A treadmill-based method of monitoring a user's health status

Abstract

This invention relates to the field of sports health monitoring technology, and more particularly to a method for monitoring the health status of users based on a treadmill. The method includes: acquiring plantar pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during the user's running process, and synchronizing these signals in time to form a raw signal set that reflects the dynamic coupling characteristics between the treadmill and the human body. In this invention, by simultaneously acquiring plantar pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during running, and synchronizing these signals in time, effective decoupling of mechanical vibration components and gait vibration components is achieved. Finally, a cardiovascular dynamic prediction network is constructed for health status assessment and adaptive adjustment of the monitoring strategy. This improves upon the problem that traditional health monitoring systems, which mostly use a single signal acquisition method, suffer from low accuracy in health assessment due to their inability to dynamically adapt to changes in the user's state.

Description

Technical Field

[0001] This invention relates to the field of sports health monitoring technology, and in particular to a method for monitoring the health status of users based on treadmills. Background Technology

[0002] With the increasing demand for health management and personalized medicine, sports health monitoring systems have been widely used in fitness, sports training, and health management for the elderly. Traditional health monitoring methods cannot comprehensively capture the dynamic health changes of users during exercise and fail to fully consider the interactions between physiological characteristics such as gait and cardiovascular function. Most existing health monitoring systems use a single signal acquisition method, which, due to its inability to dynamically adapt to changes in the user's state, results in low accuracy in health assessment. Summary of the Invention

[0003] To overcome the above shortcomings, this invention provides a method for monitoring the health status of users based on treadmills, aiming to improve the problem that most traditional health monitoring systems use a single signal acquisition method, which leads to low accuracy in health assessment due to their inability to dynamically adapt to changes in the user's status.

[0004] This invention provides the following technical solution: a method for monitoring the health status of users based on a treadmill, comprising the following steps: S1. Acquire foot pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during the user's running process, and synchronize various signals in time to form an original signal set that can reflect the dynamic characteristics of the coupling between the treadmill and the human body. S2. After obtaining the original signal set, the mechanical vibration component and the gait vibration component are decoupled based on the difference in vibration mode of different signal sources. The continuous gait cycle is divided based on the change of plantar pressure after decoupling, so as to obtain the gait cycle sequence and purification micro-vibration characteristic sequence for subsequent analysis. S3. Based on the gait cycle sequence and the purified micro-vibration characteristic sequence, a latent variable model is constructed to characterize the coupling relationship between gait dynamics and cardiovascular dynamics, and a set of latent variables that characterize the trend of cardiovascular dynamic changes are extracted from it to provide input for prediction. S4. After obtaining the set of latent variables, establish a cardiovascular dynamic prediction network based on the set of latent variables, so that the network outputs continuous prediction results of the user's cardiovascular dynamic response during running.

[0005] S5. After obtaining the prediction results, assess the user's health status based on the steady-state deviation, dynamic trend and periodic consistency of the prediction results, and generate health status results. S6. After obtaining the health status results, the signal acquisition parameters and processing flow in the monitoring are adaptively adjusted according to the results, so that the monitoring process forms a dynamic closed loop and the monitoring strategy is continuously adjusted as the operation progresses.

[0006] By adopting the above technical solution, the plantar pressure distribution signal, the micro-vibration signal of the treadmill belt, and the drive torque response signal are collected simultaneously during running. The various signals are synchronized in time, thereby achieving effective decoupling of mechanical vibration components and gait vibration components. Finally, a cardiovascular dynamic prediction network is constructed to assess health status and adaptively adjust the monitoring strategy. This improves the problem that traditional health monitoring systems mostly use a single signal acquisition method, which cannot dynamically adapt to changes in the user's state, resulting in low accuracy of health assessment.

[0007] Preferably, the time synchronization of various signals includes: Foot pressure signals, surface micro-vibration signals, and driving torque response signals were collected separately to form independent initial time series; Read the time reference of the treadmill controller, map each independent time series to a unified time axis, and perform time drift correction and interpolation processing on the mapped signal; The corrected signals are combined according to the time sequence to form a synchronous original signal set, which is used for subsequent vibration decoupling processing.

[0008] Preferably, the decoupling of the mechanical vibration component from the gait vibration component includes: Analyze the characteristics of the original signal in the time and frequency domains to identify the mechanical vibration modes transmitted by the treadmill drive system; The identified mechanical vibration patterns are separated from the original signal to obtain the purified gait vibration signal; The purified gait vibration signal is used for gait period segmentation.

[0009] Preferably, the gait period division includes: Identify the foot contact point and departure point in the vibration signal to form a key event sequence; Gait cycles are divided into multiple gait segments by the time intervals between adjacent key events; The gait segments are arranged in the order of occurrence to form a gait cycle sequence, which is used as input for the latent variable model.

[0010] Preferably, the latent variable model includes: The gait cycle sequence and the purification micro-vibration sequence are combined and input into the model according to a unified time window; The correlation between gait period and micro-vibration sequence was analyzed under a unified window. The association is converted into a set of candidate latent variables.

[0011] Preferably, the set of candidate latent variables includes: Analyze the synchronous change behavior of candidate latent variables in continuous gait cycles; Irrelevant variables were eliminated based on the degree of correlation between candidate latent variables and micro-vibration sequences; The selected latent variables are combined according to time series to form a latent variable set, which is then used as input to the prediction network.

[0012] Preferably, the cardiovascular dynamic prediction network includes: Input the set of latent variables into the prediction network in chronological order; The predictive network parses the input to generate a continuous-time cardiovascular dynamic prediction sequence; The predicted sequence output is used as input for health status assessment.

[0013] Preferably, the cardiovascular dynamic prediction network further includes: The predicted sequence is segmented into consecutive time periods to form multiple feature segments; Three types of features are extracted from the feature fragments: steady-state shift, dynamic change, and periodic structure. The three types of features are combined to form the input feature set required for health status assessment.

[0014] Preferably, the assessment of the health status includes: The feature set is input into the evaluation process in a preset order, so that the evaluation process can parse various features in sequence. The combination relationships between features are analyzed according to the evaluation process to form intermediate judgment results; The intermediate judgment results are further integrated to generate a single health status result, which serves as the basis for subsequent adaptive adjustments.

[0015] Preferably, the adaptive adjustment includes: Read the health status results and match the corresponding collection parameter adjustment strategy according to preset rules; Adjustments are made to the signal acquisition method, sampling frequency, and key steps in the processing flow based on the matched adjustment strategy. The adjusted acquisition process is applied to the next round of acquisition, thereby forming a new set of raw signals, and then entering the next monitoring process of cyclic execution.

[0016] The present invention has the following beneficial effects: 1. In this invention, by simultaneously collecting foot pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during running, and synchronizing these signals in time, the mechanical vibration components and gait vibration components are effectively decoupled. Finally, a cardiovascular dynamic prediction network is constructed to assess health status and adaptively adjust the monitoring strategy. This improves the problem that traditional health monitoring systems mostly use a single signal acquisition method, which cannot dynamically adapt to changes in the user's state, resulting in low accuracy of health assessment.

[0017] 2. In this invention, the differences in vibration modes of different signal sources are decoupled, and the gait cycle is divided based on the changes in plantar pressure after decoupling. Then, the purified micro-vibration feature sequence is extracted, thereby improving the traditional gait analysis method, which mostly relies on complex manual calibration and single vibration features, resulting in inaccurate gait cycle division and thus causing error problems in the monitoring process.

[0018] 3. In this invention, after obtaining the health status result, the signal acquisition parameters and processing flow are adaptively adjusted to form a dynamic closed loop and optimize the subsequent monitoring accuracy. This improves the problem that most traditional health monitoring systems use fixed acquisition and processing flows, which cannot be adjusted according to real-time data, resulting in fixed monitoring strategies that are difficult to meet the needs of different users, thus causing insufficient flexibility in health status monitoring. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for monitoring user health status based on a treadmill, as proposed in this invention. Figure 2 This is a flowchart of signal acquisition and processing for a treadmill-based method for monitoring user health status proposed in this invention. Figure 3 This is a flowchart of the latent variable model and prediction network for a treadmill-based user health status monitoring method proposed in this invention. Figure 4 This is a flowchart of the adaptive adjustment closed-loop process of a treadmill-based user health status monitoring method proposed in this invention. Figure 5 This is a hierarchical architecture diagram of a monitoring system for a treadmill-based user health status monitoring method proposed in this invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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. Example 1:

[0021] In a first embodiment of the present invention, the present invention provides a method for monitoring the health status of users based on a treadmill, such as... Figures 1-5 As shown, it includes the following steps: S1. Acquire foot pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during the user's running process, and synchronize various signals in time to form an original signal set that can reflect the dynamic characteristics of the coupling between the treadmill and the human body. Furthermore, time synchronization of various signals includes: Foot pressure signals, surface micro-vibration signals, and driving torque response signals were collected separately to form independent initial time series; Read the time reference of the treadmill controller, map each independent time series to a unified time axis, and perform time drift correction and interpolation processing on the mapped signal; The corrected signals are combined according to the time sequence to form a synchronous original signal set, which is used for subsequent vibration decoupling processing.

[0022] Specifically, during running, the system needs to acquire plantar pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals, and synchronize these signals in time. This step aims to create a raw signal set based on the same time reference, enabling subsequent vibration decoupling, gait cycle segmentation, and implicit feature extraction to be performed with a consistent time reference, avoiding signal mismatch caused by time axis offset.

[0023] The system independently acquires plantar pressure signals, surface micro-vibration signals, and drive torque response signals. Each type of signal originates from a different sensor, and its sampling frequency can be expressed as: plantar pressure sampling frequency. Micro-vibration sampling frequency Torque response sampling frequency Each signal forms an independent time series: plantar pressure series. Micro-vibration sequence Torque response sequence ;in , , These are the local timestamps of their respective sensors. This ensures that all three different types of signals are recorded completely in their original sampled form, without losing their respective physical characteristics.

[0024] Mapping to a unified time axis and correcting for time drift and sampling differences. The system reads the unified time reference provided by the treadmill controller, denoted as... The time mapping relationship can be represented as: ; in For the mapped unified timeline timestamp; This is the drift correction calculated based on the controller baseline. Then, interpolation is performed to handle sampling inconsistencies, using linear interpolation: ;in express , or One approach is to ensure that the three signals have corresponding values ​​at the same time point, so that the matrix inputs used for vibration decoupling remain aligned in the time dimension.

[0025] These are combined to form a synchronized original signal set. The synchronized original signals can then form a matrix input: ; in It is a collection of multi-source signals under a unified time axis; there is no time misalignment between elements. It provides synchronous information on human gait loading, treadmill structural response, and drive system load changes under a unified time reference, enabling each signal to establish a clear correspondence at the same moment.

[0026] The original synchronization signal set, after being output, will serve as the input for subsequent processing modules. Input to the vibration decoupling module: The goal of decoupling is to decompose the signal into: ;in It is a component of mechanical vibration; This is a gait vibration component. It ensures that the gait period division is not affected by mechanical vibrations.

[0027] Input to the gait period division module: This is a time series of plantar pressure. This is the purified gait vibration sequence. Periodic boundaries are extracted using plantar pressure peak-valley structure or threshold models. Output gait cycle sequence: .

[0028] Latent variable model input: gait cycle sequence ; Purification vibration characteristic sequence The model is used to derive latent dynamic variables. ,like: ;in This represents the set of implicit features derived from the correlation between gait dynamics and cardiovascular dynamics.

[0029] Predict the network input, including latent variables: Output cardiovascular dynamic values: .

[0030] Health status assessment input: Predicted sequence Output: Health status assessment information .

[0031] Closed-loop control module input: Health status result Output: The adjusted set of acquisition and processing parameters.

[0032] S2. After obtaining the original signal set, the mechanical vibration component and the gait vibration component are decoupled based on the difference in vibration mode of different signal sources. The continuous gait cycle is divided based on the change of plantar pressure after decoupling, so as to obtain the gait cycle sequence and purification micro-vibration characteristic sequence for subsequent analysis. Furthermore, the decoupling of mechanical vibration components from gait vibration components includes: Analyze the characteristics of the original signal in the time and frequency domains to identify the mechanical vibration modes transmitted by the treadmill drive system; The identified mechanical vibration patterns are separated from the original signal to obtain the purified gait vibration signal; The purified gait vibration signal is used for gait period segmentation.

[0033] Gait period division includes: Identify the foot contact point and departure point in the vibration signal to form a key event sequence; Gait cycles are divided into multiple gait segments by the time intervals between adjacent key events; The gait segments are arranged in the order of occurrence to form a gait cycle sequence, which is used as input for the latent variable model.

[0034] Specifically, after obtaining the original signal set, it is necessary to separate the mechanical vibration component from the gait vibration component based on the differences in vibration modes from different signal sources, and then use the separated plantar pressure sequence to divide the continuous gait cycle. The purpose of this step is to ensure that the gait structure information used by the subsequent model is distinguishable from the vibration information generated by the human body itself, avoiding confusion caused by treadmill equipment vibration in gait event recognition.

[0035] Vibration mode analysis. Original signal set. The input is sent to the vibration analysis module, which includes plantar pressure signals. Surface micro-vibration signal Torque response signal We perform time-domain and frequency-domain analyses on the three types of signals respectively. The time domain is used to identify periodic abrupt changes, and the frequency domain is used to identify frequency bands where energy is concentrated. Let the frequency domain representation of the signal be: ;in for , , Any signal in the middle; This is a Fourier transform operation; for Amplitude distribution in the frequency domain; For frequency, mechanical vibrations typically exhibit a fixed peak frequency, denoted as . The vibrations generated by gait have periodic variation characteristics, with their main frequency band distributed around the step frequency, denoted as step frequency. By comparing the amplitude distribution, the energy range of mechanical vibration can be obtained. and gait vibration energy range .

[0036] Vibration component separation. A frequency band filtering model is used to extract mechanical vibration from the original signal. The purified vibration signal is denoted as: ; To be based on the mechanical vibration frequency band Reconstructed mechanical vibration components The purified gait vibration signal is used for subsequent gait cycle recognition. By separating the mechanical vibration, the vibration sequence structure and the gait events in the plantar pressure sequence correspond more clearly on the time axis, making the recognition of ground contact events and ground lift events unaffected by equipment-driven vibrations.

[0037] Identify key event sequences. Input plantar pressure signals. ; purified gait vibration signal The identification of the contact point is based on the calculation of a threshold for the rising segment of plantar pressure, let the threshold be... Then the contact point satisfies: ; Location identification is based on a pressure drop threshold calculation: ;in For the time of contact point; Time from location; The plantar pressure threshold is determined by estimating the steady-state pressure mean. The identified time points, arranged in chronological order, form a sequence of critical events. : The key event sequence provides objective gait structure nodes, laying the foundation for subsequent periodic boundary calculations.

[0038] Gait cycles are defined by key events. The time interval between two adjacent key events constitutes one gait cycle. The cycle is denoted as: ;in For the first One gait cycle; For the first event in the event sequence Individual event points. A continuous gait cycle sequence is formed: The gait cycle sequence forms a complete time structure, enabling the latent variable model to independently model different gait cycles.

[0039] Output purified micro-vibration characteristic sequence. The purified vibration signal is segmented according to gait period to form a structured sequence: The output includes: gait cycle sequence ; the corresponding periodic purification vibration sequence Organizing vibration signals in a periodic form enables subsequent latent variable models to construct coupled feature representations based on the periodic structure.

[0040] Data input / output flow. Input: Raw signal set. Vibration mode analysis module; Vibration mode recognition input: , , Output: Mechanical vibration frequency band Gait vibration frequency band Vibration component separation input: , Output: Purify gait vibration Key event identification input: , Output: Event sequence Periodic input Output: Gait cycle sequence Periodic signal input: , Output: Periodic purification signal .

[0041] S3. Based on the gait cycle sequence and the purified micro-vibration characteristic sequence, a latent variable model is constructed to characterize the coupling relationship between gait dynamics and cardiovascular dynamics, and a set of latent variables that characterize the trend of cardiovascular dynamic changes are extracted from it to provide input for prediction. Furthermore, latent variable models include: The gait cycle sequence and the purification micro-vibration sequence are combined and input into the model according to a unified time window; The correlation between gait period and micro-vibration sequence was analyzed under a unified window. The association is converted into a set of candidate latent variables.

[0042] The candidate latent variable set includes: Analyze the synchronous change behavior of candidate latent variables in continuous gait cycles; Irrelevant variables were eliminated based on the degree of correlation between candidate latent variables and micro-vibration sequences; The selected latent variables are combined according to time series to form a latent variable set, which is then used as input to the prediction network.

[0043] Specifically, after obtaining the gait cycle sequence and the purified micro-vibration feature sequence, a latent variable model that reflects the coupling characteristics between gait dynamics and cardiovascular dynamics needs to be constructed. The role of this model is to extract temporally correlated latent variables from the structured cycle information and micro-vibration sequence, so that the input variables used for subsequent cardiovascular dynamic response prediction have a clear physical correspondence.

[0044] The combination of input sequences, the gait periodic sequence is denoted as: The purified micro-vibration characteristic sequence, after being periodically segmented, is denoted as: To ensure the model can align the temporal structure of gait rhythms and vibration signals, the two sequences are divided into sequences with a uniform window length. Combined into the input matrix: ;in For the first Input matrix for each time window; Gait period Structured feature sequences; This represents the purification microvibration sequence corresponding to this cycle. This combination method constitutes the input space of the latent variable model, enabling the structural characteristics and corresponding microvibrations of each cycle to be analyzed on the same time scale.

[0045] Analysis methods for the correlation between gait period and micro-vibrations. For the input matrix... This requires analyzing the relationship between the two sequences under a unified window. A linear mapping matrix is ​​used. Indicates the linear correlation between input sequences: ;in Depend on The set of candidate latent variables obtained by linear mapping; The mapping matrix learned by the model; The input matrix is ​​denoted as . Used to measure the linear contribution relationship of different features in the input within a time window. The variables in each dimension are candidate latent variables, representing the potential representation formed by the combination of gait features and micro-vibration features. It represents the joint variation pattern of gait structure and vibration structure within the same window, and is a preliminary implicit feature extracted from the joint sequence.

[0046] Analysis of synchronous change behavior across cycles. Candidate latent variable sequences. We need to analyze its temporal variation behavior over a continuous period. Let the sequence of latent variables formed by the continuous window be: ;right Each hidden variable dimension Calculate its cross-cycle correlation: ;in For the first in the sequence Candidate latent variables in the period The value of ; This is a correlation calculation function; For the first Consistency index of changes in candidate latent variables over consecutive periods. This is used to determine whether the latent variable has a stable change structure during a continuous period. If there is no periodic continuity, it cannot be used as a stable coupling feature.

[0047] Candidate variables unrelated to micro-vibrations were eliminated. For each latent variable... Correlation test with the purification vibration sequence: ;in The correlation coefficient between the latent variable and the purification vibration is used to determine whether to retain it; This is the purification vibration sequence corresponding to the cycle. Variables not directly related to micro-vibrations are eliminated, ensuring that the final retained latent variables all have clear sources and physical meanings.

[0048] Construct a set of latent variables and form the prediction input. Combine the retained latent variables using time series analysis to form the latent variable set: Where H is the set of latent variables; This is the j-th latent variable after retention. Finally... The sequence is used as input to the prediction network.

[0049] Data input and output flow, input end: gait cycle sequence Periodic purification micro-vibration sequence Sequence alignment and combination, input and Output: Combination matrix Latent variable generation, input: Output: Candidate hidden variables Cross-cycle variation analysis, input: Output: Latent variables in time series ; Relevance filtering, input: and Output: Set of hidden variables preserved Final output: Set of latent variables , as input to the cardiovascular dynamic prediction network.

[0050] S4. After obtaining the set of latent variables, establish a cardiovascular dynamic prediction network based on the set of latent variables, so that the network outputs continuous prediction results of the user's cardiovascular dynamic response during running. Furthermore, the cardiovascular dynamic prediction network includes: Input the set of latent variables into the prediction network in chronological order; The predictive network parses the input to generate a continuous-time cardiovascular dynamic prediction sequence; The predicted sequence output is used as input for health status assessment.

[0051] The cardiovascular dynamic prediction network also includes: The predicted sequence is segmented into consecutive time periods to form multiple feature segments; Three types of features are extracted from the feature fragments: steady-state shift, dynamic change, and periodic structure. The three types of features are combined to form the input feature set required for health status assessment.

[0052] Specifically, after obtaining the set of latent variables, a cardiovascular dynamics prediction network needs to be constructed so that it outputs a cardiovascular dynamics prediction sequence that changes over time after the set of latent variables is input. The core function of the network is to generate continuous-time predictive quantities that can be used for health status assessment based on the latent features extracted from gait cycles and micro-vibration sequences.

[0053] Latent variable set It consists of multiple latent variables: ;Will Arranged chronologically, forming a sequence of input vectors: ;in For the first Hidden variables in time The value of ; This is the input vector for the prediction network. Inputting it in chronological order ensures the prediction network can analyze the changes in latent variables during continuous gait.

[0054] The core structure of the prediction network uses a mapping function that takes time series data as input: ; in This is a predicted value for cardiovascular dynamic response; The time series mapping relationship established for the prediction network; To predict the set of parameters in the network. The network uses... The input latent variable sequence is mapped to a cardiovascular dynamic response, giving the output continuous-time properties.

[0055] The predicted output sequence is represented as: This sequence provides the basis for subsequent health status assessment.

[0056] After the network outputs, the predicted sequence needs to be divided into continuous time periods, and long-term changes, short-term changes, and periodic features need to be extracted from each prediction result.

[0057] The predicted sequence is divided into fixed-length windows. Divide into: ; in For the first Time series; This represents the window length for segmentation. By splitting the predicted sequence into multiple time periods, trend and structural change characteristics can be analyzed at the time scale.

[0058] Calculate the average from each time series segment: ;in For the first Steady-state offset characteristics of the segment. This represents the average level of the predicted amount within that time period.

[0059] Calculate the first-order difference sequence over the time period: Combine the sequences in chronological order to form a dynamic feature set: ;in To reflect the characteristic sequence of the predicted quantity changing over time.

[0060] Calculate the autocorrelation function of the predicted quantity over this time period: ; in To predict the sequence in lag The autocorrelation value is used to determine the periodic variation pattern of the predicted quantity within a given time period.

[0061] After extracting three types of features from each time series segment, a health assessment input feature set is formed: ;in For the first Segment feature set. Ultimately, all will be... The inputs are combined to form the health status assessment model.

[0062] Data flow direction, input end: set of latent variables Predict the network input; input: a sequence of vectors. Output: Cardiovascular dynamic prediction sequence Sequence segmentation, input: Output: Segmented sequence Feature extraction, input: Output: , , Feature combination, input: three feature sequences, output: health status assessment input feature set .

[0063] S5. After obtaining the prediction results, assess the user's health status based on the steady-state deviation, dynamic trend and periodic consistency of the prediction results, and generate health status results. Further assessment of health status includes: The feature set is input into the evaluation process in a preset order, so that the evaluation process can parse various features in sequence. The combination relationships between features are analyzed according to the evaluation process to form intermediate judgment results; The intermediate judgment results are further integrated to generate a single health status result, which serves as the basis for subsequent adaptive adjustments.

[0064] Specifically, after obtaining the cardiovascular dynamic prediction sequence output by the prediction network, the system constructs a health status assessment process based on the steady-state shift, dynamic trend, and periodic consistency in the sequence. The acquisition of various features depends on the feature set formed in the previous stage. The input to the assessment process is the feature set, and the output is a single health status result, which is used as the basis for subsequent adaptive adjustments.

[0065] Let the continuous prediction sequence generated by the prediction network be... ,in Indexed by time. The system starts from... Three basic features are extracted from the data: steady-state offset features. : ; This represents the average level of the sequence over the observation period; This represents the number of sampling points during that time period.

[0066] Dynamic change trend characteristics : ; It indicates the overall direction and magnitude of the sequence's change over the entire time period.

[0067] Periodic consistency characteristics : ;in It serves as an indicator of sequence correlation between adjacent gait cycle segments; This represents the number of periodic segment pairs used to calculate the correlation.

[0068] Three types of features are combined into a vector: This vector is the input to the health status assessment process.

[0069] The evaluation process receives feature sets in a fixed order. Feature input parsing, the system reads sequentially. The data are then sent to the corresponding parsing modules. The parsing modules determine the basic attributes of each type of feature based on the rules, such as the magnitude of the offset, the positive or negative direction of the trend value, and the degree of periodic consistency.

[0070] The intermediate judgment result is generated based on the combination relationship. The parsing results of each feature are sent to the combination judgment unit, which executes the preset logical relationship model.

[0071] The logical relationship can be formalized as follows: ; This is a pre-defined logical function used to generate an intermediate decision value. .in Used to determine steady-state level Used to determine the direction of a trend. Used to determine periodic stability The value reflects the combination of the three within the same time period.

[0072] Generate final health status results intermediate judgment quantity After output, the results are processed by the result integration module. The integration module then sorts the results according to the classification rules. Mapped to a single health status outcome: ;in This is a classification mapping function used to map continuous quantities to a single health status level. Health status results. As input for subsequent adaptive adjustment.

[0073] Data input / output flow: Input: Continuous prediction sequences from the prediction network ;Depend on calculate , , Three types of characteristics; , , The feature set is formed according to the set order. ; The input evaluation process proceeds sequentially to the feature parsing module; the parsing results are input into the combination judgment unit to generate intermediate judgment values. ;according to Generate a single health status result Output: Health status results used in subsequent adaptive adjustment processes. .

[0074] The process can transform the continuous cardiovascular dynamics sequence output by the prediction network into a structured result that can be used for subsequent processing. This allows the system to combine and judge steady-state shifts, trend changes, and periodic consistency according to fixed logic, and obtain a single health status result that can be further processed, thus achieving a complete closed loop of prediction → analysis → evaluation.

[0075] S6. After obtaining the health status results, the signal acquisition parameters and processing flow in the monitoring are adaptively adjusted according to the results, so that the monitoring process forms a dynamic closed loop and the monitoring strategy is continuously adjusted as the operation progresses. Furthermore, adaptive adjustment includes: Read the health status results and match the corresponding collection parameter adjustment strategy according to preset rules; Adjustments are made to the signal acquisition method, sampling frequency, and key steps in the processing flow based on the matched adjustment strategy. The adjusted acquisition process is applied to the next round of acquisition, thereby forming a new set of raw signals, and then entering the next monitoring process of cyclic execution.

[0076] Specifically, in the system generating health status results Then, the monitoring module based on Adaptive adjustments are made to the signal acquisition parameters and processing flow currently in operation, enabling the next round of acquisition to adjust according to changes in the current state. The input to this adaptive adjustment is the health status result. The output is a set of acquisition parameters and processing flow configuration for the next round of monitoring.

[0077] The health status result is recorded as The system has a pre-defined set of mapping rules to... This corresponds to specific parameter adjustment strategies. The mapping relationship can be represented by a function as follows: ;in As a result of health status; This is a pre-defined mapping function; This is a set of strategies for adjusting the acquired parameters. Acquisition parameter adjustment strategies. This includes parameters for the acquisition method, sampling frequency, and execution configurations for key steps in the processing flow. The system reads... Afterwards, Matched For subsequent implementation.

[0078] strategy set It includes multiple control variables related to the acquisition link, and the system performs adjustments based on these control variables: the acquisition mode is adjusted, and the parameter set includes an acquisition mode indicator. : ;in Used to indicate different sensor combinations or different acquisition paths. The system is based on... Switch the data acquisition method.

[0079] The sampling frequency is adjusted, and the sampling frequency is denoted as... The adjustment method is expressed as follows: ;in Based on the sampling frequency, For the reason A predetermined frequency adjustment coefficient is used to increase or decrease the sampling density.

[0080] The process flow can be adjusted step by step, and the parameter set includes step switch quantities and process indicator quantities, such as: ; in Indicates the first step in the process The execution configuration for each key step can be set to enabled or disabled. The system will then... Adjust the processing flow structure to ensure that the next round of data collection uses a processing path that better matches the current state. After adjustment, a new data collection configuration set is generated, denoted as: .

[0081] Configuration Collection The signal is input to the acquisition link to drive the next signal acquisition. The acquisition module relies on... Perform the acquisition to generate a new set of raw signals. : ;in This is the execution function for the acquisition link, used to generate a new set of raw signals based on the configuration set. Signal set It will be used as input for the next round of monitoring process, entering the cycle calculation, feature extraction, prediction and evaluation functional modules to achieve cyclical operation.

[0082] Data input / output process: Input: Health status result ;Will Input mapping function To generate a set of regulation strategies ;according to Adjust the acquisition method, sampling frequency, and processing flow configuration to obtain the acquisition configuration set. ;Will Input acquisition link, acquisition link performs acquisition to generate raw signal set ; Input is then fed into the next monitoring process to achieve closed-loop operation.

[0083] By adjusting the process, the monitoring system can update the collection and processing steps based on health status results, ensuring that the next round of collection is consistent with the current cardiovascular status, thereby completing a cyclical control closed loop in the monitoring link. Example 2:

[0084] This system monitors the health status of treadmill users in gyms, sports training centers, and home fitness environments. Utilizing multimodal signal acquisition technology, it acquires real-time data on plantar pressure distribution, treadmill belt micro-vibration signals, and drive torque response signals. Combined with a latent variable model of the coupling relationship between gait dynamics and cardiovascular dynamics, it performs real-time predictions of cardiovascular status. Furthermore, the system automatically adjusts acquisition parameters and processing procedures based on the health status results obtained from each monitoring session, forming a dynamic closed loop to ensure both accuracy and real-time monitoring.

[0085] In practical applications, the system faces several technical challenges. First, maintaining time synchronization and accurate acquisition during high-frequency, multi-source signal acquisition, due to the simultaneous operation of multiple sensors, is crucial. Second, the mechanical vibrations and gait signals generated by the treadmill are intertwined; effectively decoupling these two signals to extract accurate gait cycles and micro-vibration characteristics becomes a challenge for precise system analysis. Furthermore, the time-varying nature of gait and cardiovascular dynamics requires the construction of an accurate and real-time predictive model to reflect the cardiovascular status of different users. Personalized adjustments based on health status assessment results to ensure monitoring adaptability for different users are also challenges in system design. Finally, ensuring the real-time performance and stability of the adaptive adjustment mechanism in the dynamic closed loop, enabling the system to continuously optimize and adapt to different environments and user needs, is key to improving overall system reliability and user experience. To address these issues, this invention provides a treadmill-based user health status monitoring method, the structure of which is as follows: Figure 1 As shown. The specific implementation process of this method is as follows: During a user's run, sensors on the treadmill collect signals such as plantar pressure distribution, surface micro-vibration, and drive torque response. Each type of signal is collected as independent time-series data. To ensure the synchronization of different signal sources, these signals need to be time-aligned. First, by reading the treadmill controller's time reference, the collected signals are mapped onto a unified time axis. Then, time drift correction is performed to eliminate deviations caused by differences in sensor acquisition times. Finally, interpolation algorithms are used to correct these signals, ensuring they remain consistent on the same time axis, forming a raw signal set that reflects the dynamic coupling characteristics of the treadmill and the human body. This signal set provides the foundation for subsequent signal decoupling and analysis.

[0086] After obtaining the original signal set, the differences between mechanical vibration components and gait vibration components were identified by analyzing the characteristics of each signal in the time and frequency domains. Based on this, signal decoupling processing was performed. First, mechanical vibration modes generated by the treadmill drive system were identified; these modes have frequency and time-domain characteristics significantly different from gait vibrations. Through frequency-domain filtering and time-domain deconvolution techniques, these mechanical vibration modes were separated from the original signal, thus obtaining the purified gait signal. Next, the gait cycle was divided using the decoupled gait signal. By identifying the foot contact point and departure point, key events were determined, thereby dividing the purified signal into multiple gait cycles, each representing a complete gait process. Finally, the gait cycle sequence and purified micro-vibration feature sequence were obtained for subsequent analysis.

[0087] Based on gait cycle sequences and cleansing micro-vibration feature sequences, a latent variable model is constructed to characterize the coupling relationship between gait dynamics and cardiovascular dynamics. This latent variable model analyzes the temporal and spatial correlation between gait cycle sequences and cleansing micro-vibration feature sequences as inputs. The model extracts key latent variables through hidden layers; these latent variables characterize the changing trends of cardiovascular dynamics and reflect the complex interaction between gait and cardiovascular status. The extracted set of latent variables is provided as input to the subsequent cardiovascular dynamics prediction network, thus providing important input representations for the prediction of cardiovascular status.

[0088] After obtaining the set of latent variables, a cardiovascular dynamics prediction network is built using this set. This network, based on the input set of latent variables, is trained and predicted using a deep learning model, outputting the continuous trend of cardiovascular dynamics changes during running. The network model learns the temporal relationship between gait dynamics and cardiovascular dynamics to achieve real-time prediction of cardiovascular responses. The prediction results can provide necessary information for assessing the user's health status and provide foundational data for subsequent health monitoring and personalized interventions.

[0089] After obtaining the prediction results, the system assesses the user's health status based on characteristics such as steady-state deviation, dynamic trends, and periodic consistency. Specifically, by analyzing the steady-state component of the prediction results, it can identify whether the user is in a normal cardiovascular state; by analyzing the dynamic trends, it can identify possible changes or abnormalities in cardiovascular stress; and periodic consistency can be used to assess whether the user's cardiovascular response is stable at different stages of exercise. Based on these assessment indicators, the system generates a comprehensive health status result. This result serves as the basis for subsequent dynamic adjustments and health interventions, helping users monitor their health status in real time during exercise.

[0090] After obtaining the health status result, the system adaptively adjusts the signal acquisition parameters and processing flow during the monitoring process based on this result. The specific adjustment strategy is selected based on the health status assessment result. By reading the health status result and matching it with preset adjustment rules, the system adjusts the signal acquisition method, sampling frequency, and key data processing steps. For example, if the health status indicates a high cardiovascular burden, the system can reduce the sampling frequency to reduce computational burden and enhance signal stability; if the health status shows normal, the system can increase the sampling frequency to obtain higher-precision data. The adjusted acquisition and processing strategy will be applied to the next round of signal acquisition, thus forming a new raw signal set for the next monitoring cycle. The entire process forms a dynamic closed loop through an adaptive adjustment mechanism, continuously optimizing subsequent monitoring strategies and accuracy, ensuring that the monitoring system always remains consistent with the user's health status.

[0091] Through the technical implementation of the above steps, the system can automatically adjust itself based on the results of each health status assessment without relying on additional human intervention, making the monitoring system more flexible and accurate.

[0092] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the health status of users based on a treadmill, characterized in that, Includes the following steps: S1. Acquire foot pressure distribution signals, treadmill belt micro-vibration signals, and drive torque response signals during the user's running process, and synchronize various signals in time to form an original signal set that can reflect the dynamic characteristics of the coupling between the treadmill and the human body. S2. After obtaining the original signal set, the mechanical vibration component and the gait vibration component are decoupled based on the difference in vibration mode of different signal sources. The continuous gait cycle is divided based on the change of plantar pressure after decoupling, so as to obtain the gait cycle sequence and purification micro-vibration characteristic sequence for subsequent analysis. S3. Based on the gait cycle sequence and the purified micro-vibration characteristic sequence, a latent variable model is constructed to characterize the coupling relationship between gait dynamics and cardiovascular dynamics, and a set of latent variables that characterize the trend of cardiovascular dynamic changes are extracted from it to provide input for prediction. S4. After obtaining the set of latent variables, establish a cardiovascular dynamic prediction network based on the set of latent variables, so that the network outputs continuous prediction results of the user's cardiovascular dynamic response during running. S5. After obtaining the prediction results, assess the user's health status based on the steady-state deviation, dynamic trend and periodic consistency of the prediction results, and generate health status results. S6. After obtaining the health status results, the signal acquisition parameters and processing flow in the monitoring are adaptively adjusted according to the results, so that the monitoring process forms a dynamic closed loop and the monitoring strategy is continuously adjusted as the operation progresses.

2. The method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The time synchronization of various signals includes: Foot pressure signals, surface micro-vibration signals, and driving torque response signals were collected separately to form independent initial time series; Read the time reference of the treadmill controller, map each independent time series to a unified time axis, and perform time drift correction and interpolation processing on the mapped signal; The corrected signals are combined according to the time sequence to form a synchronous original signal set, which is used for subsequent vibration decoupling processing.

3. The method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The decoupling of the mechanical vibration component from the gait vibration component includes: Analyze the characteristics of the original signal in the time and frequency domains to identify the mechanical vibration modes transmitted by the treadmill drive system; The identified mechanical vibration patterns are separated from the original signal to obtain the purified gait vibration signal; The purified gait vibration signal is used for gait period segmentation.

4. The method for monitoring user health status based on a treadmill according to claim 3, characterized in that, The gait period division includes: Identify the foot contact point and departure point in the vibration signal to form a key event sequence; Gait cycles are divided into multiple gait segments by the time intervals between adjacent key events; The gait segments are arranged in the order of occurrence to form a gait cycle sequence, which is used as input for the latent variable model.

5. The method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The latent variable model includes: The gait cycle sequence and the purification micro-vibration sequence are combined and input into the model according to a unified time window; The correlation between gait period and micro-vibration sequence was analyzed under a unified window. The association is converted into a set of candidate latent variables.

6. The method for monitoring user health status based on a treadmill according to claim 5, characterized in that, The candidate latent variable set includes: Analyze the synchronous change behavior of candidate latent variables in continuous gait cycles; Irrelevant variables were eliminated based on the degree of correlation between candidate latent variables and micro-vibration sequences; The selected latent variables are combined according to time series to form a latent variable set, which is then used as input to the prediction network.

7. The method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The cardiovascular dynamic prediction network includes: Input the set of latent variables into the prediction network in chronological order; The predictive network parses the input to generate a continuous-time cardiovascular dynamic prediction sequence; The predicted sequence output is used as input for health status assessment.

8. The method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The cardiovascular dynamic prediction network also includes: The predicted sequence is segmented into consecutive time periods to form multiple feature segments; Three types of features are extracted from the feature fragments: steady-state shift, dynamic change, and periodic structure. The three types of features are combined to form the input feature set required for health status assessment.

9. A method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The assessment of the health status includes: The feature set is input into the evaluation process in a preset order, so that the evaluation process can parse various features in sequence. The combination relationships between features are analyzed according to the evaluation process to form intermediate judgment results; The intermediate judgment results are further integrated to generate a single health status result, which serves as the basis for subsequent adaptive adjustments.

10. A method for monitoring user health status based on a treadmill according to claim 1, characterized in that, The adaptive adjustment includes: Read the health status results and match the corresponding collection parameter adjustment strategy according to preset rules; Adjustments are made to the signal acquisition method, sampling frequency, and key steps in the processing flow based on the matched adjustment strategy. The adjusted acquisition process is applied to the next round of acquisition, thereby forming a new set of raw signals, and then entering the next monitoring process of cyclic execution.

Images