Abnormal monitoring alarm method, device and equipment for wearable smart underwear

Abstract

The present application relates to the field of intelligent wearable device monitoring technology, and discloses an abnormal monitoring alarm method, device and equipment for wearable intelligent underwear. The method includes: determining the spatiotemporal boundary conditions and multi-scale acquisition scheme; collecting physiological data of the wearer to obtain an initial physiological data set; performing static robust adaptive filtering processing to obtain a first physiological data set; performing dynamic robust adaptive filtering processing to obtain a second physiological data set; performing abnormal data detection and abnormal data classification to obtain multiple abnormal data sets; performing abnormal data subclassification processing through a weighted twin support vector machine model to obtain an abnormal subclassification information set; generating an abnormal assessment report based on the abnormal subclassification information set and sending alarm signal data to the wearer, performing tightness adjustment and continuous abnormality monitoring through a tightness adjustment mechanism, and the present application improves the accuracy of abnormal monitoring alarm of wearable intelligent underwear.

Description

Technical Field

[0001] The present application relates to the field of intelligent wearable device monitoring technology, and in particular to an abnormal monitoring and alarm method, device and equipment for wearable intelligent underwear. Background Art

[0002] As people's demand for health monitoring continues to increase, wearable smart devices have received widespread attention and application. As a new type of health monitoring device, wearable smart underwear can collect and monitor the wearer's physiological data in real time, such as heart rate, respiratory rate, body temperature, and exercise posture. However, the wearable smart underwear on the market still has many problems in data collection and processing, and it is difficult to meet the requirements of high precision, real-time and comprehensiveness.

[0003] In the data collection process of existing wearable smart underwear, the accuracy and integrity of data collection are insufficient due to factors such as inconsistent sensor refresh rates and environmental interference. Many devices lack effective spatiotemporal boundary conditions and multi-scale acquisition schemes, and cannot ensure data synchronization and coordination between different sensors, which in turn affects subsequent data processing and analysis. In terms of data processing, traditional signal processing methods have limited suppression effects on noise and outliers, and cannot adapt to the complexity and variability of physiological data. Especially for data such as heart rate and body temperature, the existing methods lack the ability of robust adaptive filtering, resulting in a lot of noise and errors in the processed data. In addition, the detection and classification of abnormal data often rely on simple statistical methods, which cannot accurately identify and distinguish different types of abnormal situations, limiting the reliability and effectiveness of abnormal monitoring. In terms of abnormal monitoring and alarm, existing wearable smart underwear lacks a complete abnormal evaluation and alarm mechanism. Many devices can only provide simple alarm prompts and cannot provide detailed abnormal evaluation reports and classification information, resulting in the wearer being unable to understand the specific situation and severity of the abnormality in a timely manner, affecting the effect of health monitoring. Summary of the invention

[0004] The present application provides a method, device and equipment for abnormal monitoring and alarming of wearable smart underwear, which are used to improve the accuracy of abnormal monitoring and alarming of wearable smart underwear.

[0005] In a first aspect, the present application provides a wearable smart underwear abnormality monitoring and alarm method, the wearable smart underwear abnormality monitoring and alarm method comprising:

[0006] Call multiple sensors in wearable smart underwear and establish a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition;

[0007] Based on the spatiotemporal boundary conditions and the multi-scale acquisition scheme, physiological data of the wearer is collected to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature, and movement posture;

[0008] Performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set;

[0009] Performing dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set;

[0010] performing abnormal data detection and abnormal data classification on the second physiological data set to obtain multiple abnormal data sets;

[0011] Inputting the multiple abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set;

[0012] An abnormality assessment report of the wearable smart underwear is generated according to the abnormal sub-classification information set and an alarm signal data is sent to the wearer. At the same time, the wearable smart underwear is loosened and adjusted and continuously monitored for abnormalities through a looseness adjustment mechanism.

[0013] In a second aspect, the present application provides an abnormality monitoring and alarm device for wearable smart underwear, the abnormality monitoring and alarm device for wearable smart underwear comprising:

[0014] Establish a module for calling multiple sensors in wearable smart underwear and establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition;

[0015] A collection module, used for collecting physiological data of the wearer based on the spatiotemporal boundary conditions and the multi-scale collection scheme to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature and movement posture;

[0016] A static filtering module, used for performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set;

[0017] A dynamic filtering module, used for performing dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set;

[0018] a classification module, configured to perform abnormal data detection and abnormal data classification on the second physiological data set to obtain a plurality of abnormal data sets;

[0019] A processing module, used for inputting the plurality of abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set;

[0020] A generating module is used to generate an abnormal evaluation report of the wearable smart underwear according to the abnormal sub-classification information set and send alarm signal data to the wearer, and at the same time, perform tightness adjustment and continuous abnormality monitoring on the wearable smart underwear through a tightness adjustment mechanism.

[0021] The third aspect of the present application provides an abnormality monitoring and alarm device for wearable smart underwear, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the abnormality monitoring and alarm device for the wearable smart underwear executes the above-mentioned abnormality monitoring and alarm method for the wearable smart underwear.

[0022] In the technical solution provided by the present application, multiple sensors are used for data collection. By establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale collection scheme of data collection, the accuracy and integrity of data collection are improved, and the data synchronization and coordination between different sensors are ensured. Static robust adaptive filtering and dynamic robust adaptive filtering techniques are used to preprocess the initial physiological data, eliminate the noise and outliers in the data, and obtain a more stable and reliable first physiological data set and second physiological data set. The preprocessed data is subjected to anomaly detection. By constructing a wavelet energy density function and a Gaussian mixture model, the abnormal data is accurately identified and classified, and different types of abnormal situations can be effectively distinguished, thereby improving the accuracy and reliability of anomaly detection. The weighted twin support vector machine model is used to perform fine classification processing on the detected abnormal data. Through feature extraction and weighted processing, the classification hyperplane is optimized, the generalization ability and accuracy of the classification model are improved, and detailed abnormal sub-classification information is obtained, which can accurately judge the type and severity of the abnormal situation. An abnormal assessment report is generated based on abnormal classification information, and an alarm signal is sent to the wearer through sound, vibration and mobile phone application notification, ensuring that the wearer can obtain abnormal information in time and take corresponding response measures, thereby improving the practicality and timeliness of health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0024] Figure 1 A schematic diagram of an embodiment of an abnormality monitoring and alarm method for wearable smart underwear in an embodiment of the present application;

[0025] Figure 2Schematic diagram of an embodiment of an abnormality monitoring and alarm device for wearable smart underwear in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The embodiments of the present application provide an abnormal monitoring alarm method, device and equipment for wearable smart underwear. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the abnormal monitoring and alarm method of wearable smart underwear includes:

[0028] Step S101, calling multiple sensors in the wearable smart underwear and establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition;

[0029] It is understandable that the execution subject of the present application can be the abnormal monitoring alarm device of the wearable smart underwear, or it can be a terminal or a server, which is not limited here. The embodiment of the present application is described by taking the server as the execution subject as an example.

[0030] Specifically, the state of the heart rate sensor, breathing sensor, temperature sensor and motion sensor in the wearable smart underwear is initialized. This is achieved by collecting the output data of each sensor at the initial moment. These initial data are used to set the starting state of each sensor and obtain the sensor state data. According to the sensor state data, a discrete time state space model is constructed. In this model, the change of state variables is discretized into the form of time steps, and these state variables are described by differential equations to form a state equation group. The construction of the state equation group is based on the working principle of each sensor and the change law of physiological data. Through these differential equations, the dynamic process of physiological data such as heart rate, breathing rate, body temperature and movement posture changing over time is described. The state equation group is linearized, the nonlinear state equation group is converted into a linear form, the solution process is simplified, and the linearized state equation group is obtained. According to the linearized state equation group, the spatiotemporal boundary of data acquisition is solved to obtain the spatiotemporal boundary conditions. The spatiotemporal boundary conditions define the time range and spatial distribution of data acquisition to ensure the comprehensiveness and accuracy of data acquisition. The multi-scale acquisition parameter set is determined according to the spatiotemporal boundary conditions. The multi-scale acquisition parameter set includes parameters such as the time frequency and spatial resolution of data acquisition, which determine the precision and coverage of data acquisition. According to the multi-scale acquisition parameter set, the multi-scale acquisition scheme for data acquisition is determined. The multi-scale acquisition scheme refers to the specific methods and steps for data acquisition at different time and space scales. Through multi-scale acquisition, more detailed and comprehensive physiological data can be captured to ensure accurate monitoring of the wearer's health status and timely warning of abnormal conditions.

[0031] Step S102, collecting physiological data of the wearer based on the spatiotemporal boundary conditions and the multi-scale collection scheme to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature and movement posture;

[0032] Specifically, the heart rate sensor is parameterized according to the space-time boundary conditions to obtain a heart rate acquisition parameter set. The determination of the heart rate acquisition parameter set needs to take into account factors such as the acquisition frequency, acquisition period, and acquisition accuracy of the heart rate data to ensure the accuracy and timeliness of the heart rate data. The heart rate data of the wearer is collected based on the heart rate acquisition parameter set to obtain a heart rate data set. The respiratory sensor is subjected to wavelet transform processing according to a multi-scale acquisition scheme. Wavelet transform is a mathematical tool suitable for analyzing non-stationary signals, which can capture the characteristics of respiratory signals at different time scales to obtain respiratory data transformation coefficients. The respiratory data transformation coefficients are used to describe the time and frequency characteristics of the respiratory signal. Through these coefficients, the respiratory frequency data of the wearer is collected to obtain a respiratory frequency data set. At the same time, the temperature sensor is calibrated to obtain a temperature sensor calibration coefficient. The temperature sensor is calibrated under known temperature conditions to ensure that the temperature data output by the sensor accurately reflects the actual body temperature of the wearer. According to the temperature sensor calibration coefficient, the body temperature data of the wearer is collected to obtain a body temperature data set. The motion sensor is inertially corrected to obtain an inertial correction parameter. Inertial correction is to correct the data deviation caused by tilt or movement of the motion sensor. By correcting the parameters, the sensor can accurately capture the wearer's motion posture. According to the inertial correction parameters, the wearer's motion posture data is collected to obtain a motion posture dataset. The heart rate dataset, respiratory rate dataset, body temperature dataset and motion posture dataset are time-series aligned and fused to obtain the initial physiological dataset.

[0033] Step S103, performing static robust adaptive filtering processing on the initial physiological data set to obtain a first physiological data set;

[0034] Specifically, a stationarity test is performed on the initial physiological data set to determine whether the data has constant statistical characteristics, such as mean and variance. The first stationarity test result is obtained by the test. According to the result, the non-stationary data is differentially processed to convert it into a stationary time series to obtain a first stationary time series. An autoregressive moving average model (ARMA model) is constructed through the first stationary time series. This model can capture the autocorrelation structure in the data. A Gaussian white noise term is introduced into the ARMA model to form an additive model. The additive model can better simulate the random fluctuations in the data. The autocorrelation function and partial autocorrelation function analysis are performed on the residual term of the additive model to determine the order of the model so that the model can achieve a balance between complexity and fitting effect. According to the determined model order, the parameters of the additive model are estimated to obtain the estimated values ​​of the model parameters. These parameter estimates are substituted into the additive model to construct an adaptive filter. The adaptive filter can self-adjust according to the changes in the input data, thereby improving the filtering effect. In order to enhance the anti-interference ability of the filter, a robust weight function is introduced into the adaptive filter to obtain a robust adaptive filter. The robust weight function can reduce the impact of outliers on the filtering results, thereby improving the robustness of the filter. The initial physiological data set is input into the robust adaptive filter for filtering and data normalization. The filtering process can remove noise and interference in the data, while the normalization process is to adjust the data to the same scale, so as to facilitate subsequent analysis and processing, and finally obtain the first physiological data set.

[0035] Step S104, performing dynamic robust adaptive filtering processing on the first physiological data set to obtain a second physiological data set;

[0036] Specifically, a stationarity test is performed on the first physiological data set to obtain a second stationarity test result. According to the second stationarity test result, the non-stationary data is differentially processed to obtain a second stationary time series. The trend and seasonal components in the data are eliminated to make the data stationary, which is convenient for subsequent analysis and processing. A time-varying autoregressive sliding average model is constructed for the second stationary time series. The model can capture the dynamic change characteristics in the data and adapt to the characteristics of the data changing over time. In order to enhance the filtering ability of the model, the Kalman filter state equation is introduced into the time-varying autoregressive sliding average model to form an extended Kalman filter. Kalman filtering is a recursive filter suitable for state estimation of dynamic systems. By combining the state equation and observation equation of the system, the state estimate can be updated and optimized in real time. In order to improve the robustness of the filter, a robust weight function is introduced into the extended Kalman filter to obtain a robust extended Kalman filter. The robust weight function can reduce the influence of outliers on the filtering results, thereby improving the anti-interference ability and reliability of the filter. The first physiological data set is input into the robust extended Kalman filter for filtering processing to obtain a filtered data sequence. Filtering can remove noise and interference in the data and improve the quality and accuracy of the data. Perform wavelet transform on the filtered data sequence. Wavelet transform is a time-frequency analysis method that can analyze the local characteristics of the signal at different time scales to obtain a wavelet transform coefficient matrix. Wavelet transform can effectively separate different frequency components in the data, which is helpful for further data analysis and processing. By reconstructing the wavelet transform coefficient matrix, the separated frequency components can be recombined to obtain a second physiological data set. The purpose of data reconstruction is to remove noise and outliers while retaining useful signal characteristics to obtain smoother and more reliable physiological data.

[0037] Step S105, performing abnormal data detection and abnormal data classification on the second physiological data set to obtain multiple abnormal data sets;

[0038] Specifically, the second physiological data set is subjected to wavelet transform, the signal is decomposed into different frequency scales, and the local characteristics and spectrum information of the data are obtained. Wavelet decomposition coefficients are obtained through wavelet transform, and these coefficients reflect the energy distribution of physiological data at different scales. According to the wavelet decomposition coefficients, a wavelet energy density function is constructed. The wavelet energy density function can quantify the energy distribution of data at different frequency scales, which is helpful to identify abnormal changes in the data. By analyzing the wavelet energy density function, the energy threshold of the data at different scales is determined. The wavelet energy density function is subjected to threshold processing to obtain an abnormal detection threshold. Values ​​that significantly exceed the normal range in the energy density function are identified, and these values ​​correspond to abnormal points in the data. According to the abnormal detection threshold, abnormal data point detection is performed on the second physiological data set to obtain an abnormal data point set. The abnormal data point set contains all data points identified as abnormal, and these data points may reflect the abnormal physiological state of the wearer. Cluster analysis is performed on the abnormal data point set, and similar abnormal data points are grouped, so as to find the pattern and distribution characteristics of the abnormal data points. Abnormal data clusters are obtained through cluster analysis, and these data clusters represent different types of abnormal patterns. According to the distribution characteristics of the abnormal data clusters, a Gaussian mixture model is constructed. The Gaussian mixture model is a probabilistic model used to describe the distribution of data. It can fit complex data distributions through the linear combination of multiple Gaussian distributions. The expectation maximization algorithm is used to estimate the parameters of the Gaussian mixture model. The expectation maximization algorithm is an iterative optimization algorithm that enables the model to more accurately describe the data distribution by continuously optimizing the model parameters. After obtaining the Gaussian model parameters, the abnormal data clusters are classified according to these parameters to obtain the abnormal data category labels. The abnormal data points with the same category label are divided into an abnormal data set, and these data points represent the same type of abnormal pattern. According to all the abnormal data category labels, multiple abnormal data sets are obtained, each of which represents a specific type of physiological abnormality.

[0039] Step S106: input the multiple abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set;

[0040] Specifically, feature extraction is performed on multiple abnormal data sets, and the original data is converted into a feature set suitable for model training to obtain an abnormal data feature set. According to the abnormal data feature set, a training set and a test set of a weighted twin support vector machine model are constructed. The training set is used for model training, and the test set is used to evaluate the generalization ability of the model. In the weighted twin support vector machine model, two non-parallel classification hyperplanes are set. The two hyperplanes divide the data space into different regions, and the interval region between the two classification hyperplanes is weighted to obtain a weighted interval region. The purpose of weighted processing is to emphasize the importance of data points in the interval region to improve the classification performance of the model. In the weighted interval region, abnormal data points are introduced as abnormal samples, and these abnormal samples are used to train the model so that it can identify and classify abnormal data. Based on the training set, the weighted twin support vector machine model is trained to obtain the target model parameters. The training process optimizes the objective function so that the model can maximize the interval between the classification hyperplanes and minimize the classification error rate, thereby improving the classification performance of the model. According to the target model parameters, the generalization ability of the weighted twin support vector machine model is evaluated on the test set to detect the performance of the model on unseen data. According to the evaluation results, the target model parameters are optimized to achieve the optimal performance of the weighted twin support vector machine model, and the optimized weighted twin support vector machine model is obtained. Through the optimized weighted twin support vector machine model, the decision function is calculated for the target data points in multiple abnormal data sets to obtain the function distance value of each target data point to the two classification hyperplanes. The positive and negative signs of the function distance value are used to determine the preliminary classification result of each target data point, and the preliminary classification result is weighted to calculate the probability value of each target data point falling in the weighted interval area. The probability value reflects the possibility that the data point belongs to a certain category, and the probability threshold is set, and the final classification label of each target data point is determined according to the probability value. Samples with the same classification label are classified into one category to obtain the sub-classification result, and the abnormal sub-classification information set is constructed based on the sub-classification result.

[0041] Step S107: Generate an abnormal assessment report of the wearable smart underwear according to the abnormal sub-classification information set and send alarm signal data to the wearer. At the same time, adjust the tightness of the wearable smart underwear and continuously monitor abnormalities through the tightness adjustment mechanism.

[0042] Specifically, a comprehensive analysis is performed on the abnormal sub-classification information set. The abnormal sub-classification information set contains detailed classification results of various abnormal data, and each type of abnormal data corresponds to a specific physiological abnormality type and characteristics. Through comprehensive analysis of these classification information, the distribution and severity of abnormal data are identified, and it is determined which abnormalities need to be processed immediately and which abnormalities can be temporarily observed. Based on the above analysis results, an abnormal assessment report for wearable smart underwear is generated. The abnormal sub-classification information is sorted and summarized, and different types of abnormalities are sorted and classified according to their severity, frequency of occurrence and possible health effects. A detailed description is provided for each abnormality, including the specific characteristics of the abnormality, the time and frequency of detection, and possible health problems. In the process of report generation, historical data is combined for comparative analysis. If the wearer has a record of wearing smart underwear for a long time, the current abnormal data can be compared with the historical data to analyze the change trend of the abnormality. For example, whether the abnormality of a certain physiological parameter gradually worsens over a period of time, or whether certain abnormal data points appear frequently, etc. Provide targeted suggestions. These suggestions may include reminders to seek medical attention immediately, lifestyle adjustments, drug recommendations, or further medical examinations. The recommendation section is not only based on the current abnormality detection results, but also takes into account the wearer's health history and potential risks. After the abnormality assessment report is generated, the report content is sent to the wearer through the smart underwear's supporting application or platform. At the same time, an alarm signal is set according to the urgency in the report. For serious abnormalities that need to be handled immediately, the system will trigger a high-priority alarm and notify the wearer through sound, vibration, visual prompts, etc., to ensure that they pay attention to potential health risks and take corresponding measures. This application is combined with the tightness adjustment function. When an abnormality is detected in the abnormality assessment report, the wearable smart underwear is adjusted tighter through the tightness adjustment mechanism, which can achieve a better fit between the sensor and the skin and more accurate detection results, so as to achieve continuous cycle monitoring of wearable smart underwear.

[0043] In the embodiment of the present application, multiple sensors are used for data collection. By establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale collection scheme of data collection, the accuracy and integrity of data collection are improved, and data synchronization and coordination between different sensors are ensured. Static robust adaptive filtering and dynamic robust adaptive filtering techniques are used to preprocess the initial physiological data, eliminate the noise and outliers in the data, and obtain a more stable and reliable first physiological data set and second physiological data set. The preprocessed data is subjected to abnormality detection. By constructing a wavelet energy density function and a Gaussian mixture model, the abnormal data is accurately identified and classified, and different types of abnormal situations can be effectively distinguished, thereby improving the accuracy and reliability of abnormality detection. A weighted twin support vector machine model is used to perform fine classification processing on the detected abnormal data. Through feature extraction and weighted processing, the classification hyperplane is optimized, the generalization ability and accuracy of the classification model are improved, and detailed abnormal sub-classification information is obtained, which can accurately judge the type and severity of the abnormal situation. An abnormal assessment report is generated based on abnormal classification information, and an alarm signal is sent to the wearer through sound, vibration and mobile phone application notification, ensuring that the wearer can obtain abnormal information in time and take corresponding response measures, thereby improving the practicality and timeliness of health monitoring.

[0044] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0045] (1) Initialize the state of the heart rate sensor, breathing sensor, temperature sensor and motion sensor in the wearable smart underwear to obtain sensor state data;

[0046] (2) Based on the sensor state data, a discrete-time state-space model is constructed, and in the discrete-time state-space model, the state variables are described by differential equations to obtain a set of state equations;

[0047] (3) Linearizing the state equation group to obtain a linearized state equation group, and solving the space-time boundary of data acquisition based on the linearized state equation group to obtain the space-time boundary conditions;

[0048] (4) Determine a multi-scale acquisition parameter set based on the spatiotemporal boundary conditions, and determine a multi-scale acquisition scheme for data acquisition based on the multi-scale acquisition parameter set.

[0049] Specifically, the heart rate sensor, respiration sensor, temperature sensor and motion sensor in the wearable smart underwear are initialized, and the working state of the sensor is determined by collecting the initial data of the sensor. For the heart rate sensor, the initial state is determined by measuring the wearer's heart rate at rest; for the respiration sensor, it is initialized by detecting the wearer's resting respiration rate; the temperature sensor measures the wearer's body surface temperature; and the motion sensor determines the initial state by capturing the wearer's initial posture and slight movement changes. After these initial data are aggregated, the sensor state data is formed to ensure that each sensor works normally within a predetermined range and provides a reference point for subsequent data processing. Based on the sensor state data, a discrete time state space model is constructed. The state space model is a mathematical model that describes the dynamic behavior of a system and reflects the dynamic characteristics of the system through the relationship between state variables and input variables. Among them, the state variables represent the physiological parameters measured by the sensor, such as heart rate, respiration rate, body temperature and motion posture. The discrete time state space model is constructed to discretize continuous physiological signals for processing in a digital system. The core of this model is the state equation and the observation equation. The state equation describes the evolution of the state variable, and the observation equation describes the relationship between the sensor output and the state variable. The general form of the state equation is:

[0050] ;

[0051] in, is a state variable vector, representing the system state at time step k; is the state transfer matrix, which describes the relationship between state variables; is the input matrix, describing the influence of input variables on the state; is the input variable vector, representing the external input at time step k; is the process noise, which represents the uncertainty of the model. The general form of the observation equation is:

[0052] ;

[0053] in, is the observed variable vector, representing the sensor output at time step k; is the observation matrix, describing the influence of state variables on sensor output; is the direct transfer matrix, describing the direct impact of the input variables on the sensor output; is the observation noise, which represents the uncertainty of sensor measurement. Linearize the state equations to simplify calculation and analysis. When the state equation is nonlinear, it is approximated into a linear form by methods such as Taylor expansion. Determine the multi-scale acquisition parameter set. The multi-scale acquisition parameter set includes the data acquisition frequency and resolution at different time scales and spatial scales. For example, heart rate data can be collected once per second to capture rapidly changing heart rate information; respiratory rate data can be collected once per minute to capture the rhythm of breathing; body temperature data can be collected once every ten minutes to reflect slow changes in body temperature; and motion posture data can be collected multiple times per second to capture subtle changes in motion. By setting these parameters, a multi-scale acquisition scheme is formed, making data acquisition adaptable and flexible at different time scales and spatial scales.

[0054] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0055] (1) Setting parameters of the heart rate sensor according to the spatiotemporal boundary conditions to obtain a heart rate collection parameter set, and collecting heart rate data of the wearer according to the heart rate collection parameter set to obtain a heart rate data set;

[0056] (2) Performing wavelet transform processing on the respiratory sensor according to the multi-scale acquisition scheme to obtain the respiratory data transformation coefficient, and collecting the wearer's respiratory frequency data according to the respiratory data transformation coefficient to obtain a respiratory frequency data set;

[0057] (3) Calibrate the temperature sensor to obtain the temperature sensor calibration coefficient, and collect the wearer's body temperature data based on the temperature sensor calibration coefficient to obtain a body temperature data set;

[0058] (4) Performing inertia correction on the motion sensor to obtain inertia correction parameters, and collecting motion posture data of the wearer based on the inertia correction parameters to obtain a motion posture data set;

[0059] (5) Perform time series alignment and data fusion on the heart rate dataset, respiratory rate dataset, body temperature dataset, and movement posture dataset to obtain the initial physiological dataset.

[0060] Specifically, the heart rate sensor is parameterized according to the spatiotemporal boundary conditions. The spatiotemporal boundary conditions define the time range and spatial distribution of data collection, thereby ensuring that data collection can fully and accurately reflect the wearer's physiological state. Through these boundary conditions, the acquisition frequency and other necessary parameters of the heart rate sensor are determined to obtain a heart rate acquisition parameter set. For example, if the heart rate data needs to be collected once per second, the sampling frequency of the heart rate sensor is set to 1 Hz. In this way, it is ensured that the heart rate sensor continuously collects data within a predetermined time interval. According to the set heart rate acquisition parameter set, the heart rate data of the wearer is collected to obtain a heart rate data set. The respiratory sensor is subjected to wavelet transform processing according to the multi-scale acquisition scheme. Wavelet transform is a time-frequency analysis method that can effectively capture the local characteristics of the signal and is suitable for analyzing non-stationary signals. Through wavelet transform processing, the respiratory data transformation coefficients are obtained, which reflect the characteristics of the respiratory signal at different scales. According to these transformation coefficients, the respiratory frequency data is accurately collected. For example, the acquisition frequency of the respiratory frequency data is set to once per minute, and then the acquisition frequency and time interval are adjusted according to the results of the wavelet transform to ensure that the collected data can accurately reflect the changes in the wearer's respiratory frequency, and obtain a respiratory frequency data set. Calibrate the temperature sensor. Calibrate the sensor by measuring under known temperature conditions to ensure that the output of the sensor accurately reflects the actual temperature. Through the calibration process, the calibration coefficients of the temperature sensor are obtained. After the calibration is completed, the wearer's body temperature data is collected according to these calibration coefficients to obtain a body temperature data set. Perform inertia correction on the motion sensor. By correcting the data deviation caused by the sensor's tilt or movement, ensure that the data output by the sensor accurately reflects the actual motion state. Through the inertia correction process, the inertia correction parameters are obtained. The wearer's motion posture data is collected according to these correction parameters. For example, if the motion sensor is affected by tilt during measurement, the correction parameters are used to correct the measurement results to obtain accurate motion posture data. Perform time series alignment and data fusion on the heart rate data set, respiratory rate data set, body temperature data set, and motion posture data set. Time series alignment refers to synchronizing the data of different sensors according to the same time base to ensure that the time corresponding to each data point is consistent. For example, if the heart rate data is collected once a second and the body temperature data is collected every ten minutes, the two data sets need to be aligned to ensure that they can match at the same time point. Data fusion refers to the integration of data from different sensors to form a complete initial physiological data set.

[0061] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0062] (1) Performing a stationary test on the initial physiological data set to obtain a first stationary test result, and performing differential processing on the non-stationary data based on the first stationary test result to obtain a first stationary time series;

[0063] (2) An autoregressive moving average model is constructed through the first stationary time series, and a Gaussian white noise term is introduced into the autoregressive moving average model to obtain an additive model;

[0064] (3) Analyze the autocorrelation function and partial autocorrelation function of the residual term of the additive model to determine the model order;

[0065] (4) Estimate the parameters of the additive model according to the model order, obtain the estimated values ​​of the model parameters, and substitute the estimated values ​​of the model parameters into the additive model to obtain the adaptive filter;

[0066] (5) A robust weight function is introduced into the adaptive filter to obtain a robust adaptive filter, and the initial physiological data set is input into the robust adaptive filter for filtering and data normalization to obtain a first physiological data set.

[0067] Specifically, the data set is preprocessed to ensure that it is suitable for time series analysis. The purpose of the stationarity test is to determine whether the data has constant statistical properties, such as mean and variance. Common stationarity test methods include unit root tests, such as ADF tests. By performing an ADF test on the initial physiological data set, the first stationarity test result is obtained. If the test result shows that the data is non-stationary, differential processing is required to eliminate non-stationarity and transform the data into a stationary time series. Assume that the initial heart rate data series is If the ADF test result shows If it is non-stationary, then difference processing is performed to obtain the first stationary time series ,Right now:

[0068] ;

[0069] Through differential processing, the trend and seasonal components in the data are eliminated and converted into a stationary time series. Construct an autoregressive moving average model (ARMA model). The ARMA model is a statistical model commonly used in time series analysis, combining the ideas of autoregression (AR) and moving average (MA). The autoregressive part uses past values ​​to predict the future, and the moving average part uses past prediction errors to adjust the prediction results. When constructing the ARMA model, the Gaussian white noise term is introduced. As random errors, we get the additive model:

[0070] ;

[0071] in, are the parameters of the autoregressive part, is the parameter of the sliding average part, is a Gaussian white noise term. The autocorrelation function and partial autocorrelation function analysis of the residual term of the additive model are performed to determine the order of the model, that is, the order of the autoregressive and sliding average parts. and . The autocorrelation function reflects the correlation between each time point in the sequence, while the partial autocorrelation function is used to identify higher-level correlation structures. By analyzing the autocorrelation function and partial autocorrelation function graphs, the most suitable order is determined, thereby constructing a suitable ARMA model. After determining the model order, the parameters of the additive model are estimated according to these orders to obtain the model parameter estimates. Parameter estimation can be achieved by maximum likelihood estimation or least squares method. These parameter estimates are substituted into the additive model to obtain an adaptive filter. The adaptive filter can adjust the filter parameters in real time according to the changes in the input data, thereby improving the filtering effect. In order to improve the anti-interference ability and robustness of the filter, a robust weight function is introduced into the adaptive filter to obtain a robust adaptive filter. The robust weight function is used to reduce the impact of outliers on the filtering results and ensure that the filter can still work effectively in the presence of noise and outliers. The initial physiological data set is input into the robust adaptive filter for filtering and data normalization. The filtering process can remove noise and interference in the data, and the normalization process can adjust the data to the same scale for subsequent analysis.

[0072] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0073] (1) Performing a stationarity test on the first physiological data set to obtain a second stationarity test result;

[0074] (2) Perform differential processing on non-stationary data based on the results of the second stationarity test to obtain the second stationary time series;

[0075] (3) A time-varying autoregressive moving average model is constructed for the second stationary time series, and the Kalman filter state equation is introduced into the time-varying autoregressive moving average model to obtain an extended Kalman filter;

[0076] (4) introducing a robust weight function into the extended Kalman filter to obtain a robust extended Kalman filter, and inputting the first physiological data set into the robust extended Kalman filter for filtering to obtain a filtered data sequence;

[0077] (5) Performing wavelet transform on the filtered data sequence to obtain a wavelet transform coefficient matrix, and reconstructing the wavelet transform coefficient matrix to obtain a second physiological data set.

[0078] Specifically, a suitable statistical method is used to determine the stationarity of the data. Common methods for stationarity tests include unit root tests, such as ADF tests. By performing an ADF test on the first physiological data set, it can be determined whether the data has statistical characteristics such as a constant mean and variance. After obtaining the second stationarity test result, if the result shows that the data is non-stationary, differential processing is required. The purpose of differential processing is to eliminate the trend and seasonal components in the data by calculating the difference between consecutive observations, thereby converting the data into a stationary time series. A time-varying autoregressive moving average model is constructed through the second stationary time series. The time-varying autoregressive moving average model can capture the dynamic change characteristics in time series data. The state equation and observation equation of the time-varying autoregressive moving average model respectively describe the relationship between the system state and the observed data. In the time-varying autoregressive moving average model, the Kalman filter state equation is introduced to form an extended Kalman filter. Kalman filtering is a recursive algorithm suitable for dynamic system state estimation, which can provide optimal estimation under noise and uncertainty conditions. The state equation of the model can be expressed as:

[0079] ;

[0080] in, represents the state variable vector of the system, is the state transition matrix, is the control input matrix, is the control input vector, is the process noise. The observation equation is expressed as:

[0081] ;

[0082] in, represents the observed variable vector, is the observation matrix, is the observation noise. Through these equations, the extended Kalman filter can recursively update the system state and provide the optimal state estimation. In order to enhance the anti-interference ability of the filter, a robust weight function is introduced into the extended Kalman filter to obtain a robust extended Kalman filter. The introduction of the robust weight function can reduce the influence of outliers on the filtering results, thereby improving the robustness of the filter. The first physiological data set is input into the robust extended Kalman filter for filtering, which can effectively remove the noise and interference in the data and obtain a smoother filtered data sequence. The filtered data sequence is subjected to wavelet transform. Wavelet transform is a time-frequency analysis tool that can analyze the local characteristics of the signal at different time scales and is suitable for analyzing non-stationary signals. Through wavelet transform, the filtered data is decomposed into different frequency scales to obtain a wavelet transform coefficient matrix. These coefficients reflect the characteristics of the data at each frequency scale. Data reconstruction is performed according to the wavelet coefficient matrix, and the wavelet coefficients at different scales are combined to restore the characteristics of the original signal. The reconstructed data sequence is smoother and noise-free, and the second physiological data set is obtained.

[0083] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0084] (1) performing wavelet transform on the second physiological data set to obtain wavelet decomposition coefficients, and constructing a wavelet energy density function based on the wavelet decomposition coefficients;

[0085] (2) performing threshold processing on the wavelet energy density function to obtain an abnormality detection threshold, and performing abnormal data point detection on the second physiological data set based on the abnormality detection threshold to obtain an abnormal data point set;

[0086] (3) Perform cluster analysis on the set of abnormal data points to obtain abnormal data clusters, and construct a Gaussian mixture model based on the distribution characteristics of the abnormal data clusters;

[0087] (4) Use the expectation maximization algorithm to estimate the parameters of the Gaussian mixture model to obtain the Gaussian model parameters, and classify the abnormal data clusters according to the Gaussian model parameters to obtain the abnormal data category labels;

[0088] (5) The abnormal data points with the same category label are divided into an abnormal data set, and multiple abnormal data sets are obtained based on all abnormal data category labels.

[0089] Specifically, the second physiological data set is subjected to wavelet transform to obtain wavelet decomposition coefficients. Wavelet transform is a time-frequency analysis tool that can analyze signals in both the time domain and the frequency domain and is suitable for processing non-stationary signals. Through wavelet transform, the second physiological data set is decomposed into details and approximation parts at different scales to obtain wavelet decomposition coefficients. Assume that the second physiological data set is , a series of wavelet coefficients can be obtained through wavelet transform ,in Indicates scale, Indicates the time position. The result of wavelet transform can reflect the energy distribution characteristics of data at different time scales. According to the obtained wavelet decomposition coefficients, the wavelet energy density function is constructed. The wavelet energy density function represents the energy distribution by calculating the square sum of the wavelet coefficients at each scale. Wavelet energy density function It can be expressed as:

[0090] ;

[0091] in, Indicated in scale The energy density on is in scale and time location The wavelet coefficients on . By calculating the wavelet energy density function, the energy distribution at each scale is obtained, so as to identify abnormal data points. The wavelet energy density function is thresholded to determine the abnormal detection threshold. The purpose of threshold processing is to identify values ​​whose energy density significantly exceeds the normal range, which correspond to abnormal points in the data. A threshold is set, and when the wavelet energy density exceeds the threshold, the corresponding time point is marked as an abnormal data point. In this way, abnormal data points are detected from the second physiological data set to form a set of abnormal data points. Cluster analysis is performed on the set of abnormal data points to group similar abnormal data points and find the pattern and distribution characteristics of the abnormal data points. Commonly used methods for cluster analysis include K-means clustering and DBSCAN. Through cluster analysis, abnormal data clusters are obtained, which represent different types of abnormal patterns. According to the distribution characteristics of the abnormal data clusters, a Gaussian mixture model is constructed. The Gaussian mixture model is a probability model used to describe the distribution of data. The Gaussian mixture model assumes that the data is composed of a linear combination of multiple Gaussian distributions, and each Gaussian distribution corresponds to a data cluster. The probability density function of the Gaussian mixture model can be expressed as:

[0092] ;

[0093] in, is the number of Gaussian distributions, It is The mixing coefficient of the Gaussian distribution, is the mean vector, is the covariance matrix. The expectation maximization (EM) algorithm is used to estimate the parameters of the Gaussian mixture model. The EM algorithm is an iterative optimization algorithm that continuously optimizes the model parameters so that the model can more accurately describe the data distribution. In the EM algorithm, the E step calculates the expectation and the M step maximizes the expectation. By alternating the E step and the M step, the model parameters are gradually optimized. After the EM algorithm is iterated, the Gaussian model parameters are obtained. , and . According to the Gaussian model parameters, the abnormal data clusters are classified to obtain abnormal data category labels. Each data point is classified according to its probability of belonging to different Gaussian distributions to form abnormal data category labels. Abnormal data points with the same category label are divided into an abnormal data set. For example, some abnormal points in the heart rate data may be classified as high heart rate, while other abnormal points may be classified as unstable heart rate. Through this classification, the abnormal data points are divided into different abnormal data sets, each of which represents a specific type of abnormality. According to all abnormal data category labels, multiple abnormal data sets are obtained. For example, in a wearable smart underwear monitoring system, it is assumed that two types of abnormalities appear in the heart rate data set, one is high heart rate and the other is unstable heart rate. Through wavelet transform and Gaussian mixture model analysis, the abnormal points of high heart rate are clustered into one abnormal data set, and the abnormal points of unstable heart rate are clustered into another abnormal data set. Finally, the system generates two abnormal data sets, each corresponding to an abnormal type, and records their respective abnormal data points.

[0094] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0095] (1) Extract features from multiple abnormal data sets to obtain abnormal data feature sets;

[0096] (2) Based on the abnormal data feature set, construct the training set and test set of the weighted twin support vector machine model;

[0097] (3) In the weighted twin support vector machine model, two non-parallel classification hyperplanes are set, and the interval area between the two classification hyperplanes is weighted to obtain a weighted interval area;

[0098] (4) In the weighted interval region, abnormal data points are introduced as abnormal samples, and the weighted twin support vector machine model is trained based on the training set to obtain the target model parameters;

[0099] (5) According to the target model parameters, the generalization ability of the weighted twin support vector machine model is evaluated on the test set, and the target model parameters are optimized so that the weighted twin support vector machine model achieves the optimal performance, thereby obtaining the optimized weighted twin support vector machine model;

[0100] (6) The optimized weighted twin support vector machine model is used to calculate the decision function of the target data points in multiple abnormal data sets to obtain the function distance value from each target data point to the two classification hyperplanes;

[0101] (7) Determine the preliminary classification result of each target data point based on the sign of the function distance value, perform weighted processing on the preliminary classification results, and calculate the probability value of each target data point falling in the weighted interval area;

[0102] (8) Set the probability threshold, determine the final classification label of each target data point according to the probability value, and classify the samples with the same classification label into one category to obtain the sub-classification results. Based on the sub-classification results, construct the abnormal sub-classification information set.

[0103] Specifically, feature analysis is performed on the data points in each abnormal data set, and the original data is converted into a feature set suitable for model training. These features may include time domain features, frequency domain features, and other statistical features. For example, for the heart rate abnormality data set, features such as average heart rate, heart rate variability, maximum heart rate, minimum heart rate, etc. are extracted; for the respiratory rate abnormality data set, features such as average respiratory rate, respiratory rate variability, etc. are extracted. In this way, abnormal data feature sets are obtained, which can fully reflect the abnormal features of each data point. According to the abnormal data feature set, the training set and test set of the weighted twin support vector machine model are constructed. The training set is used to train the model, while the test set is used to evaluate the performance of the model. The abnormal data feature set is randomly divided into two parts, one for training and the other for testing. This ensures the independence of model training and testing and improves the reliability of model evaluation. In the weighted twin support vector machine model, two non-parallel classification hyperplanes are set, which divide the data space into different regions. The support vector machine seeks a hyperplane that can maximize the interval, while the weighted twin support vector machine enhances the classification ability of the model by introducing weights. Set two classification hyperplanes and ,in and is the normal vector of the hyperplane, and is the bias term. The interval area between the two classification hyperplanes is weighted to obtain a weighted interval area, which enhances the classification model's ability to distinguish different types of data points, especially the ability to identify abnormal data points. In the weighted interval area, abnormal data points are introduced as abnormal samples, and the weighted twin support vector machine model is trained based on the training set. The purpose of training is to optimize the model parameters so that the classification hyperplane can maximize the interval and correctly classify abnormal samples in the training data. Through training, the target model parameters are obtained. , , and . According to the target model parameters, the generalization ability of the weighted twin support vector machine model is evaluated on the test set. The purpose of generalization ability evaluation is to detect the performance of the model on unseen data to ensure that the model has good predictive ability. Through evaluation on the test set, the model's classification accuracy, recall rate, F1 value and other indicators are calculated. According to the evaluation results, the target model parameters are optimized to achieve the optimal performance of the weighted twin support vector machine model, and the optimized weighted twin support vector machine model is obtained. Through the optimized weighted twin support vector machine model, the decision function is calculated for the target data points in multiple abnormal data sets to obtain the function distance value of each target data point to the two classification hyperplanes. The calculation formula of the decision function is:

[0104] ;

[0105] in, Represents data points The function distance value to the two classification hyperplanes, and They are the normal vectors and The norm of . According to the positive and negative signs of the function distance value, the preliminary classification result of each target data point is determined. A positive value indicates that the data point is closer to one classification hyperplane, and a negative value indicates that it is closer to another classification hyperplane. The preliminary classification results are weighted to calculate the probability value of each target data point falling in the weighted interval area. Weighted processing enhances the model's ability to distinguish by assigning different weights to different data points, especially for data points on the boundary. In this way, the classification probability of each data point can be calculated more accurately. Set a probability threshold and determine the final classification label of each target data point based on the probability value. For example, set a probability threshold. When the classification probability of a data point is greater than the threshold, it is marked as one category, otherwise it is marked as another category. In this way, samples with the same classification label are classified into one category to obtain a detailed classification result. Finally, based on the detailed classification results, an abnormal detailed classification information set is constructed.

[0106] The above describes the abnormal monitoring alarm method of the wearable smart underwear in the embodiment of the present application. The following describes the abnormal monitoring alarm device of the wearable smart underwear in the embodiment of the present application. Figure 2 In the embodiment of the present application, an abnormal monitoring and alarm device for wearable smart underwear includes:

[0107] Establishing module 201, for calling multiple sensors in the wearable smart underwear and establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition;

[0108] The acquisition module 202 is used to acquire physiological data of the wearer based on the spatiotemporal boundary conditions and the multi-scale acquisition scheme to obtain an initial physiological data set, where the initial physiological data set includes heart rate, respiratory rate, body temperature and movement posture;

[0109] A static filtering module 203, used for performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set;

[0110] A dynamic filtering module 204, configured to perform dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set;

[0111] A classification module 205, configured to perform abnormal data detection and abnormal data classification on the second physiological data set to obtain a plurality of abnormal data sets;

[0112] The processing module 206 is used to input the multiple abnormal data sets into the preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set;

[0113] The generating module 207 is used to generate an abnormal evaluation report of the wearable smart underwear according to the abnormal sub-classification information set and send alarm signal data to the wearer. At the same time, the wearable smart underwear is adjusted in tension and continuously monitored for abnormalities through the tension adjustment mechanism.

[0114] Through the cooperation of the above components, multiple sensors are used for data collection. By establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme of data collection, the accuracy and integrity of data collection are improved, and the data synchronization and coordination between different sensors are ensured. Static robust adaptive filtering and dynamic robust adaptive filtering techniques are used to preprocess the initial physiological data, eliminate the noise and outliers in the data, and obtain more stable and reliable first and second physiological data sets. Anomaly detection is performed on the preprocessed data. By constructing wavelet energy density function and Gaussian mixture model, abnormal data can be accurately identified and classified, which can effectively distinguish different types of abnormal situations and improve the accuracy and reliability of anomaly detection. The weighted twin support vector machine model is used to perform fine classification processing on the detected abnormal data. Through feature extraction and weighted processing, the classification hyperplane is optimized, the generalization ability and accuracy of the classification model are improved, and detailed abnormal sub-classification information is obtained, which can accurately judge the type and severity of abnormal situations. An abnormal assessment report is generated based on abnormal classification information, and an alarm signal is sent to the wearer through sound, vibration and mobile phone application notification, ensuring that the wearer can obtain abnormal information in time and take corresponding response measures, thereby improving the practicality and timeliness of health monitoring.

[0115] The present application also provides an abnormality monitoring and alarm device for wearable smart underwear, the abnormality monitoring and alarm device for wearable smart underwear includes a memory and a processor, the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the abnormality monitoring and alarm method for wearable smart underwear in the above-mentioned embodiments.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0117] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0118] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A wearable smart underwear abnormal monitoring and alarm method, characterized in that: The method comprises: Call multiple sensors in wearable smart underwear and establish a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition; Based on the spatiotemporal boundary conditions and the multi-scale acquisition scheme, physiological data of the wearer is collected to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature, and movement posture; Performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set; Performing dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set; performing abnormal data detection and abnormal data classification on the second physiological data set to obtain multiple abnormal data sets; Inputting the multiple abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set; An abnormality assessment report of the wearable smart underwear is generated according to the abnormal sub-classification information set and an alarm signal data is sent to the wearer. At the same time, the wearable smart underwear is loosened and adjusted and continuously monitored for abnormalities through a looseness adjustment mechanism.

2. The abnormal monitoring and alarm method for wearable smart underwear according to claim 1 is characterized in that: The method of calling multiple sensors in the wearable smart underwear and establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition includes: Initialize the states of the heart rate sensor, breathing sensor, temperature sensor and motion sensor in the wearable smart underwear to obtain sensor state data; According to the sensor state data, a discrete time state space model is constructed, and in the discrete time state space model, a differential equation is described for the state variables to obtain a state equation group; Linearizing the state equation group to obtain a linearized state equation group, and solving the space-time boundary of data acquisition according to the linearized state equation group to obtain space-time boundary conditions; A multi-scale acquisition parameter set is determined according to the spatiotemporal boundary conditions, and a multi-scale acquisition scheme for data acquisition is determined according to the multi-scale acquisition parameter set.

3. The abnormal monitoring and alarm method for wearable intelligent underwear according to claim 2 is characterized in that: The physiological data of the wearer is collected based on the spatiotemporal boundary conditions and the multi-scale collection scheme to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature and movement posture, including: Setting parameters of the heart rate sensor according to the spatiotemporal boundary conditions to obtain a heart rate collection parameter set, and collecting heart rate data of the wearer according to the heart rate collection parameter set to obtain a heart rate data set; Performing wavelet transform processing on the respiratory sensor according to the multi-scale acquisition scheme to obtain a respiratory data transformation coefficient, and collecting respiratory frequency data of the wearer according to the respiratory data transformation coefficient to obtain a respiratory frequency data set; Calibrate the temperature sensor to obtain a temperature sensor calibration coefficient, and collect body temperature data of the wearer according to the temperature sensor calibration coefficient to obtain a body temperature data set; Performing inertia correction on the motion sensor to obtain inertia correction parameters, and collecting motion posture data of the wearer according to the inertia correction parameters to obtain a motion posture data set; The heart rate dataset, the respiratory frequency dataset, the body temperature dataset and the movement posture dataset are time-series aligned and fused to obtain an initial physiological dataset.

4. The abnormal monitoring and alarm method for wearable smart underwear according to claim 1, characterized in that: The performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set includes: Performing a stationary test on the initial physiological data set to obtain a first stationary test result, and performing a non-stationary data differential process according to the first stationary test result to obtain a first stationary time series; An autoregressive moving average model is constructed through the first stationary time series, and a Gaussian white noise term is introduced into the autoregressive moving average model to obtain an additive model; Performing autocorrelation function and partial autocorrelation function analysis on the residual term of the additive model to determine the model order; According to the model order, the parameters of the additive model are estimated to obtain model parameter estimation values, and the model parameter estimation values ​​are substituted into the additive model to obtain an adaptive filter; A robust weight function is introduced into the adaptive filter to obtain a robust adaptive filter, and the initial physiological data set is input into the robust adaptive filter for filtering and data normalization to obtain a first physiological data set.

5. The abnormal monitoring and alarm method for wearable intelligent underwear according to claim 4 is characterized in that: The step of performing dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set includes: Performing a stationarity test on the first physiological data set to obtain a second stationarity test result; Performing non-stationary data difference processing according to the second stationary test result to obtain a second stationary time series; Constructing a time-varying autoregressive moving average model for the second stationary time series, and introducing a Kalman filter state equation into the time-varying autoregressive moving average model to obtain an extended Kalman filter; Introducing a robust weight function into the extended Kalman filter to obtain a robust extended Kalman filter, and inputting the first physiological data set into the robust extended Kalman filter for filtering to obtain a filtered data sequence; The filtered data sequence is subjected to wavelet transform to obtain a wavelet transform coefficient matrix, and the wavelet transform coefficient matrix is ​​subjected to data reconstruction to obtain a second physiological data set.

6. The abnormal monitoring and alarm method for wearable smart underwear according to claim 1, characterized in that: The performing abnormal data detection and abnormal data classification on the second physiological data set to obtain multiple abnormal data sets includes: Performing wavelet transform on the second physiological data set to obtain wavelet decomposition coefficients, and constructing a wavelet energy density function according to the wavelet decomposition coefficients; Performing threshold processing on the wavelet energy density function to obtain an abnormality detection threshold, and performing abnormal data point detection on the second physiological data set according to the abnormality detection threshold to obtain an abnormal data point set; Performing cluster analysis on the abnormal data point set to obtain abnormal data clusters, and constructing a Gaussian mixture model based on the distribution characteristics of the abnormal data clusters; Using an expectation maximization algorithm to estimate the parameters of the Gaussian mixture model to obtain Gaussian model parameters, and classifying the abnormal data clusters according to the Gaussian model parameters to obtain abnormal data category labels; The abnormal data points with the same category label are divided into one abnormal data set, and multiple abnormal data sets are obtained according to all abnormal data category labels.

7. The abnormal monitoring and alarm method for wearable smart underwear according to claim 1 is characterized in that: The step of inputting the plurality of abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set includes: Extracting features from the multiple abnormal data sets to obtain an abnormal data feature set; Constructing a training set and a test set of a weighted twin support vector machine model according to the abnormal data feature set; In the weighted twin support vector machine model, two non-parallel classification hyperplanes are set, and the interval area between the two classification hyperplanes is weighted to obtain a weighted interval area; In the weighted interval region, abnormal data points are introduced as abnormal samples, and based on the training set, the weighted twin support vector machine model is trained to obtain target model parameters; According to the target model parameters, the generalization ability of the weighted twin support vector machine model is evaluated on the test set, and the target model parameters are optimized so that the weighted twin support vector machine model achieves optimal performance, thereby obtaining an optimized weighted twin support vector machine model; Performing decision function calculation on the target data points in the multiple abnormal data sets by using the optimized weighted twin support vector machine model to obtain a function distance value from each target data point to two classification hyperplanes; Determine the preliminary classification result of each target data point according to the positive and negative signs of the function distance value, and perform weighted processing on the preliminary classification result to calculate the probability value of each target data point falling in the weighted interval area; A probability threshold is set, and the final classification label of each target data point is determined according to the probability value, and samples with the same classification label are classified into one category to obtain a detailed classification result, and an abnormal detailed classification information set is constructed according to the detailed classification result.

8. An abnormality monitoring and alarm device for wearable smart underwear, used to execute the abnormality monitoring and alarm method for wearable smart underwear as claimed in claim 1, characterized in that: The device comprises: Establish a module for calling multiple sensors in wearable smart underwear and establishing a discrete time state space model to determine the spatiotemporal boundary conditions and multi-scale acquisition scheme for data acquisition; A collection module, used for collecting physiological data of the wearer based on the spatiotemporal boundary conditions and the multi-scale collection scheme to obtain an initial physiological data set, wherein the initial physiological data set includes heart rate, respiratory rate, body temperature and movement posture; A static filtering module, used for performing static robust adaptive filtering on the initial physiological data set to obtain a first physiological data set; A dynamic filtering module, used for performing dynamic robust adaptive filtering on the first physiological data set to obtain a second physiological data set; a classification module, configured to perform abnormal data detection and abnormal data classification on the second physiological data set to obtain a plurality of abnormal data sets; A processing module, used for inputting the plurality of abnormal data sets into a preset weighted twin support vector machine model to perform abnormal data sub-classification processing to obtain an abnormal sub-classification information set; A generating module is used to generate an abnormal evaluation report of the wearable smart underwear according to the abnormal sub-classification information set and send alarm signal data to the wearer, and at the same time, perform tightness adjustment and continuous abnormality monitoring on the wearable smart underwear through a tightness adjustment mechanism.

9. An abnormal monitoring and alarm device for wearable smart underwear, characterized in that: The abnormal monitoring and alarm device of the wearable smart underwear comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instruction in the memory so that the abnormality monitoring and alarm device of the wearable intelligent underwear executes the abnormality monitoring and alarm method of the wearable intelligent underwear as described in any one of claims 1-7.

Images