Exercise training management system based on multi-sensor fusion

Abstract

The invention relates to the field of medical care informatics, and discloses an exercise training management system based on multi-sensor fusion, which is characterized in that the system acquires calibration data to establish an individualized exercise physiological response baseline model, inputs actual exercise performance into the baseline model in training, and determines expected physiological signs; calculating a time sequence deviation value between the expected physiological sign and the actual physiological sign; when it is monitored that the athletic performance declines, attribution is carried out based on a time sequence deviation value, and the athletic performance declines are judged to be derived from conventional fatigue or potential physiological health risks, by introducing a time sequence deviation attribution mechanism, the technical problem that in the prior art, due to data isolation, the athletic performance decline causes cannot be judged is solved, and the probability that the athletic performance declines is lowered is lowered. Therefore, the method has the capability of distinguishing conventional fatigue from potential health risks, and decision conflicts between invalid exercise guidance and health risk early warning are avoided.

Description

Technical Field

[0001] This invention relates to a sports training management system based on multi-sensor fusion, belonging to the field of healthcare informatics technology. Background Technology

[0002] Currently, collecting users' physiological and athletic performance data through various sensors, such as smart wearable devices or intelligent training equipment, has become a common technological approach. These systems typically aim to provide feedback to users or coaches using the collected data. However, those skilled in the art have noted that existing information processing systems generally tend to treat users' athletic performance data streams and their physiological data streams as two independent and parallel information sources. The system analyzes these two types of data separately and provides conclusions on athletic assessment or physiological monitoring for each. This data-isolated processing architecture has inherent limitations at the information processing level. Specifically, when a decline in user athletic performance is detected, the system cannot accurately determine, from an information processing perspective, whether the decline is caused by routine exercise fatigue or by potential physiological health risks. This limitation in information processing has not been effectively addressed in some existing systems that integrate multiple sensors.

[0003] For example, Chinese invention patent CN118319571A discloses a knee joint load reduction orthopedic system based on multi-sensor information fusion. Although the system also uses posture sensors, electromyography (EMG) sensors, and pressure sensors, its information processing approach is still limited to fusing kinematic posture data with physiological EMG data to output an assessment score (such as a WOMAC score) or monitoring EMG, pressure, and other data independently. This architecture does not essentially establish an individualized coupling benchmark between motor performance and physiological response. When abnormal motor performance is detected, it is also unable to fundamentally distinguish whether the abnormality is due to muscle fatigue or other potential pathological risks, resulting in a lack of clear attribution basis for its intervention instructions, such as training feedback.

[0004] Therefore, the technical problem to be solved by this invention is how to establish a unique coupling relationship between individual motor performance and physiological response in information processing methods, and use this model to analyze the correlation between changes in motor performance and deviations in physiological state in real time, so as to enable the system to distinguish between routine fatigue and potential health risks, and avoid conflicting intervention instructions caused by isolated data processing. Summary of the Invention

[0005] This invention provides a sports training management system based on multi-sensor fusion. Its main purpose is to solve the problem that existing technologies, due to the isolation of sports and physiological data processing, cannot accurately determine the causes of declining sports performance, thus causing conflicts between sports guidance and health warning decisions.

[0006] To achieve the above objectives, the present invention provides a sports training management system based on multi-sensor fusion, including a sensor fusion module, an individualized baseline model construction module, an expected physiological response determination module, a time-series deviation calculation module, and a health risk attribution module;

[0007] The sensor fusion module is used to acquire calibrated motion performance data and calibrated physiological sign data of the user in the calibration state, as well as actual motion performance data and actual physiological sign data of the user in the training state.

[0008] The individualized baseline model construction module, connected to the sensor fusion module, is used to establish an individualized exercise physiological response baseline model that characterizes the coupling relationship between exercise performance and physiological response based on calibrated exercise performance data and calibrated physiological sign data, using a preset baseline model algorithm.

[0009] The expected physiological response determination module connects the sensor fusion module and the individualized baseline model construction module. It is used to input actual exercise performance data into the individualized exercise physiological response baseline model to determine the expected physiological signs data.

[0010] The timing deviation calculation module connects the sensor fusion module and the expected physiological response determination module, and is used to calculate the timing deviation value based on the actual physiological sign data and the expected physiological sign data.

[0011] The health risk attribution module connects the sensor fusion module and the time-series deviation calculation module. When a decline in athletic performance is detected based on actual athletic performance data, the module determines whether the decline in athletic performance is attributable to routine exercise fatigue or potential physiological health risks based on the comparison between the time-series deviation value and a preset attribution threshold, and outputs corresponding instructions.

[0012] Preferably, the calibration data and actual performance data include at least one of the following: strength data, speed data, power output sequence, movement trajectory data, force exertion pattern data, or plantar pressure distribution data collected by intelligent training equipment; the calibration data and actual physiological signs data include at least one of the following: heart rate sequence, skin conductance sequence, body temperature data, or bioelectrical signal data collected by intelligent wearable equipment.

[0013] Preferably, the preset baseline model algorithm includes: a multivariate time series regression model or a dynamic time warping algorithm. The individualized baseline model construction module is specifically used to: obtain calibrated exercise performance data and calibrated physiological sign data of the user during a set of standardized training actions in a confirmed healthy state, and train the model using the preset baseline model algorithm to obtain an individualized exercise physiological response baseline model.

[0014] Preferably, the health risk attribution module is specifically used to: determine the preset attribution threshold as the first threshold; if the time series deviation value is within the first threshold, determine that the decline in athletic performance is attributable to routine exercise fatigue and output an exercise guidance instruction; if the time series deviation value exceeds the first threshold, determine that the decline in athletic performance is attributable to potential physiological health risks and output a health risk warning instruction.

[0015] Preferably, the time-series deviation calculation module is specifically used to: at any given time point, obtain the actual physiological response value of the actual physiological sign data and the expected physiological response value of the expected physiological sign data; and calculate the time-series deviation value. Timing deviation value Determined by the following formula: ,in This represents the actual physiological response value. This represents the expected physiological response value.

[0016] Preferably, the intelligent training equipment includes at least one of the following: intelligent human target, intelligent biomechanical insole, intelligent wall-mounted reaction training system or digital comprehensive strength trainer, and the intelligent wearable equipment includes smartwatch or intelligent armband.

[0017] Preferably, the intelligent human target has an embedded array of force sensors to measure striking force and speed; the intelligent biomechanical insole integrates a flexible pressure sensor network to capture plantar pressure distribution and gait cycle; and the digital integrated strength trainer uses a servo motor resistance system to record force-speed curves and power output.

[0018] Preferably, the exercise guidance instructions include: instructions to prompt the user to rest between sets or adjust the training intensity; health risk warning instructions include: high-priority instructions to prompt the user to immediately stop training, check physiological signs, or seek medical assistance; the health risk attribution module is also used to: prevent the output of exercise guidance instructions when it is determined that the risk is attributable to a potential physiological health risk.

[0019] Preferably, the sensor fusion module is also used to acquire user limb data collected by the visual inspection device; the individualized baseline model construction module is also used to assist in establishing an individualized motion physiological response baseline model based on the user limb data.

[0020] Preferably, the system also includes a health record management module, which is used to: store individualized exercise physiological response baseline models; and continuously record time-series deviation values, using the time-series deviation values ​​as a dimension of dynamic health information for long-term health trend analysis or early risk identification.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] 1. By calibrating the individualized response benchmark between a user's athletic performance and physiological signs under healthy conditions, and using this benchmark to calculate the expected physiological response corresponding to the actual athletic performance in subsequent training, and then comparing the expected response with the measured physiological response; this information processing method enables the system to distinguish between two similar phenomena (decline in athletic performance) and the different underlying causes (routine physical fatigue or potential physiological abnormalities), thus avoiding ineffective or even harmful intervention instructions due to the inability to accurately attribute causes.

[0023] 2. Based on the ability to attribute the causes of declining athletic performance, a dynamic decision-making switching mechanism is established in the system's feedback logic. When the deviation between athletic performance and physiological response is judged to be within the normal range, exercise technique guidance is executed. When the deviation is judged to exceed a specific range, the intervention instruction is switched to a high-priority health risk warning. This avoids potential conflicts between the two instruction goals of exercise guidance and health protection from the information processing flow, ensuring that the intervention output by the system always points to the correct management goal that matches the user's current state.

[0024] 3. The individualized response benchmark model established and stored in this invention, along with the temporal deviation data between the expected and actual physiological responses continuously generated during training, together constitute a dimension of personal health information. It is no longer a simple record of isolated exercise parameters or physiological indicators, but rather a dynamic reflection of the coupling relationship and stable state of an individual's internal physiological regulatory mechanisms under specific loads. This provides far more profound and effective informatics support than static data for long-term health trend analysis, early risk identification, and quantitative assessment of the rehabilitation process. Attached Figure Description

[0025] Figure 1 This is a diagram of the multi-sensor fusion data processing architecture of the system of the present invention;

[0026] Figure 2 This is a schematic diagram illustrating the long-term monitoring of timing deviation values ​​and the triggering of model updates in this invention;

[0027] Figure 3 This is a use case diagram of training attribution and feedback for the system of this invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining this invention and are not intended to limit the scope of protection of this invention.

[0029] This invention discloses a sports training management system based on multi-sensor fusion, including a sensor fusion module, an individualized baseline model construction module, an expected physiological response determination module, a time-series deviation calculation module, and a health risk attribution module. The sensor fusion module aggregates and synchronizes multi-source heterogeneous data. The individualized baseline model construction module establishes a sports-physiological coupling benchmark representing individual characteristics based on calibration data. The expected physiological response determination module and the time-series deviation calculation module work together to calculate in real time the deviation between the actual physiological state and the expected state of the benchmark model. Finally, the health risk attribution module judges the cause of the decline in athletic performance based on this deviation and outputs decision instructions. In specific engineering implementation, the sensor fusion module is an interface and processing unit with multi-channel data access and clock synchronization functions. It is used to acquire and process two types of core data streams: calibration data of the user in the calibration state and actual data of the user in the training state. The calibration of athletic performance data and actual athletic performance data can originate from intelligent training devices. For example, a digital comprehensive strength trainer uses its servo motor resistance system's built-in sensor to collect the user's power output sequence at a sampling rate of no less than 100Hz. The calibration of physiological characteristic data and actual physiological characteristic data can originate from intelligent wearable devices. For example, a smart armband uses its optical heart rate sensor to synchronously collect the user's heart rate sequence. A key task of this module is ensuring data timestamp alignment. When receiving data from different devices, such as plantar pressure distribution data from smart biomechanical insoles and heart rate sequences from smartwatches, this module uses a unified master clock or time synchronization protocol to calibrate all data streams to the same time reference, providing a data foundation for subsequent coupled modeling. In some implementations, this sensor fusion module can also be used to acquire user limb data collected by visual inspection devices for auxiliary analysis.

[0030] The individualized baseline model building module, connected to the sensor fusion module, has the core function of executing a calibration process to establish an individualized exercise physiological response baseline model. During calibration, this module acquires data on the user performing a set of standardized training movements while in a confirmed healthy state. For example, this involves three sets of standardized hitting exercises using a smart human target, each set lasting one minute. During this calibration process, the sensor fusion module simultaneously collects calibration exercise performance data, such as hitting force and speed sequences, and calibration physiological data, such as heart rate sequences. and skin conductance response sequence This module employs a preset baseline model algorithm, trained using the aforementioned calibration data. Taking a multivariate time series regression model as an example, this module will use calibration data such as power output sequences... Synchronized calibrated physiological data, such as heart rate sequences, will be used as input variables. As the output dependent variable, regression analysis is used to determine the coupling coefficient and time delay relationship between the two; alternatively, a dynamic time warping algorithm is used to calculate the shortest warped path between the two sets of time series data, which itself defines the nonlinear correspondence between changes in exercise performance and changes in physiological response. This module ultimately outputs and stores an individualized exercise physiological response baseline model. The model This objectively characterizes the stable coupling relationship between exercise load and physiological stress for a specific user in a healthy state. In the specific implementation of the individualized baseline model construction module, to ensure the consistency and reproducibility of the initial physiological state used for calibration, the health status is confirmed through an assessment process. This process is used to obtain the user's resting heart rate and heart rate variability (HRV) data and compare them with the user's personal baseline data stored in the health record management module to confirm that the current data is within the preset normal fluctuation range of the baseline. At the same time, combined with the response results of the exercise preparation activity questionnaire (PAR-Q), it is determined that the user has not experienced high-intensity training and has no subjective discomfort symptoms within a specific time window before calibration.

[0031] When the sensor fusion module is also used to acquire user limb data collected by the visual inspection device, the individualized baseline model building module, in its specific implementation, uses this user limb data to assist in establishing an individualized motion physiological response baseline model. One way to assist in establishing this model is by using limb data acquired by the visual inspection device, such as joint angles, range of motion, or posture stability parameters. After quantization, it is compared with calibrated motion performance data, such as power sequences. Together, they serve as the multivariate input variables for the pre-defined baseline model algorithm, used to train an individualized motion physiological response baseline model coupled with action patterns. The model Characterized and The coupling relationship between them; in another implementation, the module uses user limb data to perform normative judgment on movements in a calibrated state, filters out calibrated motion performance data and calibrated physiological sign data corresponding to movement segments that meet preset standards (such as squat depth greater than 90 degrees), and then uses these filtered data subsets to train an individualized motion physiological response baseline model. This improves the accuracy of the baseline model. During the user's regular training, the expected physiological response determination module obtains real-time actual motion performance data from the sensor fusion module, such as the actual work done or power output of the user's current squat. This module then inputs the actual motion performance data into the previously stored individualized exercise physiological response baseline model. ,Model After the calculation, the expected physiological data is output. This data represents the physiological state that the user should achieve after completing this exercise, assuming the user is still in a baseline healthy state. Specifically, it can be the expected physiological response value. Meanwhile, the timing deviation calculation module obtains actual physiological sign data at the same time point (or after considering physiological time delay) from the sensor fusion module, which can specifically be the actual physiological response value. This module then calculates the temporal deviation between the actual physiological data and the expected physiological data. In the specific implementation, the timing deviation calculation module determines the timing deviation value using the following formula. : in, This represents the actual physiological response value. This refers to the expected physiological response value; here... It is an objective indicator that quantifies the deviation of a user's current physiological state from their individual health baseline.

[0032] The health risk attribution module connects the sensor fusion module and the time-series deviation calculation module. Its core function is to attribute the risk based on the time-series deviation value when a decline in athletic performance is detected. Attribution judgment is performed; this module presets an attribution threshold, which is determined as the first threshold in the specific implementation; this first threshold is not a fixed value, but can be determined through the calibration process. By guiding the user to repeat the calibration test in a confirmed state of mild fatigue, such as immediately starting the next set of exhaustion tests after resting between standard sets, the timing deviation value measured in this state is used. The statistical upper limit (e.g., mean plus three standard deviations) is set as the first threshold to define the reasonable deviation range for individual routine exercise fatigue; during training, when this module detects actual exercise performance data from the sensor fusion module (e.g., prolonged reaction time in a reaction training system or decreased power output in a strength trainer), the attribution logic is immediately activated; if the time-series deviation value Within the first threshold, i.e. A smaller value indicates that the actual heart rate is basically consistent with the expected heart rate, thus the decline in athletic performance is determined to be due to routine exercise fatigue, and an exercise guidance instruction is output. This instruction can be a prompt for the user to rest between sets or adjust the training intensity; if the time sequence deviation value is smaller... Exceeding the first threshold, If the value deviates, it indicates that the actual physiological response does not match the expected response under the same exercise performance. Therefore, the decline in exercise performance is determined to be attributable to potential physiological health risks, and a health risk warning instruction is output. This instruction is a high-priority instruction to prompt the user to immediately stop training, check physiological signs, or seek medical assistance. Furthermore, when the health risk attribution module determines that the decline is attributable to potential physiological health risks, it will also proactively prevent the output of exercise guidance instructions, thereby avoiding decision-making conflicts between ineffective exercise guidance and health risk warnings from the information processing flow perspective. To achieve long-term health trend analysis, the system of this invention may also include a health record management module. This module is used to store individualized exercise physiological response baseline models and continuously record time-series deviation values ​​generated during training. This timing deviation value No longer isolated physiological indicators, but dynamically reflecting the stable state of an individual's internal physiological regulatory mechanisms under specific loads, this module can convert time-series deviation values... As a dynamic dimension of health information, it is used for long-term health trend analysis or early risk identification.

[0033] Example 1: This example demonstrates the operation of the technical solution in a specific healthcare informatics monitoring scenario. In a health monitoring scenario for athletes undergoing high-intensity training, a user is using the sports training management system of this invention. The system includes a digital integrated strength training device and a smart armband for collecting heart rate sequences. The system has been calibrated according to the specific implementation procedure. Under the user's confirmed healthy condition, an individualized exercise physiological response baseline model characterizing the coupling relationship between power output and heart rate response has been established. During training, the sensor fusion module monitors the user's actual exercise performance data. Specifically, when performing a set of squats, the user's power output sequence begins to decline compared to previous sets. This decline in performance triggers the analysis logic of the health risk attribution module. The system then performs attribution analysis, and the expected physiological response determination module inputs the actual exercise performance data. The expected physiological characteristics corresponding to this athletic performance are calculated, namely the expected physiological response value. Simultaneously, the timing deviation calculation module obtains the user's actual physiological sign data for the same period from the smart armband, and obtains the actual physiological response value. Then, the timing deviation value is calculated. In this calculation, the timing deviation value It is determined to be within the preset first threshold.

[0034] This result indicates that although users' athletic performance declines, their actual physiological responses match their current achievable athletic performance level; based on this time-series deviation value... If the result does not exceed the first threshold, the health risk attribution module attributes the decline in athletic performance to routine exercise fatigue. The system then outputs exercise guidance instructions, prompting the user to rest between sets and replenish fluids as planned. After the user rests, the next set of training continues. During the exercise, the sensor fusion module detects a non-progressive, significant decrease in the user's actual athletic performance data. The system then executes the attribution analysis process again. Based on this significant decrease in actual athletic performance data, the expected physiological response determination module calculates a new expected physiological response value. However, the timing deviation calculation module obtains the actual physiological response values ​​from the smart armband. Instead of matching the expected level, it exhibited an abnormal state that was severely inconsistent with its athletic performance, leading to an incorrectly calculated timing deviation value. It exceeded the first threshold; given this timing deviation value Upon detecting the deviation, the health risk attribution module determines that the decline in athletic performance is highly likely due to potential physiological health risks beyond routine fatigue. The system's information processing and decision-making logic immediately switch. The health risk attribution module determines that the decline in athletic performance is attributed to potential physiological health risks and immediately blocks the output of any exercise guidance instructions, instead outputting a high-priority health risk warning instruction, prompting the user to immediately stop training and monitor physiological signs. This mechanism, by establishing a coupling benchmark between exercise and physiology and calculating the time-series deviation, enables the system to distinguish between different underlying causes behind two similar phenomena.

[0035] Example 2: This example aims to objectively verify the ability of the sports training management system of the present invention, especially its health risk attribution module, to distinguish the decline in athletic performance caused by different factors. Ten healthy male subjects were selected for the experiment. All subjects had completed standardized calibration tests in advance according to the specific implementation procedure, and individualized baseline models of their exercise physiological responses were established. They also determined their respective first thresholds for distinguishing between routine fatigue and potential risk. These first thresholds were calibrated for all subjects. The mean is at Within the bpm range; the test platform includes an electromagnetic resistance-controlled power bicycle that can precisely adjust resistance and output power data in real time, and a device for synchronously collecting heart rate ( The medical-grade dynamic electrocardiogram (ECG) monitor provides ECG data to provide actual physiological signs and serves as an objective standard for verifying the physiological state of subjects in the experiment. Two systems were set up for comparison: a control group system, which simulates existing technology and can independently monitor and display exercise performance data (power) and physiological sign data (heart rate), providing fixed exercise guidance prompts when a decrease in power is detected; and the prototype system of this invention, which implements all modules, including an individualized baseline model construction module, an expected physiological response determination module, a time-series deviation calculation module, and a health risk attribution module. The experiment was conducted in two phases: the first phase was a routine fatigue test, where subjects were at normal fitness levels. In the first phase, a high-intensity intermittent exhaustion program was performed under both electrolyte and kinetic states until the target power could no longer be maintained, resulting in a decline in actual exercise performance data. The second phase was a simulated health risk test. On another day, the same group of subjects performed a standard thermal exercise dehydration program, reaching a state of mild dehydration (weight loss of about 2%), and then performed the same exhaustion program as in the first phase. This state can trigger cardiovascular system compensatory abnormalities and is a potential physiological health risk. In both phases, the control group system and the sample group system of this invention simultaneously received data from the power bicycle and the electrocardiogram monitor and independently output instructions. Two typical attribution scenarios were observed in the experiment, and their key data are shown in Table 1.

[0036] Table 1: Comparison of Attribution and Command Output of the Two Systems under Different Scenarios

[0037]

[0038] Referring to Table 1, during the fatigue phase, when performance declines (power approximately 191.9W), the actual heart rate ( (approximately 177.9 bpm) and based on Expected physiological response value calculated by the model Similarly, the timing deviation value calculated by the sample system of this invention The value is +4.2 bpm, which falls within the first threshold. Within a certain range (bpm), the health risk attribution module attributes this to routine exercise fatigue and outputs exercise guidance instructions; during the risk phase (dehydration), when exercise performance also declines (power approximately 189.7W), the actual heart rate ( (approximately 191.5 bpm) higher than The model extrapolates the expected physiological response based on the current power. (Approximately 174.8 bpm), the timing deviation value calculated by the sample system of this invention. When the value reached +16.7 bpm, it exceeded the first threshold. The health risk attribution module attributed this to a potential physiological health risk and output a high-priority health risk warning instruction. In the same scenario, the control group system, lacking this attribution mechanism, still output exercise guidance instructions.

[0039] Example 3: This example combines Figures 1 to 3 This describes a motion training management system based on multi-sensor fusion, such as... Figure 1 As shown, the system's data sources include an inertial measurement unit (IMU) for acquiring attitude, angular velocity, and acceleration; a physiological sensor (HRM / EMG) for acquiring physiological signals such as heart rate and electromyography; a plantar pressure sensor for acquiring ground reaction force and distribution; and an optional optical motion capture camera for acquiring high-precision spatial position markers. These multi-source data enter the data preprocessing module for data cleaning, filtering, and time alignment. The multi-sensor information fusion engine performs spatiotemporal domain data alignment and state estimation. This engine obtains training parameters from the training scheme and user database, and its estimation results are sent to the kinematics / dynamics solution module to construct a biomechanical model. The output data of this solution module is used to archive data to the training scheme and user database, and also drives three output modules: a training data visualization module for 3D model reconstruction and data charting; a training report generation module for generating periodic data analysis reports; and a real-time motion feedback module for providing immediate guidance to users or coaches.

[0040] like Figure 2 As shown in the figure, three key metrics are illustrated: the time-series bias value (bpm) varying with the training day X-axis, the 30-day moving average, and a fixed model update trigger threshold of 15 bpm. In this diagram, the 30-day moving average of the time-series bias value shows continuous drift and eventually crosses the set 15 bpm model update trigger threshold. This phenomenon indicates that the user's physiological state may have undergone a systematic change, requiring the triggering of the baseline model update procedure. Figure 3 As shown, users can perform calibration tests to establish individualized baseline models. In daily use, users perform regular training, during which the system performs real-time monitoring and attribution. Based on the real-time attribution results, the system will trigger two types of feedback to the user: outputting exercise guidance instructions or outputting health risk warning instructions.

[0041] Example 4: In the implementation of the individualized baseline model construction module using a multivariate time series regression model as the preset baseline model algorithm, the specific execution steps include: First, acquiring the calibration motion performance data synchronously collected under calibration conditions, using power sequences... For example, compared with calibrated physiological data, using heart rate sequences Taking this as an example, the two sets of sequences are preprocessed with time alignment, denoising, and normalization; the second step is to... and one or more preset time delay terms, such as , etc., as input variables, will As the output dependent variable, the least squares method or gradient descent method is used to fit a regression equation characterizing the temporal coupling relationship between the two, which constitutes the individualized exercise physiological response baseline model. The third step is to store the data. The regression coefficients, time delay parameters, and intercept terms of the model are used for subsequent calculations. In another implementation, when the preset baseline model algorithm is a dynamic time warping algorithm, the specific execution steps of the individualized baseline model construction module include: First, obtaining the values ​​under calibration conditions. and Time series data; second step, calculation and The first step is to calculate the local distance between all pairs of data points, using Euclidean distance as an example, and construct a cumulative distance matrix. The second step is to use a dynamic programming algorithm to find a regular path that passes through this matrix and minimizes the total cumulative distance. The path The path itself defines the nonlinear temporal correspondence between the two sequences. The parameters and related parameters are stored as a baseline model of individualized exercise physiological response. .

[0042] The standardized calibration procedure for the first threshold is used to reproducibly determine the timing deviation value corresponding to routine exercise fatigue in users in an engineering context. The statistical boundaries of this procedure are defined after the user has completed an individualized baseline model of the physiological response to exercise. Execution after construction; the initial state is defined as: the user is in a non-fatigue state with sufficient replenishment of water, electrolytes, and glycogen, using a training device with precise power control, such as a power bicycle, and wearing a heart rate monitoring device with a sampling frequency of no less than 1Hz; the process judgment is quantified as follows: First step, the user uses their maximum oxygen uptake... The first step involves a 5-minute warm-up at 50% of the power level. The second step involves starting with a preset base power of 200W, increasing the power by 25W every 2 minutes, and continuing this incremental load program. The third step involves the system determining that the user has entered the initial stage of routine exercise fatigue when they first report a subjective fatigue level (RPE) of 17 or their heart rate fails to stabilize within one minute at the current power level, exhibiting fluctuations greater than 5 bpm. The system then immediately records the actual exercise performance data for the following 60 seconds. Compared with actual physiological data The parameter determination steps are as follows: The system will collect the group of parameters... As input, substitute the already stored... The model calculates a set of corresponding expected physiological response values. Calculate the time-series deviation value sequence under this fatigue state. Finally, the system... Perform statistical analysis on the sequence and calculate its mean. and standard deviation And the first threshold is determined as The first threshold is determined through this standardized physiological load test.

[0043] Example 5: This example is a specific supplement to the functions of the health record management module, which is used to ensure the individualized exercise physiological response baseline model. To ensure long-term effectiveness, the health record management module continuously records and statistically analyzes the time-series deviation values ​​generated by users during multiple training sessions. The sequence is analyzed, and the statistical mean of the sequence over a preset long-term time window, taking 30 consecutive training days as an example, is calculated. With volatility indicators; when the system detects this A sustained, unidirectional drift exceeding a preset statistical threshold, such as continuously maintaining a value above +5.0 bpm, indicates a systemic change in the user's physiological health or exercise capacity, leading to a disruption of the original... The system can no longer accurately represent the current exercise-physiological coupling relationship. After this systematic drift is confirmed, the system automatically triggers the model update procedure. The system prompts the user through its interactive interface to re-execute the standardized calibration process in the specific implementation method or Example 4 at the next appropriate time point, that is, when the user confirms that they are in a healthy and non-fatigued state. The individualized baseline model construction module then uses the newly collected calibration exercise performance data and calibration physiological sign data to retrain and generate an updated individualized exercise-physiological response baseline model. And accordingly, the updated first threshold is recalculated. The health record management module will integrate the original Archive the first threshold and activate it. and This serves as the analytical benchmark for subsequent training.

[0044] Example 6: In the specific implementation of the sensor fusion module processing multi-source heterogeneous data, to ensure data quality and timing consistency, the module executes a data packet verification and clock synchronization procedure when acquiring force sensor array data from a smart human target and optical heart rate sensor data from a smart wearable device. In this procedure, each data source uses a unified data frame format, which includes data payload, timestamp, and cyclic redundancy check (CRC) code. After receiving a data frame, the sensor fusion module calculates its CRC code and compares it with the check code embedded in the frame. If the comparison is inconsistent, the data frame is determined to have an error during transmission, and the system marks it as a bad frame and discards it to avoid abnormal data contaminating subsequent model calculations. After the data frame passes verification, the sensor fusion module further... The first step involves clock synchronization processing to correct for minute time drifts caused by differences in the internal crystal oscillators of different sensors. This module uses a high-precision network time protocol source or a local master clock to periodically broadcast synchronization signals to all sensor nodes. Upon receiving the signal, the sensor nodes compare their local timestamp with the synchronization time to calculate the time offset. This offset is then used to correct the locally generated timestamp when reporting data in subsequent updates. After aggregating these corrected data, the sensor fusion module unifies them to a global time reference and stores them in a data buffer sorted by timestamp. This buffer is then used by the subsequent individualized baseline model building module or the expected physiological response determination module. This procedure is used to establish a temporal correspondence between motion performance data and physiological characteristic data.

[0045] Example 7: In a lower limb strength training scenario using intelligent biomechanical insoles, the intelligent biomechanical insoles integrate a flexible pressure sensor network to capture the user's plantar pressure distribution data, which is used as part of the actual athletic performance data. When the system detects an abnormal shift in the plantar pressure distribution during three consecutive deadlifts, and this shift is determined to be a decline in athletic performance, the health risk attribution module initiates analysis. The system inputs this pressure distribution data (combined with power data from a digital comprehensive strength trainer) into... The model calculates the expected physiological response value. and compared with the actual physiological response values ​​obtained from the smart armband. Compare; if the calculated timing deviation value If the pressure exceeds the first threshold, the system determines that the abnormal decline in athletic performance is not due to routine fatigue, but may be caused by potential physiological health risks such as lower back compensation, and then outputs a high-priority health risk warning. In another scenario using an intelligent wall-mounted reaction training system, the system is used to train the user's reaction speed, and the measured reaction time series is used as actual athletic performance data. When the system detects that the user's average reaction time in continuous testing has increased by 150 milliseconds, constituting a decline in athletic performance, the health risk attribution module initiates analysis; the system inputs this athletic performance data... Model calculation and the actual measured Compare; if the calculated timing deviation value If the reaction speed is within the first threshold range, the system determines that the decrease is due to routine exercise fatigue or lack of concentration, and outputs exercise guidance instructions to prompt the user to take a rest.

[0046] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A sports training management system based on multi-sensor fusion, characterized by, The system comprises a sensor fusion module, an individualized baseline model construction module, an expected physiological response determination module, a timing deviation calculation module, and a health risk attribution module. The sensor fusion module is configured to acquire calibration performance data and calibration physiological data of a user in a calibration state, and actual performance data and actual physiological data of the user in a training state. The individualized baseline model construction module is connected to the sensor fusion module and is configured to establish an individualized exercise physiological response baseline model representing a coupling relationship between performance and physiological response based on the calibration performance data and the calibration physiological data, and using a preset baseline model algorithm. The expected physiological response determination module is connected to the sensor fusion module and the individualized baseline model construction module, and is configured to input the actual performance data into the individualized exercise physiological response baseline model to determine expected physiological data. The timing deviation calculation module is connected to the sensor fusion module and the expected physiological response determination module, and is configured to calculate a timing deviation value based on the actual physiological data and the expected physiological data. The health risk attribution module is connected to the sensor fusion module and the timing deviation calculation module, and is configured to determine whether the performance decline is due to regular exercise-induced fatigue or potential physiological health risk based on a comparison result of the timing deviation value and a preset attribution threshold when the performance decline is detected based on the actual performance data, and output a corresponding instruction.

2. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The calibration performance data and the actual performance data include at least one of force data, speed data, power output sequence, motion trajectory data, force mode data, or foot pressure distribution data collected by an intelligent training device; and the calibration physiological data and the actual physiological data include at least one of heart rate sequence, skin electric response sequence, body temperature data, or bioelectric signal data collected by an intelligent wearable device.

3. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The preset baseline model algorithm includes a multivariate time series regression model or a dynamic time warping algorithm, and the individualized baseline model construction module is specifically configured to acquire calibration performance data and calibration physiological data of a user during execution of a set of standardized training actions in a confirmed healthy state, and train the preset baseline model algorithm to obtain the individualized exercise physiological response baseline model.

4. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The health risk attribution module is specifically configured to determine the preset attribution threshold as a first threshold value, determine that the performance decline is due to regular exercise-induced fatigue and output a movement guidance instruction if the timing deviation value is within the first threshold value, and determine that the performance decline is due to potential physiological health risk and output a health risk warning instruction if the timing deviation value exceeds the first threshold value.

5. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The time sequence deviation calculation module is specifically configured to: at any time point, acquire an actual physiological response value of actual physiological sign data and an expected physiological response value of expected physiological sign data; and calculate a time sequence deviation value D, which is determined by the following formula: wherein is the actual physiological response value, is the expected physiological response value.

6. The motion training management system based on multi-sensor fusion according to claim 2, characterized in that, The intelligent training device includes at least one of an intelligent human target, an intelligent biomechanical insole, an intelligent wall-mounted reaction training system, or a digital comprehensive strength training device, and the intelligent wearable device includes a smartwatch or a smart armband.

7. The motion training management system based on multi-sensor fusion according to claim 6, characterized in that, The intelligent human target is embedded with a force sensor array for measuring striking force and speed; the intelligent biomechanical insole is integrated with a flexible pressure sensor network for capturing foot pressure distribution and gait cycle; and the digital comprehensive strength training device uses a servo motor resistance system to record force-speed curve and power output.

8. The motion training management system based on multi-sensor fusion according to claim 4, characterized in that, The exercise guidance instruction comprises: an instruction for prompting the user to take a break or adjust the training intensity between groups; the health risk warning instruction comprises: a high-priority instruction for prompting the user to immediately pause the training, detect physiological signs, or seek medical assistance; and the health risk attribution module is further configured to: prevent the output of the exercise guidance instruction when it is determined that the potential physiological health risk is attributed.

9. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The sensor fusion module is further configured to acquire user limb data collected by a visual detection device; and the individualized baseline model construction module is further configured to assist in establishing the individualized exercise physiological response baseline model based on the user limb data.

10. The motion training management system based on multi-sensor fusion according to claim 1, characterized in that, The system further comprises a health record management module, which is configured to: store the individualized exercise physiological response baseline model; and continuously record the timing deviation value, taking the timing deviation value as a dynamic health information dimension.

Images