A wearable adaptive acquisition energy management method guided by physiological motion response

Abstract

The application discloses a wearable adaptive acquisition energy efficiency management method guided by physiological motion response, and relates to the technical field of physiological motion data processing.The application collects motion data and physiological data and completes time sequence alignment, adopts two differentiated algorithms to extract motion features and physiological features respectively, and constructs a physiological motion response correlation index time sequence through a correlation coefficient; an individual dynamic stability domain is established through a model with time sequence attenuation weight, and the current response deviation degree is calculated; a multilayer regularization LSTM model is used to predict the subsequent index change trend and divide a three-level risk grade; three value indexes of data anomaly, trend and scarcity are calculated in sequence, and a decision score is obtained through a nonlinear fusion model to generate multi-gear regulation parameters; and the sensor start-stop, sampling frequency, edge computing power and wireless transmission cycle are linked and adjusted.The application dynamically reduces the equipment power consumption under the premise of ensuring the physiological monitoring accuracy, and has wide application value.

Description

Technical Field

[0001] This invention relates to the field of physiological motion data processing technology, specifically to a wearable adaptive data acquisition and energy efficiency management method guided by physiological motion response. Background Technology

[0002] With the rapid development of intelligent sensing technology, edge computing technology, and the human health monitoring industry, wearable smart devices, with their advantages of lightweight design, portability, and all-weather monitoring, have become the core carrier for collecting human motion physiological data and assessing health status. Currently, wearable devices generally integrate multiple acquisition modules, including motion sensors and physiological sensors, enabling continuous collection of multi-dimensional vital signs data such as human motion posture, heart rate, and blood oxygenation. This provides crucial data support for human motion health analysis, daily status monitoring, and exercise load assessment. Currently, a large amount of related research focuses on sensor sampling control and power consumption optimization strategies for wearable devices. By dynamically adjusting sensor operating states, sampling frequencies, and data transmission mechanisms, the short battery life of wearable devices can be alleviated. However, significant technical bottlenecks remain in areas such as individualized adaptation accuracy, dynamic response correlation analysis, and intelligent energy efficiency synergistic control, making it difficult to simultaneously meet the dual requirements of high-precision monitoring and low-energy operation. In this context, research on high-precision, individualized, and dynamically adaptive wearable data acquisition energy efficiency management technology, addressing the homogeneity and universality deficiencies of existing wearable device energy efficiency control technologies, has significant practical significance and application value.

[0003] Chinese patent (publication number CN108628217A) discloses a power consumption control method for wearable devices. This method identifies the user's motion state and stationary state, and switches the sensor working mode according to the device's operating scenario to achieve basic power consumption control of wearable devices. However, this method only regulates energy consumption through a single motion state, without associating it with the changes in human physiological response, and ignores the dynamic correlation between motion and physiological state. This results in a single and fixed control logic, which cannot adapt to the different physiological fluctuations of the human body, and has poor monitoring accuracy and energy consumption balance.

[0004] Chinese patent (publication number CN114756114B) discloses a method and device for saving energy in wearable devices. This method reduces the ineffective energy consumption of the device by collecting human physiological and exercise data and adjusting the sensor sampling frequency and data transmission frequency according to a preset fixed threshold. However, this method uses a globally fixed discrimination threshold and does not build an individualized steady-state model adapted to different users. It cannot adapt to the individual differences in users' long-term physiological state and exercise habits, and is prone to high misjudgment and missed judgment problems.

[0005] Chinese patent (publication number CN116321040A) discloses a joint optimization method for energy management and network performance of smart clothing. This method achieves basic scheduling optimization of computing power and transmission resources of wearable devices through time-series data monitoring and state classification. However, this method only performs superficial regulation on device energy consumption parameters, lacks risk prediction and data value quantification mechanisms, and has obvious lag in energy consumption regulation, making it impossible to achieve forward-looking and refined energy efficiency management.

[0006] In summary, existing energy efficiency management technologies for wearable devices are insufficient to meet the current demands for high-precision, low-power, and personalized intelligent monitoring of human movement and physiology. Therefore, there is an urgent need to propose a novel wearable adaptive data acquisition energy efficiency management method to address the technical problems existing in the aforementioned technologies. Summary of the Invention

[0007] Based on the above-mentioned technical problems, this application discloses a wearable adaptive energy management method guided by physiological motion response, specifically including:

[0008] The user's motion and physiological data are collected synchronously through motion sensors and physiological sensors in wearable devices;

[0009] Motion state features are extracted from the motion data, physiological response features are extracted from the physiological data, and a physiological motion response correlation index is calculated based on the dynamic response relationship between the motion state features and the physiological response features, thereby constructing a physiological motion response correlation index sequence.

[0010] A response deviation analysis model is constructed. The analysis model is based on the physiological motor response correlation index sequence obtained during historical monitoring. An individualized response stability domain is constructed. The degree of deviation of the current physiological motor response is obtained by comparing the physiological motor response correlation index at the current moment with the individualized response stability domain.

[0011] A physiological motor response prediction model is constructed. The prediction model is based on the physiological motor response correlation index sequence to predict the trend of response changes in future time periods and analyze the risk level of future responses.

[0012] Based on the current deviation of physiological and motor response, the risk level of future response, and the information value indicators of current monitoring data, a decision score is calculated to generate energy efficiency management decision parameters.

[0013] The working status, data sampling frequency, data processing strategy, and wireless transmission cycle of each sensor in the wearable device are dynamically adjusted according to the energy efficiency management decision parameters to achieve adaptive optimization between monitoring accuracy and device energy consumption.

[0014] Preferably, the motion state features and physiological response features are as follows: the motion state features are based on motion time series data, and the mean amplitude, variance of fluctuation, and frequency of the data within the window are calculated using a time domain statistical feature extraction algorithm to extract motion state features, including motion intensity, motion frequency, and motion stability.

[0015] Physiological response features are based on stationary physiological time-series data. An adaptive smoothing difference feature extraction algorithm is used. After removing impulse noise from the physiological data through median filtering, the real-time rate of change of the data is calculated through first-order difference operation to extract physiological response features, including heart rate fluctuation coefficient, blood oxygen change rate, and body temperature offset.

[0016] Preferably, the calculation of the physiological motor response correlation index specifically involves quantifying the dynamic correlation between motor characteristics and physiological characteristics using the Pearson correlation coefficient, as shown in the formula:

[0017]

[0018] in, For single feature pair correlation, The covariance of motor characteristics and physiological characteristics. , These are the variances of motor characteristics and physiological characteristics, respectively.

[0019] The correlation degree of multiple sets of features is weighted and fused to calculate the physiological motor response correlation index.

[0020] Preferably, the response deviation analysis model specifically involves: collecting the user's historical physiological motor response correlation index sequence, fitting the historical CI sequence with a Gaussian distribution, and combining it with a time-series decay weighting factor to construct an individualized response stability region, as shown in the formula:

[0021]

[0022]

[0023]

[0024] in, For the first Each time-decay weight The time decay coefficient, The total number of historical physiological motor response association index sequences. For the weighted steady-state mean, For the first A physiological motor response correlation index, For weighted steady-state standard deviation, , These are the upper and lower limits of the stability region of the individualized response, respectively. This is the individual steady-state tolerance coefficient;

[0025] The deviation of the real-time physiological motor response is quantified by the deviation between the current response correlation index and the stable region interval, using the following formula:

[0026]

[0027] in, The degree of deviation in response at the current moment. The term represents the correlation index of physiological and motor response at the current moment. In the formula, the first term represents the dispersion of the data relative to the steady-state mean, and the second term is the steady-state boundary offset correction coefficient. The larger the value, the greater the deviation of the current physiological motor response state from the user's individualized dynamic steady-state range, and the more significant the abnormal response characteristics.

[0028] Preferably, the physiological motor response prediction model specifically comprises: constructing a continuous time-series response correlation index sequence. The LSTM time series prediction model is used to predict the time series of events. Perform iterative predictions and output a predicted sequence of future response correlation indices. ;

[0029] The response risk level is determined based on the fluctuation amplitude and frequency of deviations from the stable region of the predicted sequence.

[0030] Preferably, the information value index is specifically: the data anomaly value, data trend value, and data scarcity and uniqueness value are calculated separately using independent algorithms, and then weighted and fused to obtain the overall data information value;

[0031] Among them, the value of data anomaly is quantified based on the degree of deviation of the current response from the steady-state boundary. When the deviation exceeds the individual steady-state tolerance, the value increases significantly, which represents the high information gain of the abnormal data.

[0032] The value of data trends is based on the slope of the change in the time series response correlation index to quantify the significance of dynamic trends, and the dynamic evolution value of the current data sequence is characterized by the time series average rate of change.

[0033] The value of data scarcity and uniqueness is quantified by the proportion of historical steady-state sample distribution, which determines the degree of sample scarcity.

[0034] Based on the value of data anomalies, the value of data trends, and the value of data scarcity and uniqueness, the information value index is calculated using the following formula.

[0035] Preferably, the calculation of the decision score specifically involves: performing interval normalization on the response deviation degree, response risk level, and data information value to obtain evaluation factors within the same quantification interval; configuring power-law nonlinear transformations for different evaluation factors to achieve adaptive adjustment of low-value weak response and high-value strong gain; combining a multi-indicator joint situational enhancement term to adaptively weight and strengthen highly sensitive scenarios with simultaneous increases in multi-dimensional indicators; and fusing these to obtain a decision score that accurately represents the current monitoring scenario's needs and energy consumption control priorities. The formula is as follows:

[0036]

[0037] in, , , These are the normalized values ​​for the degree of response deviation, risk level, and data information value, respectively. These are the corresponding normalized weights. For dynamic enhancement coefficients, It is a very small constant; the first half of the formula is the weighted value of indicators of different dimensions, and the second half is the joint situation enhancement term, which adaptively amplifies the decision level when multiple indicators rise at the same time.

[0038] Preferably, the generation of energy efficiency management decision parameters specifically involves: based on the decision score, matching sensor sleep / standby / high-precision wake-up level, low / medium / high sampling frequency level, lightweight / conventional / high-precision edge computing level, and latency / conventional / real-time data transmission level according to the three threshold ranges corresponding to low, medium, and high risks, respectively, to achieve multi-dimensional linkage adaptive control of device parameters. The higher the score, the higher the priority of device monitoring accuracy and real-time performance, and the more aggressive the energy consumption adaptation strategy.

[0039] Preferably, the adaptive optimization between monitoring accuracy and equipment energy consumption specifically includes adaptive optimization of monitoring accuracy and adaptive optimization of equipment energy consumption;

[0040] When the decision score increases, the sampling frequency of the motion sensor and physiological sensor is adaptively increased; when the decision score decreases, the sampling frequency is decreased. The adaptive adjustment formula for the sampling frequency during the process is as follows:

[0041]

[0042] in, For adaptive sampling frequency, Equipment basic sampling frequency, This is the frequency adjustment coefficient. Rate the current decision. This is the maximum decision score threshold.

[0043] Preferably, the adaptive optimization of the device's energy consumption specifically involves: dividing edge computing into three levels—lightweight computing, conventional computing, and high-precision computing—and dynamically allocating edge computing resources based on decision parameters. In low-risk scenarios, lightweight filtering and simple feature extraction algorithms are used to reduce computing power consumption; in high-risk scenarios, high-precision response correlation index calculation and trend prediction algorithms are enabled to ensure data processing accuracy. The data transmission cycle control formula during the process is as follows:

[0044]

[0045] in, For data transmission cycle, Based on the transmission period, S is the transmission adjustment coefficient, S is the current decision score, and Smax is the maximum decision score threshold.

[0046] Compared with the prior art, the technical solution of this application has the following technical effects:

[0047] This invention constructs a dynamic correlation analysis mechanism between motion characteristics and physiological response characteristics. By quantifying the physiological motion response correlation index, it accurately characterizes the matching relationship between human motion stimulation and physiological feedback, effectively improving the matching accuracy between equipment operating condition control and the actual state of the human body.

[0048] This invention constructs an individualized dynamic response stability domain model with time-decay weights, which can form a unique individual steady-state range by combining historical monitoring data of users, and adapt the physiological fluctuation characteristics under different exercise intensities through the individual steady-state tolerance coefficient, thus greatly improving the individualized adaptation capability of human physiological exercise state discrimination and energy efficiency regulation.

[0049] This invention achieves forward-looking prediction of physiological motion response trends by optimizing the LSTM time series prediction model through multi-level regularization, and establishes a multi-level response risk classification system. At the same time, based on a multi-dimensional information value quantification mechanism that includes data anomaly, data trend, and data scarcity, it realizes the upgrade from passive energy consumption control to proactive predictive refined energy efficiency management.

[0050] This invention combines a multi-level nonlinear fusion decision model with a multi-device parameter collaborative control strategy. It can adaptively adjust the sensor's operating level, sampling frequency, edge computing power, and wireless transmission cycle based on dynamic decision scores. While ensuring high accuracy and integrity in human motion physiological monitoring, it minimizes the device's ineffective energy consumption and achieves a dynamic adaptive balance between monitoring accuracy and device power consumption.

[0051] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0052] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0054] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0055] Figure 1 A flowchart illustrating the overall process of a wearable adaptive energy management method guided by physiological motion response;

[0056] Figure 2 This is an architecture diagram of a wearable adaptive energy management method guided by physiological motion response;

[0057] Figure 3 This is a diagram of the architecture of the response deviation analysis model in this application;

[0058] Figure 4 This is a diagram illustrating the architecture of the physiological motor response prediction model in this application;

[0059] Figure 5 This is a time-series diagram of the original monitoring data and exercise intensity analysis during the exercise period in the embodiments of this application;

[0060] Figure 6 This is a schematic diagram illustrating the prediction results for future time periods during the motion process in an embodiment of this application;

[0061] Figure 7 This is a probability comparison chart of the power consumption data of each method in the embodiments of this application;

[0062] Figure 8 This is a comparison chart of the overall performance data of the various methods in the embodiments of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0064] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0065] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0066] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0067] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0068] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0069] Example 1 describes a wearable adaptive energy management method guided by physiological motion response, such as... Figure 1 , Figure 2 As shown, it specifically includes:

[0070] The user's motion and physiological data are collected synchronously through motion sensors and physiological sensors in wearable devices;

[0071] Motion state features are extracted from the motion data, physiological response features are extracted from the physiological data, and a physiological motion response correlation index is calculated based on the dynamic response relationship between the motion state features and the physiological response features, thereby constructing a physiological motion response correlation index sequence.

[0072] A response deviation analysis model is constructed. The analysis model is based on the physiological motor response correlation index sequence obtained during historical monitoring. An individualized response stability domain is constructed. The degree of deviation of the current physiological motor response is obtained by comparing the physiological motor response correlation index at the current moment with the individualized response stability domain.

[0073] A physiological motor response prediction model is constructed. The prediction model is based on the physiological motor response correlation index sequence to predict the trend of response changes in future time periods and analyze the risk level of future responses.

[0074] Based on the current deviation of physiological and motor response, the risk level of future response, and the information value indicators of current monitoring data, a decision score is calculated to generate energy efficiency management decision parameters.

[0075] The working status, data sampling frequency, data processing strategy, and wireless transmission cycle of each sensor in the wearable device are dynamically adjusted according to the energy efficiency management decision parameters to achieve adaptive optimization between monitoring accuracy and device energy consumption.

[0076] Furthermore, the motion data includes motion acceleration, angular velocity, number of steps, and motion posture parameters; the physiological data includes heart rate, blood oxygen saturation, body temperature, and electrocardiogram waveform parameters.

[0077] The motion data and physiological data are time-aligned according to a unified time reference, and the original sampling time series sets of motion sensors and physiological sensors are obtained respectively. A linear interpolation algorithm is then used to complete the time series of the two types of data.

[0078] Furthermore, the motion state characteristics and physiological response characteristics are as follows: The motion state characteristics are based on motion time series data. The time domain statistical feature extraction algorithm is used to calculate the mean amplitude, variance of fluctuation, and frequency of the data within the window to extract motion state characteristics, including motion intensity, motion frequency, and motion stability.

[0079] Physiological response features are based on stationary physiological time-series data. An adaptive smoothing difference feature extraction algorithm is used. After removing impulse noise from the physiological data through median filtering, the real-time rate of change of the data is calculated through first-order difference operation to extract physiological response features, including heart rate fluctuation coefficient, blood oxygen change rate, and body temperature offset.

[0080] Furthermore, the physiological-motor response correlation index is calculated as follows: the dynamic correlation between motor characteristics and physiological characteristics is quantified using the Pearson correlation coefficient, with the following formula:

[0081]

[0082] in, For single feature pair correlation, The covariance of motor characteristics and physiological characteristics. , These are the variances of motor characteristics and physiological characteristics, respectively.

[0083] The physiological motor response correlation index is calculated by weighted fusion of multiple feature correlations, using the following formula:

[0084]

[0085] in, It is a physiological motor response correlation index. For the first The weight coefficients of the feature pairs. The total number of feature groups. For the first Group-single feature correlation degree.

[0086] Furthermore, such as Figure 3 The diagram shown illustrates the architecture of the response deviation analysis model. Specifically, the model involves collecting the user's historical physiological motion response correlation index (CI) sequence, fitting the historical CI sequence using a Gaussian distribution, and combining this with a time-decay weighting factor to construct an individualized response stability region. The formula is as follows:

[0087]

[0088]

[0089]

[0090] in, For the first Each time-decay weight The time decay coefficient, The total number of historical physiological motor response association index sequences. For the weighted steady-state mean, For the first A physiological motor response correlation index, For weighted steady-state standard deviation, , These are the upper and lower limits of the stability region of the individualized response, respectively. This is the individual steady-state tolerance coefficient;

[0091] The deviation of the real-time physiological motor response is quantified by the deviation between the current response correlation index and the stable region interval, using the following formula:

[0092]

[0093] in, The degree of deviation in response at the current moment. The term represents the correlation index of physiological and motor response at the current moment. In the formula, the first term represents the dispersion of the data relative to the steady-state mean, and the second term is the steady-state boundary offset correction coefficient. The larger the value, the greater the deviation of the current physiological motor response state from the user's individualized dynamic steady-state range, and the more significant the abnormal response characteristics.

[0094] Furthermore, such as Figure 4 The diagram shown illustrates the architecture of a physiological motor response prediction model. Specifically, the physiological motor response prediction model involves constructing a continuous time-series response correlation index sequence. The LSTM time series prediction model is used to predict the time series of events. Perform iterative predictions and output a predicted sequence of future response correlation indices. ;

[0095] The response risk level is determined based on the fluctuation amplitude and frequency of deviations from the stable region of the predicted sequence.

[0096] Furthermore, the specific classification of response risk levels is as follows:

[0097] Low-risk (L1) predictions show that all CI sequences are within the stable region, with volatility less than the preset threshold.

[0098] Medium-risk (L2) predictions show that the CI sequence slightly deviates from the stable region, but there is no sustained deviation trend.

[0099] High-risk (L3) predictions indicate that the CI sequence continuously deviates from the stable region, with fluctuations exceeding the preset threshold, posing a risk of abnormal physiological motor response.

[0100] Furthermore, the LSTM time series prediction model is a multi-level regularized optimized LSTM. Its input layer has an 8-dimensional time series feature dimension, matching the multi-dimensional input of response correlation index time series features, fluctuation features, and offset features. The first LSTM hidden layer has 64 neurons, uses the ReLU function as the activation function, and configures the dropout regularization coefficient to 0.2 to suppress time series overfitting. The second LSTM hidden layer has 32 neurons, uses the ReLU function as the activation function, and configures the dropout regularization coefficient to 0.2, compressing time series features layer by layer and extracting core change patterns. The fully connected layer has 16-dimensional feature mapping nodes to complete the fusion and dimensionality reduction of high-dimensional time series features. The output layer has a single-dimensional output node, uses a linear activation function, and outputs the predicted value of the response correlation index at future time moments. The model iterative training uses an adaptive Adam optimizer with an initial learning rate of 0.001, an iteration batch size of 16, and a maximum number of iterations of 100, ensuring a balance between time series prediction accuracy and computational efficiency.

[0101] Furthermore, the information value indicators are as follows: the data anomaly value, data trend value, and data scarcity and uniqueness value are calculated separately using independent algorithms, and then weighted and fused to obtain the overall data information value;

[0102] The data anomaly value is quantified based on the deviation of the current response from the steady-state boundary. The value increases significantly when the deviation exceeds the individual steady-state tolerance, representing the high information gain of the anomalous data. The formula is:

[0103]

[0104] in, Value of data anomalies In response to the degree of deviation, This is the individual steady-state tolerance coefficient;

[0105] The value of data trends quantifies the significance of dynamic trends based on the slope of the change in the time-series response correlation index. The dynamic evolution value of the current data sequence is characterized by the time-series average rate of change, as shown in the formula:

[0106]

[0107] in, For the value of data trends, This represents the total duration of the time series statistics window. This is the index of the sampling time within the time window. This is the physiological motion response correlation index at the current sampling time. This is the physiological motion response correlation index at the previous adjacent sampling time. A fixed time interval between adjacent sampling times;

[0108] The value of data scarcity and uniqueness is quantified by the proportion of historical steady-state sample distribution, using the following formula:

[0109]

[0110] in, For the value of data scarcity and uniqueness, This represents the number of samples in the historical data that are highly similar to the current data in terms of characteristics. The total number of historical samples is used to dynamically represent the uniqueness of the current monitoring data;

[0111] Based on the value of data anomalies, data trends, and the value of data scarcity and uniqueness, the information value index is calculated using the following formula:

[0112]

[0113] in, As an indicator of information value, These are the weighting coefficients.

[0114] Furthermore, the decision score calculation specifically involves: normalizing the response deviation, response risk level, and data information value across intervals to obtain evaluation factors within the same quantification interval; differentially configuring power-law nonlinear transformations for different evaluation factors to achieve adaptive adjustment for low-value, weak-response and high-value, strong-gain scenarios; combining a multi-indicator joint situational enhancement term to adaptively weight and reinforce highly sensitive scenarios with simultaneous increases in multi-dimensional indicators; and fusing these to obtain a decision score that accurately characterizes the current monitoring scenario's needs and energy consumption control priorities. The formula is:

[0115]

[0116] in, , , These are the normalized values ​​for the degree of response deviation, risk level, and data information value, respectively. These are the corresponding normalized weights. For dynamic enhancement coefficients, It is a very small constant; the first half of the formula is the weighted value of indicators of different dimensions, and the second half is the joint situation enhancement term, which adaptively amplifies the decision level when multiple indicators rise at the same time.

[0117] Furthermore, the energy efficiency management decision parameters are generated as follows: based on the decision score, according to the three threshold ranges corresponding to low, medium and high risks, the sensor sleep / standby / high-precision wake-up level, low / medium / high sampling frequency level, lightweight / conventional / high-precision edge computing level, and latency / conventional / real-time data transmission level are matched respectively, so as to realize the linkage and adaptive control of multi-dimensional device parameters. The higher the score, the higher the priority of device monitoring accuracy and real-time performance, and the more aggressive the energy consumption adaptation strategy.

[0118] Furthermore, achieving adaptive optimization between monitoring accuracy and equipment energy consumption specifically includes adaptive optimization of both monitoring accuracy and equipment energy consumption.

[0119] When the decision score increases, the sampling frequency of the motion sensor and physiological sensor is adaptively increased; when the decision score decreases, the sampling frequency is decreased. The adaptive adjustment formula for the sampling frequency during the process is as follows:

[0120]

[0121] in, For adaptive sampling frequency, Equipment basic sampling frequency, This is the frequency adjustment coefficient. Rate the current decision. This is the maximum decision score threshold.

[0122] Furthermore, the adaptive optimization of monitoring accuracy is divided into at least four levels: dormancy level, low-power monitoring level, normal operation level, and high-precision acquisition level.

[0123] When the response deviation is low, the risk level is L1, and the data information value is low, the non-core sensors are controlled to enter a sleep state, and the core sensors enter a low-power guard state.

[0124] When the response deviation is high, the risk level is L3, and the data information value is high, all sensors are woken up and switched to high-precision acquisition mode, realizing on-demand sensor wake-up and dynamic start-stop.

[0125] Furthermore, the adaptive optimization of equipment energy consumption is as follows: edge computing is divided into three levels: lightweight computing, regular computing, and high-precision computing. Edge computing resources are dynamically allocated based on decision parameters. In low-risk scenarios, lightweight filtering and simple feature extraction algorithms are used to reduce computing power consumption; in high-risk scenarios, high-precision response correlation index calculation and trend prediction algorithms are enabled to ensure data processing accuracy. The data transmission cycle control formula during the process is:

[0126]

[0127] in, For data transmission cycle, Based on the transmission period, For transmission regulation coefficient, Rate the current decision. This is the maximum decision score threshold.

[0128] This embodiment details a wearable adaptive energy management method guided by physiological motion response. It collects motion and physiological data and aligns them temporally. Two differentiated algorithms are used to extract motion and physiological features respectively. A time-series sequence of physiological motion response correlation index is constructed using correlation coefficients. An individualized dynamic stability domain is established using a model with temporal decay weights to calculate the degree of deviation of the current response. A multi-level regularized LSTM model is used to predict the subsequent index change trend and classify it into three risk levels. Three value indicators—data anomaly, trend, and scarcity—are calculated sequentially, and a decision score is obtained through a nonlinear fusion model. Multi-level control parameters are generated in stages. Sensor start / stop, sampling frequency, edge computing power, and wireless transmission cycle are adjusted in conjunction.

[0129] Example 2, based on Example 1, details how, using publicly available datasets and this method, a wearable adaptive data acquisition and energy efficiency optimization simulation experimental environment was built to conduct a comparative experiment on simultaneous monitoring of human motion physiology and dynamic energy efficiency regulation. The specific details are as follows:

[0130] The publicly available dataset used is the WESAD wearable multi-source physiological motion dataset (https: / / archive.ics.uci.edu / dataset / 465 / wesad+wearable+stress+and+affect+detection). The data selected is continuous millisecond-level synchronous time-series data of 15 healthy subjects in multiple working conditions from the WESAD public dataset, covering daily working conditions such as sitting, slow walking, brisk walking, and light progressive exercise. It synchronously records inertial motion parameters and physiological time-series data such as heart rate, body surface temperature, and vascular volume fluctuations, supporting the analysis of motion-physiology coupling relationship.

[0131] Taking a 30-minute progressive load exercise monitoring process as an example, the raw monitoring data during the exercise was monitored, and the exercise intensity was analyzed based on the raw monitoring data, resulting in the exercise intensity analysis data shown in Table 1 below:

[0132] Table 1. Exercise Data

[0133]

[0134] According to Table 1 and Figure 5As shown in the time series diagram of the original monitoring data during the exercise, the exercise intensity obtained by this method can accurately reflect the instantaneous changes of external exercise stimuli within 30 minutes of exercise. This proves that the quantitative results of exercise intensity output by this method can accurately distinguish between external exercise stimulation and the cumulative physiological response in the body, and can provide an accurate working condition input basis for the subsequent construction of exercise-physiological correlation index.

[0135] Multi-source motion and physiological data were time-aligned based on a unified millisecond-level time benchmark. Linear interpolation was used to complete missing time-series data, and outlier samples were removed to form a standard data sequence with synchronized time series. Differential feature extraction was performed using a sliding window with a duration of 5 seconds and an overlap rate of 50%. For motion time-series data, a time-domain statistical feature algorithm was used to calculate the motion intensity as 0.513, the motion frequency as 2.31 Hz, and the motion stability as 0.847 within the current window. For physiological data, median filtering was first used to eliminate impulse noise, and then first-order difference operations were used to extract dynamic change features, resulting in a heart rate fluctuation coefficient of 0.136, a blood oxygen saturation rate of -0.41%, and a body temperature shift of 0.42℃.

[0136] The correlation index was calculated to quantify the dynamic correlation between exercise characteristics and physiological response characteristics. Among them, the Pearson correlation coefficient between exercise intensity and heart rate fluctuation coefficient was 0.91, the Pearson correlation coefficient between exercise frequency and blood oxygen change rate was 0.73, and the Pearson correlation coefficient between exercise stability and body temperature deviation was 0.68. The weight coefficients of the three sets of features were set to 0.5, 0.3, and 0.2, respectively. The weighted fusion calculation yielded a physiological exercise response correlation index CI of 0.810 at the current time.

[0137] The response deviation analysis model of the present invention is further invoked, and based on the correlation index sequence monitored in the past 30 days of the volunteer, a time-decayed weighted Gaussian fitting algorithm is used to construct an individualized dynamic response stability domain.

[0138] The calculated user-weighted steady-state mean μ = 0.621, the weighted steady-state standard deviation σ = 0.074, and the individual steady-state tolerance coefficient δ is 2.0, resulting in an individualized response stability domain interval of [0.473, 0.769].

[0139] The correlation index of 0.810 at the current moment exceeds the upper limit of the steady-state region, indicating that the current human body's physiological response deviates from the individual's steady state. The deviation degree of the current response is calculated to be ΔCI=1.287 using the nonlinear deviation degree calculation formula, which shows that the matching state between the current motion stimulus and the physiological response deviates significantly from the daily steady-state level.

[0140] The real-time continuous temporal correlation index sequence was input into a multi-layer regularized LSTM prediction model. The model strictly followed the structural configuration specified in the claims: a first hidden layer with 64 neurons and a second hidden layer with 32 neurons, both with a Dropout coefficient of 0.2. The Adam optimizer was used with an initial learning rate of 0.001. The response correlation trend was predicted in a rolling manner for the next 10 minutes. The prediction results are shown in Table 2 below.

[0141] Table 2. Prediction results for the next 10 minutes during exercise.

[0142]

[0143] According to Table 2 and Figure 6 As shown in the schematic diagram of the prediction results for future periods during the movement process, the prediction results are highly consistent with the trend of the measured correlation index obtained synchronously during the corresponding period, which verifies the accuracy and reliability of the time series prediction model. The data shows that the correlation index for future periods is consistently higher than the upper limit of the individual stability domain and shows a continuous upward trend. The fluctuation range exceeds the preset threshold. Therefore, the system determines that the current monitoring scenario is a high-response risk state at level L3.

[0144] Based on this, a multi-dimensional value quantification algorithm was used to calculate the value of data anomaly, data trend, and data scarcity and uniqueness, yielding results of 0.84, 0.78, and 0.66, respectively; weighting coefficients were then set. The weighted fusion yields a comprehensive information value of 0.775 for the current monitoring data.

[0145] The normalized response deviation, risk level quantification, and data information value are input into a multi-level nonlinear fusion energy efficiency decision model. A dynamic enhancement coefficient and a minimum zero-prevention constant are introduced. After differentiated nonlinear weighted fusion, a decision score of S=0.872 is obtained. Compared with the preset decision thresholds: S≤0.3 is the low-energy consumption monitoring mode, 0.3<S≤0.8 is the conventional monitoring mode, and S>0.8 is the high-precision enhanced monitoring mode. The current score meets the high-precision monitoring conditions.

[0146] Based on the decision score, an adaptive full-dimensional energy efficiency control strategy is triggered, dynamically updating the equipment operating parameters. The sampling frequency of the PPG sensor is increased from 50Hz to 100Hz, the sampling frequency of the blood oxygen sensor is increased from 25Hz to 50Hz, and the dormant ECG acquisition module is actively awakened. Edge computing is switched from a lightweight computing mode to a high-precision computing mode, enabling full feature analysis and trend prediction calculations. The BLE wireless transmission cycle is shortened from 60s to 10s, enabling real-time uploading of high-value abnormal data and completing the dynamic adaptive optimization of monitoring accuracy and equipment energy consumption.

[0147] The test data of this method during the statistical test period showed that, within the same monitoring period, the recognition rate of abnormal events by this invention reached 96.7%, which is higher than the recognition rate of 88.4% of the conventional sampling strategy; the average power consumption of this invention was 89.5mW, which is better than the average power consumption of 123.6mW of the conventional sampling strategy; the average battery life of the device of this invention reached 79.6h, which is 36.5% higher than the 58.3h of the conventional solution; at the same time, the data effectiveness rate increased from 91.2% to 98.4%.

[0148] This embodiment details a comparative experiment on synchronous monitoring of human motion physiology and dynamic energy efficiency regulation, conducted using a smart health monitoring wristband as the hardware experimental carrier to build a complete wearable adaptive acquisition and energy efficiency optimization experimental platform. The experimental results show that the present invention can identify potential abnormal risks in advance based on the dynamic coupling relationship between human motion state and physiological response, and adaptively adjust the sensor working state, sampling frequency, edge computing resources, and data transmission strategy. While ensuring monitoring accuracy, it significantly reduces device energy consumption, improves device battery life and monitoring reliability, verifying that the present invention has good engineering application value and promotion prospects.

[0149] Example 3 details a comparative experiment conducted using this method and existing mainstream methods under the experimental conditions and data of Example 2, as follows:

[0150] Current mainstream methods include conventional constant frequency sampling method, motion state triggered energy-saving regulation method, heart rate threshold triggered regulation method, timed polling low power regulation method, and weighted energy efficiency regulation method;

[0151] Among them, the conventional constant frequency sampling method sets a fixed sampling frequency and data transmission cycle for various sensors throughout the entire monitoring period. The equipment maintains a working mode of continuous acquisition and timed upload, without changing the equipment acquisition and transmission parameters throughout the process, thus completing the data monitoring task at all times.

[0152] The motion state-triggered energy-saving adjustment method collects human motion parameters in real time through motion sensors, classifies the user's motion state into levels, matches the corresponding sensor sampling level and data transmission rules according to different motion levels, and switches the device working mode according to the motion conditions to realize the power consumption adaptation and adjustment during the monitoring process.

[0153] The heart rate threshold trigger adjustment method uses real-time collected human heart rate data as the basis for regulation, presets multiple heart rate threshold ranges, matches the corresponding device acquisition frequency and data processing mode according to the heart rate range, and realizes graded regulation of device operating parameters through physiological indicator conditions.

[0154] The timed polling low-power adjustment method pre-sets a fixed working time cycle, and performs sensor data acquisition, data processing and device sleep operation alternately according to the cycle. It allocates the device's operating energy consumption through time-sharing operation to achieve long-term continuous monitoring.

[0155] The weighted energy efficiency control method simultaneously collects multi-dimensional monitoring data of human movement and physiology, selects multiple state assessment indicators, and integrates the data of each indicator through a fixed linear weighting formula to obtain a comprehensive assessment value. Based on the assessment value, the sensor sampling frequency and data transmission cycle are adjusted in stages to complete the energy efficiency control of the equipment.

[0156] The experimental test metrics include abnormal event recognition rate, average operating power consumption, average device battery life, and effective data utilization rate.

[0157] The abnormal event recognition rate was determined by manual calibration before the experiment, recording all real physiological movement abnormalities and load change moments during the monitoring period as a standard abnormal sample set. Each comparison method and the method of this invention independently completed the abnormal discrimination statistics under the same monitoring time sequence. The abnormal event recognition rate is the ratio of the number of abnormal samples correctly identified by the model to the total number of all manually calibrated abnormal samples, which is used to quantify the accuracy of each method in identifying abnormal states of exercise physiology.

[0158] Throughout the average operating power consumption experiment, a high-precision power analyzer was used to continuously sample the wearable device's power in real time at a sampling frequency of 1Hz, and the instantaneous operating power consumption of the device was continuously recorded during the 7-day monitoring period. The average operating power consumption is the arithmetic mean of the instantaneous power consumption at all sampling moments, which objectively represents the overall operating energy consumption level of the device.

[0159] Under the premise of a uniform initial charge of 300mAh, the same experimental environment, and the same monitoring task, the average battery life of the device was tested continuously for each strategy. The continuous working time of the device from full charge to low battery shutdown was recorded. The average battery life of the device is the average of the battery life of multiple repeated experiments, which is used to evaluate the battery life optimization capability of each energy efficiency strategy.

[0160] Effective data utilization rate involves screening all raw monitoring data collected by the equipment for validity, and data that is complete in time sequence, free of impulse noise, without missing or distorted data, and conforms to the laws of human physiological movement changes are judged as valid data; effective data pass rate is the ratio of the number of valid data samples to the total number of all raw data samples collected by the equipment, and is used to characterize the data acquisition quality and data redundancy control capability under different control strategies.

[0161] The power consumption data for each method over seven days was statistically analyzed, resulting in the power consumption data shown in Table 3 below:

[0162] Table 3 Operating power consumption data

[0163]

[0164] According to Table 3 and Figure 7 As shown in the power consumption data probability comparison chart of each method, the power consumption probability density data of the adaptive strategy of this invention is concentrated in the low power consumption range of 85.2mW~93.5mW. Not only is the average power consumption lower than the other five comparison schemes, but the sample dispersion is also smaller and the power consumption is more stable. This proves that the adaptive control method can dynamically match the collected parameters according to the monitoring conditions, effectively suppress the ineffective energy consumption of the equipment, and ensure the stability of the working state while reducing the power consumption.

[0165] The overall performance data for each method was obtained by statistical analysis and is shown in Table 4 below:

[0166] Table 4 Overall Performance Data

[0167]

[0168] According to Table 4 and Figure 8 As shown in the comparison chart of the comprehensive performance data of the various methods, the present invention achieves an abnormal event identification rate of 96.7%, which is higher than all the comparison schemes; the average operating power consumption is only 89.5mW, which is the lowest energy consumption level; the average battery life of the device is increased to 79.6h, which is the best battery life capability; and the effective data utilization rate reaches 98.4%, which is stronger in data collection integrity, verifying the significant comprehensive performance advantages of the present method.

[0169] This embodiment describes in detail a comparative experiment using this method and existing mainstream methods. In the experiment, dynamic adaptive acquisition is achieved through the coupling relationship between motion and physiology, which reduces ineffective energy consumption, extends the device's battery life, and ensures data integrity and anomaly identification accuracy. It solves the contradiction of mutual constraints on the performance of multiple objectives under the traditional control mode, and the overall application performance is significantly better than various existing control strategies.

[0170] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A wearable adaptive energy management method guided by physiological motion response, characterized in that, include: The user's motion and physiological data are collected synchronously through motion sensors and physiological sensors in wearable devices; Motion state features are extracted from the motion data, physiological response features are extracted from the physiological data, and a physiological motion response correlation index is calculated based on the dynamic response relationship between the motion state features and the physiological response features, thereby constructing a physiological motion response correlation index sequence. A response deviation analysis model is constructed. The analysis model is based on the physiological motor response correlation index sequence obtained during historical monitoring. An individualized response stability domain is constructed. The degree of deviation of the current physiological motor response is obtained by comparing the physiological motor response correlation index at the current moment with the individualized response stability domain. A physiological motor response prediction model is constructed. The prediction model is based on the physiological motor response correlation index sequence to predict the trend of response changes in future time periods and analyze the risk level of future responses. Based on the current deviation of physiological and motor response, the risk level of future response, and the information value indicators of current monitoring data, a decision score is calculated to generate energy efficiency management decision parameters. The working status, data sampling frequency, data processing strategy, and wireless transmission cycle of each sensor in the wearable device are dynamically adjusted according to the energy efficiency management decision parameters to achieve adaptive optimization between monitoring accuracy and device energy consumption.

2. The wearable adaptive energy management method guided by physiological motion response according to claim 1, characterized in that, The motion state features and physiological response features are as follows: The motion state features are based on motion time series data. The time domain statistical feature extraction algorithm is used to calculate the mean amplitude, fluctuation variance, and period frequency of the data within the window to extract motion state features, including motion intensity, motion frequency, and motion stability. Physiological response features are based on stationary physiological time-series data. An adaptive smoothing difference feature extraction algorithm is used. After removing impulse noise from the physiological data through median filtering, the real-time rate of change of the data is calculated through first-order difference operation to extract physiological response features, including heart rate fluctuation coefficient, blood oxygen change rate, and body temperature offset.

3. The wearable adaptive energy management method guided by physiological motion response according to claim 2, characterized in that, The calculation of the physiological motor response correlation index specifically involves quantifying the dynamic correlation between motor characteristics and physiological characteristics using the Pearson correlation coefficient, as shown in the formula: in, For single feature pair correlation, The covariance of motor characteristics and physiological characteristics. , These are the variances of motor characteristics and physiological characteristics, respectively. The correlation degree of multiple sets of features is weighted and fused to calculate the physiological motor response correlation index.

4. The wearable adaptive energy management method guided by physiological motion response according to claim 1, characterized in that, The response deviation analysis model specifically involves: collecting the user's historical physiological motor response correlation index sequence, fitting the historical CI sequence with a Gaussian distribution, and combining it with a time-series decay weighting factor to construct an individualized response stability region, as shown in the formula: in, For the first Each time-decay weight The time decay coefficient, The total number of historical physiological motor response association index sequences. For the weighted steady-state mean, For the first A physiological motor response correlation index, For weighted steady-state standard deviation, , These are the upper and lower limits of the stability region of the individualized response, respectively. This is the individual steady-state tolerance coefficient; The deviation of the real-time physiological motor response is quantified by the deviation between the current response correlation index and the stable region interval, using the following formula: in, The degree of deviation in response at the current moment. The term represents the correlation index of physiological and motor response at the current moment. In the formula, the first term represents the dispersion of the data relative to the steady-state mean, and the second term is the steady-state boundary offset correction coefficient. The larger the value, the greater the deviation of the current physiological motor response state from the user's individualized dynamic steady-state range, and the more significant the abnormal response characteristics.

5. The wearable adaptive energy management method guided by physiological motion response according to claim 1, characterized in that, The physiological motor response prediction model specifically involves: constructing a continuous time-series response correlation index sequence. The LSTM time series prediction model is used to predict the time series of events. Perform iterative predictions and output a predicted sequence of future response correlation indices. ; The response risk level is determined based on the fluctuation amplitude and frequency of deviations from the stable region of the predicted sequence.

6. The wearable adaptive energy management method guided by physiological motion response according to claim 1, characterized in that, The information value indicators are specifically: data anomaly value, data trend value, and data scarcity and uniqueness value are calculated separately using independent algorithms, and then weighted and fused to obtain the overall data information value; Among them, the value of data anomaly is quantified based on the degree of deviation of the current response from the steady-state boundary. When the deviation exceeds the individual steady-state tolerance, the value increases significantly, which represents the high information gain of the abnormal data. The value of data trends is based on the slope of the change in the time series response correlation index to quantify the significance of dynamic trends, and the dynamic evolution value of the current data sequence is characterized by the time series average rate of change. The value of data scarcity and uniqueness is quantified by the proportion of historical steady-state sample distribution, which determines the degree of sample scarcity. Based on the value of data anomalies, the value of data trends, and the value of data scarcity and uniqueness, the information value index is calculated using the following formula.

7. The wearable adaptive energy management method for data acquisition guided by physiological motion response according to claim 6, characterized in that, The calculation of the decision score specifically involves: performing interval normalization on the response deviation degree, response risk level, and data information value to obtain evaluation factors within the same quantification interval; configuring power-law nonlinear transformations for different evaluation factors to achieve adaptive adjustment for low-value weak response and high-value strong gain; combining a multi-indicator joint situational enhancement term to adaptively weight and strengthen highly sensitive scenarios with simultaneous increases in multi-dimensional indicators; and fusing these to obtain a decision score that accurately represents the current monitoring scenario's needs and energy consumption control priorities. The formula is as follows: in, , , These are the normalized values ​​for the degree of response deviation, risk level, and data information value, respectively. These are the corresponding normalized weights. For dynamic enhancement coefficients, It is a very small constant; the first half of the formula is the weighted value of indicators of different dimensions, and the second half is the joint situation enhancement term, which adaptively amplifies the decision level when multiple indicators rise at the same time.

8. The wearable adaptive energy management method for data acquisition guided by physiological motion response according to claim 7, characterized in that, The generated energy efficiency management decision parameters are specifically as follows: based on the decision score, according to the three threshold ranges corresponding to low, medium and high risks, the sensor sleep / standby / high precision wake-up level, low / medium / high sampling frequency level, lightweight / conventional / high precision edge computing level, and latency / conventional / real-time data transmission level are matched respectively, so as to realize the linkage and adaptive control of multi-dimensional device parameters. The higher the score, the higher the priority of device monitoring accuracy and real-time performance, and the more aggressive the energy consumption adaptation strategy.

9. The wearable adaptive energy management method guided by physiological motion response according to claim 1, characterized in that, The aforementioned adaptive optimization between monitoring accuracy and equipment energy consumption specifically includes adaptive optimization of monitoring accuracy and adaptive optimization of equipment energy consumption. When the decision score increases, the sampling frequency of the motion sensor and physiological sensor is adaptively increased; when the decision score decreases, the sampling frequency is decreased. The adaptive adjustment formula for the sampling frequency during the process is as follows: in, For adaptive sampling frequency, Equipment basic sampling frequency, This is the frequency adjustment coefficient. Rate the current decision. This is the maximum decision score threshold.

10. The wearable adaptive energy management method for data acquisition guided by physiological motion response according to claim 9, characterized in that, The adaptive optimization of device energy consumption is as follows: the edge computing level is divided into three levels: lightweight computing, regular computing, and high-precision computing. The edge computing resources are dynamically allocated according to the decision parameters. In low-risk scenarios, lightweight filtering and simple feature extraction algorithms are used to reduce computing power consumption. In high-risk scenarios, high-precision response correlation index calculation and trend prediction algorithms are used to ensure data processing accuracy. The formula for controlling the data transmission cycle during the process is: in, For data transmission cycle, Based on the transmission period, S is the transmission adjustment coefficient, S is the current decision score, and Smax is the maximum decision score threshold.

Images