive with traditional visual solutions, thus providing a deeper and more complete evaluation of the user's motion performance. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the functional modules of a personalized sports training guidance system with intelligent sensing and feedback according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a dynamic interactive field sensing module according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the internal functional division of the data processing and analysis module according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the personalized model evolution process according to an embodiment of the present invention; Figure 5 This is a flowchart illustrating a personalized sports training guidance method based on intelligent sensing and feedback, according to an embodiment of the present invention.

[0019] Among them, 10. Dynamic interactive field perception module; 11. Smart training mat; 111. Pressure sensor; 112. Vibration sensor; 12. Depth camera; 13. Central control unit; 20. Data processing and analysis module; 21. Predictive intent engine; 30. Immersive feedback module; 40. Personalized model evolution module. Detailed Implementation

[0020] To enable those skilled in the art to more clearly understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The following is in conjunction with the appendix Figure 1 -Appendix Figure 5 The present invention will be described in detail below.

[0022] See attached document Figure 1 , Figure 1This is a functional module diagram of a personalized sports training guidance system with intelligent perception and feedback according to an embodiment of the present invention. The present invention provides a personalized sports training guidance system with intelligent perception and feedback, which constructs an intelligent closed loop from perception, analysis, feedback to evolution. Specifically, it may include: a dynamic interactive field perception module 10, a data processing and analysis module 20, an immersive feedback module 30, and a personalized model evolution module 40.

[0023] The dynamic interactive field sensing module 10 serves as the system's data input source, continuously collecting multi-dimensional, high-fidelity dynamic interactive data generated between the user and a specific interactive surface during physical training. This data constitutes a dynamic dataset capable of fully describing the user's movement process. In this embodiment, the interactive data specifically includes a real-time changing pressure distribution matrix. Vibration time-domain signals of one or more channels .

[0024] The data processing and analysis module 20 communicates with the dynamic interactive field sensing module 10 to perform in-depth processing and analysis of the received raw data stream. This module has a dual function: First, the action data that has already occurred is analyzed to extract key dynamic features that characterize the user's balance stability and action quality. The core of this is to calculate the actual path of the user's movement. Secondly, instead of passively waiting for the action to complete, it actively analyzes the sequence of subtle signals before the action occurs in order to predict the user's intention to perform the action.

[0025] The immersive feedback module 30 serves as the system's output execution terminal and is connected to the data processing and analysis module 20. This module is used to transform the abstract information produced by the analysis module into visual signals that users can intuitively understand.

[0026] On the one hand, it provides feedforward early warnings based on the predicted movement intentions; On the other hand, it provides a visual replay of the factual path of a user's actual behavior alongside the counterfactual path of their ideal behavior.

[0027] This feedback is presented directly on the user's interactive surface, achieving spatial unity between information presentation and user attention.

[0028] The personalized model evolution module 40, coupled with the data processing and analysis module 20, is crucial for the system's personalized and adaptive capabilities. This module internally stores and maintains a personalized baseline model for each user. The data processing and analysis module 20 uses this model as a reference when generating counterfactual paths. More importantly, after a user completes a training exercise, the evolution module 40 evaluates the quality of their performance and, based on a preset update strategy, decides whether to use this high-quality exercise data to optimize and iterate the user's baseline model. This allows the system's guidance capabilities to dynamically improve as the user's skill level increases.

[0029] In the entire system's workflow, these four modules form a complete closed loop of information flow.

[0030] First, the dynamic interaction field perception module 10 captures the user's original action information; Subsequently, the data processing and analysis module 20 decodes and infers the information, generating analysis results and prediction conclusions; Next, the immersive feedback module 30 provides these conclusions to the user in a visual form; Finally, users adjust their actions based on feedback, generating new interaction data. At the same time, their high-quality action data is also learned by the personalized model evolution module 40, thereby improving the accuracy of the next analysis and feedback.

[0031] This cycle enables the system to continuously, intelligently, and personally guide the user's training process.

[0032] See attached document Figure 2 , Figure 2 This is a schematic diagram of the structure of a dynamic interactive field sensing module according to an embodiment of the present invention. In this embodiment, the dynamic interactive field sensing module 10 is the foundation of the entire system's sensing layer, which captures rich, multi-dimensional physical interaction information between the user and the training field in a non-intrusive manner.

[0033] In this embodiment, the core of the dynamic interactive field sensing module 10 is a smart training pad 11 serving as an interactive surface. Physically, the smart training pad 11 comprises a flexible substrate material with good elasticity and durability, and internally integrates or attaches a plurality of sensors in a preset array. Specifically, a uniformly distributed array of sensors is located beneath the surface of the smart training pad 11. A sensor array consisting of individual pressure sensors 111. These pressure sensors 111 can be piezoelectric or piezoresistive sensors, and they are sensitive to the vertical force applied to them, thus collectively forming a real-time pressure distribution matrix. : ; in, Represents a point in time Located at the The pressure values ​​measured by the pressure sensor 111 at each location.

[0034] In addition to the array of pressure sensors 111 for static and quasi-static pressure sensing, the smart training mat 11 also integrates one or more high-sensitivity vibration sensors 112, such as triaxial accelerometers, at several key locations. These vibration sensors 112 are specifically designed to capture transient mechanical waves propagating within the interactive surface when the user performs high-dynamic actions such as jumping, landing, and sudden stops. Their output signals are multi-channel vibration time-domain signals. This signal is crucial for subsequent analysis of the impact quality and buffering techniques of user actions.

[0035] To enhance the robustness of the data source and accurately attribute the perceived data to the target user, the dynamic interactive field perception module 10 in this embodiment may further include a top-mounted depth camera 12 as an auxiliary component. It should be specifically noted that the function of this depth camera 12 is not to perform high-precision 3D human pose reconstruction, but rather to acquire the approximate outline and spatial center of the user on the smart training mat 11 with lower spatial resolution. Its primary use lies in data association; when multiple targets or environmental interference exist on the training ground, the system can utilize... The information collected by sensors 111 and 112 will be accurately linked to the target user, thereby effectively filtering out irrelevant signals.

[0036] During data acquisition, the central control unit 13 is used for the synchronous acquisition of all data streams. This unit uses a unified high-frequency clock to synchronously trigger data sampling from the pressure sensor array 111, vibration sensor 112, and depth camera 12, and assigns a common timestamp to each frame of data. This strict time synchronization mechanism is the fundamental guarantee that the subsequent data processing and analysis module 20 can accurately perform multimodal data fusion, causal relationship inference, and event alignment.

[0037] See attached document Figure 3 , Figure 3 This is a schematic diagram illustrating the internal functional division of the data processing and analysis module 20 according to an embodiment of the present invention. In this embodiment, when the system detects the start and end of a complete macroscopic action (such as a squat or a jump), the module initiates the extraction process of its motion performance dynamics characteristics. This process aims to transform the raw, high-dimensional interaction data received from the dynamic interaction field perception module 10 into a set of low-dimensional key performance indicators that can quantify the execution quality of the action.

[0038] A key aspect of this process is generating a factual path that characterizes the user's actual balance control ability. In this embodiment, this factual path is specifically the trajectory of the user's center of pressure (CoP) during the execution of an action. The calculation of the CoP trajectory is entirely based on the pressure distribution matrix acquired by the pressure sensor array 111. At any given time The two-dimensional position of CoP in the interactive surface coordinate system It is obtained by weighted averaging of the readings from all sensors based on their respective position coordinates. The specific calculation formula is as follows: ; ; In this group of formulas It is the first A pressure sensor 111 at time point Pressure reading; It is the first The pressure sensor 111 has a pre-calibrated, fixed position coordinate in the coordinate system of the smart training mat 11; This refers to the total number of pressure sensors 111. This is determined by analyzing a complete operating cycle (from time...). arrive By continuously calculating the CoP positions within the range, the complete trajectory can be obtained. The smoothness, length, and offset range of the trajectory are direct indicators for evaluating the user's balance stability and will serve as important inputs for generating immersive feedback.

[0039] Another core aspect of this process is the quantitative analysis of the impact quality of the user's actions, which is crucial for assessing the skill and safety of high-risk maneuvers such as jumping and landing. This analysis is based on multi-channel vibration time-domain signals acquired by vibration sensor 112. When the system detects an impact event (e.g., by detecting a sharp change in pressure signal), the data processing and analysis module 20 will analyze the vibration signals within a short time window before and after the impact. (This is) The system performs spectral analysis on one or more channels (through fusion). Specifically, it applies the Fast Fourier Transform (FFT) algorithm to calculate the frequency domain representation of the time-domain signal. Its theoretical basis is the Fourier transform: ; in, It is a vibration signal. It is time. It's frequency. It is the imaginary unit. From the calculated spectrum... In the subsequent analysis, the system further extracts a series of quantitative features, such as peak frequency, spectral centroid, and the energy proportion of high-frequency bands (e.g., frequency ranges above a preset threshold). These spectral features effectively reveal the nature of the impact force: a smooth, well-buffered landing will generate vibrational energy concentrated in the low-frequency region, while a stiff, high-risk landing will excite more high-frequency vibrations. These spectral features, together with the CoP trajectory features, constitute a comprehensive evaluation of the impact.

[0040] Unlike the retrospective analysis of completed actions described above, the data processing and analysis module 20 of this invention further includes a forward-looking predictive intent engine 21. This engine is designed to overcome the lag inherent in traditional motion analysis systems. Instead of summarizing after an error occurs, it anticipates the user's upcoming motion intent by interpreting subtle preparatory movements before a complete macroscopic action is executed, thus providing a basis for decision-making in achieving feedforward proactive guidance.

[0041] The predictive intent engine 21 analyzes not the data of the entire action cycle, but a sequence of interaction field data within a very short preset time window (e.g., hundreds of milliseconds) before the macro-action occurs significantly. This data sequence is constructed into a feature vector, which contains high-value information that can predict impending action. In this embodiment, the feature vector specifically integrates two key features: one is the time rate of change of the pressure distribution matrix. The first is the unconscious pre-adjustment of the center of gravity by the user in order to initiate the action; the second is the micro-vibration pattern in the multi-channel vibration time domain signal, which reveals the pre-activation and isometric contraction of specific muscle groups to stabilize the joint before exerting force.

[0042] The engine operates based on a probabilistic inference framework. Internally, the engine maintains a... A set of predefined candidate motion intentions Examples include "preparing to squat," "preparing to move left," and "preparing to jump." When capturing a sequence of interactive field data... At that time, the engine's goal is to calculate which candidate intent has the highest posterior probability given the observed data. Based on Bayes' theorem, the predicted intent... Determined by the following formula: ; In this formula, the symbols are defined as follows: It is the final predicted movement intention.

[0043] It is the first The intention of each candidate movement.

[0044] It is an intention The prior probability. This probability can be set based on the user's training plan or historical action frequency, or it can be set to a uniform distribution when there is no specific information.

[0045] It is the likelihood function, which represents the likelihood of a given user intent. Under these conditions, the current specific data sequence is observed. The probability of.

[0046] The effectiveness of this inference framework hinges on the likelihood function. The accuracy of this function is ensured. It is not a fixed, generalized model, but rather learned and personalized from the user's own historical motion data. Specifically, during system initialization calibration or daily use, when a user performs a known, tagged action (e.g., a standard jump), the system extracts the interaction field data sequence from the short period preceding the action and uses it as training samples to update the "preparing to jump" intention. The associated probability distribution model (e.g., a Gaussian mixture model or a hidden Markov model). In this way, the predictive intent engine 21 can learn the user's unique "signal fingerprint" when preparing to perform a specific action, thus achieving highly personalized and accurate intent prediction. The output of the engine, i.e., the predicted intent... This will be immediately passed to the immersive feedback module 30.

[0047] The system extracts dynamic features from the user's current motion performance and obtains their actual CoP trajectory, i.e., the factual path. Subsequently, the data processing and analysis module 20 performs another function: generating an idealized counterfactual path for comparison. This counterfactual path is not a unified, universal standard applicable to all users, but rather a realistically achievable personalized optimization goal generated entirely based on the user's own historical best performance.

[0048] This functionality relies on a personalized baseline model within the system. This model is stored and maintained independently for each registered user, its physical repository located in the personalized model evolution module 40, but its invocation and application are completed within the data processing and analysis module 20. Essentially, this model is a structured knowledge base that stores information about the user's performance of various training actions (e.g., action type). When performing a high-quality action, the idealized dynamic pattern corresponds to that action. Specifically, for each action type... The model stores a set of parameters to characterize the ideal CoP trajectory shape. This set of parameters defines a normalized trajectory in the normalized time domain. .

[0049] When it is necessary to generate a counterfactual path, the data processing and analysis module 20 first identifies the type of action the user has just completed. (Based on the output of the predictive intent engine 21 or the fact path) (This is determined through morphological analysis), and then the corresponding paradigm trajectory is retrieved from the personalized model. This paradigm trajectory Describes the action in normalized time. The ideal CoP path shape during the transition from 0 to 1.

[0050] Subsequently, the module concretizes this abstract paradigm trajectory, generating a counterfactual path that is spatiotemporally aligned with the current factual path. A direct generation method is to translate the starting point of the paradigm trajectory to align it with the starting point of the fact path. If the fact path occurs within a time interval... [ t start , t end ] Within this interval, for any time within that interval... Points on the counterfactual path It can be calculated using the following formula: in: At any moment The counterfactual CoP position.

[0051] The starting position of the fact path serves as the reference point for alignment.

[0052] It was obtained from the personalized baseline model at normalized time. The position of the paradigm CoP at that location.

[0053] It is a time normalization function that normalizes absolute time. Mapped to relative normalized time .

[0054] This method ensures that the start point and duration of the counterfactual path remain consistent with the factual path, making the comparison between the two paths intuitive. In more complex implementations, the system can also employ non-linear alignment algorithms such as Dynamic Time Warping to better match the rhythm of the paradigm trajectory with the rhythm of the factual trajectory, thereby generating an ideal reference that is more closely aligned with the user's current state in both form and temporal dynamics. Ultimately, this symbiotic, spatiotemporally precisely aligned pair of factual and counterfactual paths will be transmitted as a data pair to the immersive feedback module 30 for subsequent visualization.

[0055] This embodiment describes the specific implementation of the immersive feedback module 30 and the personalized model evolution module 40 that collaborates with it. These two modules together constitute the "action" and "learning" ends of the system, used to effectively transmit the results of data analysis to the user and to realize the system's self-evolution using the user's feedback data.

[0056] The immersive feedback module 30 operates in a dual-state mode, providing both pre-emptive, feedforward alerts and post-event, counterfactual playback. The pre-emptive alert function is triggered by the output of the predictive intent engine 21. This engine responds when it predicts the user's movement intent. When the system determines that the interaction field data sequence during the preparation phase deviates significantly from the health pattern stored in the personalized baseline model (e.g., excessive swaying of the center of gravity during the preparation for jumping), the immersive feedback module 30 immediately activates the visual feedback device to provide an immediate warning before the user completes the action. In this embodiment, the warning can be specifically manifested as a flashing red circle projected onto the smart training mat 11 around the user's current center of gravity pressure point, thereby indicating to the user that their preparation posture is unstable and needs to be adjusted first.

[0057] In terms of hardware implementation, the warning feedback can be accomplished through a controlled visual presentation device, which can be a controllable light-emitting unit array integrated inside the smart training pad, or an external projector capable of projecting dynamic graphics onto the surface of the training pad.

[0058] The replay function is activated after the user completes a full training session. At this time, the immersive feedback module 30 receives a pair of spatiotemporally aligned path data from the data processing and analysis module 20: the factual path. Counterfactual path Subsequently, the module drives a visual feedback device (e.g., a ground projector or a controllable light-emitting array integrated into the training mat) to synchronously replay the two paths on the smart training mat 11 in the form of dynamic light effects. To achieve clear visual differentiation, the two paths are assigned different visual attributes; for example, the factual path is rendered as a red light trail, while the counterfactual path is rendered as a green light trail. Users can intuitively observe, like watching a video replay, where and how their actual trajectory (red) deviates from their ideal trajectory (green), thus gaining clear and actionable directions for improvement.

[0059] See attached document Figure 4 , Figure 4 This is a schematic diagram of a personalized model evolution process according to an embodiment of the present invention. The system's personalization and adaptive capabilities are realized by the personalized model evolution module 40. This module ensures that the "counterfactual path" provided by the system is always a challenging yet achievable goal that matches the user's current ability level. The implementation of this module includes two stages: model initialization and calibration, and continuous adaptive updates.

[0060] When a user first uses the system, the personalized model evolution module 40 guides the user through an initial calibration process. During this process, the user needs to follow the demonstration and complete a series of basic training actions as accurately as possible. The module records the complete interaction data of these high-quality actions and extracts the paradigm trajectory corresponding to each action type. And related dynamic characteristics, thereby constructing an initial personalized baseline model for the user.

[0061] In subsequent regular training, the adaptive update mechanism is activated. Each time the user completes an action, the personalized model evolution module 40 calculates a comprehensive motion performance score based on the dynamic characteristics obtained from the data processing and analysis module 20 (e.g., the stability of the CoP trajectory, the energy distribution of the vibration spectrum, etc.). This score can be quantified using the following formula: ; in, Representing the Various dynamic characteristics (such as trajectory smoothness, impact force magnitude, etc.). It is a function that normalizes the score for this feature, and This is its corresponding weight. The module will then use this score. With a dynamically updated threshold Compare. The threshold. It can be set to a high percentile (e.g., 90th percentile) of a user's recent scores for similar actions. Only when... Only when this happens will the system determine that the action is a breakthrough, high-quality performance, and use the interaction data from this action to update the corresponding paradigm trajectory and features in the personalized baseline model. This mechanism ensures that the model only evolves in a better direction, thereby achieving adaptive guidance that grows together with the user.

[0062] See attached document Figure 5 , Figure 5 This is a flowchart of a personalized sports training guidance method based on intelligent sensing and feedback according to an embodiment of the present invention. The method organically integrates the various functional modules of the aforementioned system into a coherent, closed-loop guidance process, specifically including the following steps: S501, System Initialization and Calibration of User Personalization Model.

[0063] The system performs this initialization step when a user first uses the system or when a new user is added. The user first completes identity registration, and then the system guides the user into calibration mode. In this mode, the user needs to perform a set of preset basic movements (e.g., standard squat, vertical jump, lateral slide, etc.) as accurately as possible, following on-screen or voice prompts. During this process, the personalized model evolution module 40 captures and processes the complete interactive data of these high-quality demonstration movements, extracts key dynamic patterns, and uses this to build an initial version of the personalized baseline model for the user. This model includes the paradigmatic trajectory and ideal feature parameters for each movement.

[0064] S502, Start real-time training and continuously collect interactive field data.

[0065] Once initialization is complete, the user can begin formal training. Once the user steps onto the smart training mat 11 and starts the training program, the dynamic interactive field sensing module 10 begins working continuously at a high sampling rate, acquiring the pressure distribution matrix generated between the user and the interactive surface in real time. and multi-channel vibration time-domain signals This forms a continuous data stream, which is then transmitted to the data processing and analysis module 20.

[0066] S503 predicts movement intentions and provides feedforward early warning feedback.

[0067] Before a user performs any specific, macroscopic training action, the predictive intent engine 21 in the data processing and analysis module 20 analyzes the interaction field data sequence within a short, preset time window. If the engine successfully predicts the user's upcoming movement intent (e.g., "preparing to jump") based on probabilistic inference, and simultaneously determines that the dynamic characteristics of the user's ready posture (e.g., the amplitude of a slight sway in the center of gravity) deviate significantly from the health pattern in the personalized baseline model, then the immersive feedback module 30 is immediately triggered to project a warning visual signal under the user's feet before the user jumps, guiding the user to adjust and stabilize beforehand.

[0068] S504. Execute the training actions and generate fact paths.

[0069] After receiving (or not receiving) warning feedback, the user continues to perform their training actions. Throughout the entire process of the action occurring and being completed, the data processing and analysis module 20 records complete interaction data and analyzes it based on the stress distribution matrix. The continuous changes in the center of gravity (CoP) are calculated in real time through weighted averaging to generate a trajectory representing the user's actual balance control performance; this is the actual path. .

[0070] S505: Invoke the baseline model and generate a counterfactual path.

[0071] In the factual path Upon completion of the generation, the data processing and analysis module 20 will retrieve the corresponding paradigm trajectory from the personalized baseline model based on the identified action type, and after spatiotemporal alignment transformation, generate a trajectory representing the ideal performance that the user could have achieved in that action; this is the counterfactual path. .

[0072] S506, synchronous visual playback, providing comparative feedback.

[0073] After receiving the paired factual and counterfactual path data, the immersive feedback module 30 immediately activates the visual feedback device to synchronously and visually replay the complete dynamic process of the two paths on the ground of the smart training mat 11 using light effects with significant visual differences (e.g., different colors). By observing the real-time separation and overlap of the two light trails, the user can intuitively understand the deviation in their actions.

[0074] S507. Evaluate athletic performance and decide whether to evolve the model.

[0075] After the feedback is complete, the personalized model evolution module 40 will perform a comprehensive quantitative score on the action just completed. It will calculate a comprehensive motion performance score S and compare it with the user's historical performance score threshold for this action. The system compares the scores. If the score S exceeds the threshold, the action is considered a high-quality, breakthrough performance. The system then uses the data from this action to update and optimize the corresponding paradigm trajectory in the personalized baseline model, thereby improving the model's performance. If the score does not exceed the threshold, the model remains unchanged.

[0076] After this, the entire process returns to step S502, and the system repeats the loop from S502 to S507 for the user's next action until the user finishes the current training. Through this series of cyclical steps, the method of the present invention realizes a personalized training guidance closed loop that can continuously perceive, intelligently analyze, intuitively provide feedback, and adaptively improve.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A personalized sports training guidance system with intelligent sensing and feedback, characterized in that, include: The dynamic interactive field sensing module is configured to collect interactive data generated when a user moves on the interactive surface. The interactive data includes a pressure distribution matrix and a multi-channel vibration time-domain signal. The data processing and analysis module, connected to the dynamic interactive field sensing module, is configured to: The interaction data is processed to determine the factual path on the interaction surface that represents the user's actual motion performance. A preset personalized baseline model that stores the user's ideal motion performance pattern is invoked to generate a counterfactual path on the interaction surface that corresponds to the factual path. An immersive feedback module, connected to the data processing and analysis module, is configured to receive the factual path and the counterfactual path, and drive a visual feedback device to synchronize and visually distinguish the factual path and the counterfactual path on the interactive surface. The data processing and analysis module further includes a predictive intent engine, which is configured to: Before a user initiates a macroscopic action, analyze the sequence of interaction field data composed of the interaction data within a preset time window; Based on the interaction field data sequence and according to the preset probability model, predict the user's next movement intention; The immersive feedback module is also configured to: After the predictive intent engine predicts the motion intent, and before the user completes the corresponding action, if it is determined that the preparation mode of the motion intent deviates from the health mode in the personalized baseline model, the visual feedback device is driven to provide feedforward warning feedback. The interactive field data sequence includes the time rate of change of the pressure distribution matrix and the micro-vibration patterns of the multi-channel vibration time-domain signal. The predictive intent engine predicts the motion intent by identifying specific combinations of the time rate of change of the pressure distribution matrix and the micro-vibration patterns.

2. The personalized sports training guidance system with intelligent sensing and feedback according to claim 1, characterized in that, The dynamic interactive field perception module also includes a top-mounted depth camera, which is configured to acquire spatial position data of the user on the interactive surface in order to associate the interactive data with the user entity.

3. The personalized sports training guidance system with intelligent sensing and feedback according to claim 1, characterized in that, The factual path is the user's center of gravity pressure point trajectory. The data processing and analysis module determines the center of gravity pressure point trajectory by performing a weighted average calculation on the pressure distribution matrix.

4. The personalized sports training guidance system with intelligent sensing and feedback according to claim 1, characterized in that, The data processing and analysis module is also configured to perform spectral analysis on the multi-channel vibration time-domain signal to extract the impact and vibration spectral characteristics of the user's motion impact force, and to use the impact and vibration spectral characteristics as one of the bases for evaluating the user's motion performance and updating the personalized baseline model.

5. The personalized sports training guidance system with intelligent sensing and feedback according to claim 1, characterized in that, The visual feedback device is a ground projector or a controllable light-emitting array integrated into the interactive surface. The immersive feedback module synchronously replays the factual path and the counterfactual path with dynamic lighting effects of two different colors or line types to provide a visual reference for direct comparison by the user.

6. The personalized sports training guidance system with intelligent sensing and feedback according to claim 1, characterized in that, It also includes a personalized model evolution module, which is configured to: After a user completes a movement, a score for the movement performance is calculated based on the actual path and the multi-channel vibration time-domain signal. When the performance score of the movement exceeds the dynamic update threshold, the personalized baseline model is updated using the interaction data of this movement.

7. A personalized sports training guidance method based on intelligent sensing and feedback, using the system described in any one of claims 1-6, characterized in that, Includes the following steps: The dynamic interactive field sensing module collects interactive data generated when the user moves on the interactive surface. The interactive data includes a pressure distribution matrix and multi-channel vibration time domain signals. Before a user initiates a macroscopic action, the interaction field data sequence composed of the interaction data within a preset time window is analyzed to predict the user's next movement intention; The interaction data is processed to determine the factual path on the interaction surface that represents the user's actual motion performance. A preset personalized baseline model that stores the user's ideal motion performance pattern is invoked to generate a counterfactual path on the interaction surface that corresponds to the factual path. A visual feedback device is driven to synchronize and visually distinguish the factual path and the counterfactual path on the interactive surface for a visually distinct playback.

Images