Intelligent fiber clothing remote health monitoring and abnormal state early warning method and system

Abstract

The invention provides an intelligent fiber clothing remote health monitoring and abnormal state early warning method and system, and relates to the field of intelligent wearable equipment, and the method comprises the steps: collecting standardized physiological data, constructing a multi-dimensional collaborative analysis matrix and a physiological index causal network, recognizing an abnormal source index and a diffusion path, and generating an abnormal conduction topological graph. Setting multiple levels of early warning thresholds, forming an early warning rule table, monitoring the physiological data in real time, and when the indexes exceed the thresholds, determining the early warning level according to the association degree and sending an early warning signal. According to the invention, early accurate early warning of the abnormal health state can be realized, the monitoring efficiency is improved, and the false alarm rate is reduced.

Description

Technical Field

[0001] The present invention relates to smart wearable device technology, and in particular to a method and system for remote health monitoring and abnormal state early warning of smart fiber clothing. Background Art

[0002] With the development of wearable technology, smart fiber clothing based on textile-integrated sensors is increasingly being used in the health monitoring field. These garments can continuously collect a variety of physiological parameters, such as the wearer's heart rate, respiratory rate, and body temperature, during daily wear. Compared to traditional wristband and patch-based devices, they offer advantages such as a wide monitoring range, improved comfort, and high wear compliance. However, existing technologies still have significant limitations in data processing and early warning mechanisms. Most systems lack the ability to collaboratively analyze the various types of collected physiological data and often only set thresholds for single indicators to identify anomalies, making it difficult to promptly identify potential health risks. Furthermore, physiological indicators exhibit complex causal relationships and dynamic changes, making a single threshold trigger mechanism difficult to reflect the development path and transmission characteristics of anomalies, which can easily lead to delayed warnings or frequent false alarms. Furthermore, existing early warning mechanisms often lack a graded response strategy, making it difficult to take targeted measures based on the severity and spread of anomalies. Therefore, there is an urgent need for an intelligent fiber clothing monitoring method that can integrate multi-dimensional data collaborative modeling, causal network analysis, and abnormal conduction topology construction to achieve accurate identification of abnormal conditions and multi-level early warning, and send early warning information to the remote monitoring terminal in real time through wireless communication to improve the accuracy and timeliness of health monitoring. Summary of the Invention

[0003] The embodiments of the present invention provide a method and system for remote health monitoring and abnormal state early warning of smart fiber clothing, which can solve the problems in the prior art.

[0004] A first aspect of an embodiment of the present invention provides a method for remote health monitoring and abnormal state warning of smart fiber clothing, comprising:

[0005] Collect the wearer's physiological signal data and preprocess it to obtain standardized physiological data;

[0006] Construct a multidimensional physiological data collaborative analysis matrix based on standardized physiological data;

[0007] Based on the multi-dimensional physiological data collaborative analysis matrix, a physiological indicator causal network is constructed. The influence weights and propagation delays between each physiological indicator in the physiological indicator causal network are calculated. An indicator association data table is constructed. Based on the indicator association data table, abnormal source indicators are identified. Starting from the abnormal source indicator, the abnormal diffusion path is determined according to the numerical value of the propagation delay, and an abnormal conduction topology map is generated.

[0008] Set multi-level warning trigger thresholds based on the abnormal conduction topology diagram and generate a warning rule table containing warning trigger conditions and warning levels;

[0009] The wearer's standardized physiological data is monitored in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

[0010] In an optional embodiment,

[0011] Collecting the wearer's physiological signal data and preprocessing it to obtain standardized physiological data includes:

[0012] Physiological signal data is collected through a multimodal sensor array in the smart fiber clothing. The multimodal sensor array is provided with sensing units at the nodes of the conductive fiber mesh woven structure to collect the original signal waveform output by the sensing units;

[0013] A wavelet basis function is selected based on the frequency distribution of the original signal waveform, the original signal waveform is decomposed using the selected wavelet basis function, an adaptive threshold is calculated based on the local variance of the coefficients of each layer after decomposition, a soft threshold processing is performed on the signal coefficients using the adaptive threshold, and the signal is reconstructed using the selected wavelet basis function to obtain a denoised signal;

[0014] The signal segment length is determined according to the fluctuation characteristics of the denoised signal, the denoised signal is divided into multiple signal segments, the mean and variance of each signal segment are calculated, each signal segment is normalized using the mean and variance, and the normalized adjacent signal segments are smoothly connected to obtain standardized physiological data.

[0015] In an optional embodiment,

[0016] Constructing a multidimensional physiological data collaborative analysis matrix based on standardized physiological data includes:

[0017] Acquiring a local extreme value sequence of standardized physiological data, determining a characteristic period based on a time interval distribution of the local extreme value sequence, and constructing an adaptive decomposition basis function according to the characteristic period;

[0018] Adaptive decomposition basis functions are used to decompose the standardized physiological data to obtain multiple frequency components;

[0019] Calculating the phase correlation between the frequency component and the standardized physiological data, and selecting the frequency component whose phase correlation is greater than a preset correlation threshold as the periodic component;

[0020] Extracting the instantaneous phase from the periodic component as a periodic feature, and determining a time warping constraint according to a temporal variation law of the periodic feature;

[0021] Introducing the time warping constraint into a time warping cost function to limit the deformation range of adjacent sampling points, and calculating an optimal deformation path based on the time warping cost function;

[0022] The standardized physiological data are time-aligned according to the optimal deformation path to obtain time-series aligned physiological data, and the time-series aligned physiological data are reorganized into a multi-dimensional physiological data collaborative analysis matrix.

[0023] In an optional embodiment,

[0024] Based on the multi-dimensional physiological data collaborative analysis matrix, a physiological indicator causal network is constructed, and the influence weights and propagation delays between the physiological indicators in the physiological indicator causal network are calculated. The indicator association data table is constructed, including:

[0025] Divide the physiological indicator data in the multi-dimensional physiological data collaborative analysis matrix into multiple time windows according to the fluctuation period;

[0026] Perform amplitude analysis on the physiological indicator data within each time window, identify the transition time point based on the fluctuation threshold, and divide the physiological indicator data into a stable data interval and a fluctuating data interval;

[0027] Calculate the information entropy value of the physiological indicators in the stable data interval and the information gain value of the physiological indicators in the fluctuating data interval, and determine the information transmission direction between the physiological indicators according to the information entropy value and the information gain value;

[0028] The initial topological structure of the physiological indicator causal network is constructed based on the information transmission direction, and the transmission correlation strength between physiological indicators is obtained by forward recursive verification.

[0029] Classifying the physiological indicators in the physiological indicator causal network according to the transfer association strength, determining the physiological indicators with a transfer association strength higher than a preset strength threshold as source indicators, and determining the physiological indicators with a transfer association strength lower than the preset strength threshold as target indicators;

[0030] Extract the time series data of adjacent indicators in the physiological indicator causal network, calculate the synchronization coefficient of indicator fluctuations in adjacent time windows as the influence weight, and calculate the time difference of the peak value of indicator fluctuations as the propagation delay;

[0031] The corresponding relationship between the source indicator and the target indicator, the impact weight and the propagation delay are written into the indicator association data table to complete the construction of the indicator association data table.

[0032] In an optional embodiment,

[0033] Based on the indicator association data table, the abnormal source indicator is identified. Taking the abnormal source indicator as the starting point, the abnormal diffusion path is determined according to the value of the propagation delay. The abnormal conduction topology diagram is generated, including:

[0034] Extract the association paths of abnormal indicators, as well as the impact weights and propagation delays corresponding to the association paths, from the indicator association data table. Construct a nonlinear attenuation function based on the impact weights. Calculate the cumulative impact weight of each association path. Determine the starting indicator of the association path with the largest cumulative impact weight as the abnormal source indicator.

[0035] Taking the anomaly source indicator as the starting point, constructing an anomaly diffusion path based on the propagation delay, sorting the anomaly diffusion path in layers according to the numerical value of the propagation delay, and generating a time-series progressive anomaly diffusion branch tree based on a recursive iterative algorithm;

[0036] Perform dynamic propagation modeling on each branch in the abnormal diffusion branch tree, calculate the propagation attenuation coefficient and spatial superposition coefficient of the abnormal signal, determine the signal propagation reachability probability based on the propagation attenuation coefficient and spatial superposition coefficient, and select the abnormal diffusion branch with a propagation reachability probability higher than the preset probability threshold as a valid branch;

[0037] An abnormal conduction topology diagram is generated based on effective branches, in which nodes represent indicators, directed connecting lines represent the direction of abnormal diffusion, the influence weight is mapped to the width parameter of the connecting line, the propagation delay is mapped to the length parameter of the connecting line, and the topological hierarchy of the nodes is set according to the timing characteristics of abnormal propagation.

[0038] In an optional embodiment,

[0039] According to the abnormal conduction topology, a multi-level warning trigger threshold is set, and a warning rule table containing warning trigger conditions and warning levels is generated, including:

[0040] Extracting the connection relationship between nodes in the abnormal transmission topology graph, determining the topological hierarchy of the nodes based on the node connection relationship, and dividing the nodes into abnormal source nodes, propagation path nodes, and terminal nodes based on the topological hierarchy;

[0041] Based on the node division results, the warning level system is determined, and the warning types are divided into source warning, process warning and terminal warning. The propagation characteristics of each node are extracted to set the warning response level including first-level warning, second-level warning and third-level warning.

[0042] For abnormal source nodes, adaptive kernel density estimation is used to calculate the probability distribution of abnormal fluctuations to obtain the fluctuation warning threshold. The influence threshold is set based on the node out-degree information. The combination of the fluctuation warning threshold and the influence threshold is set as the source warning trigger condition.

[0043] For propagation path nodes, the propagation delay and cumulative impact weight are extracted based on the propagation path position. The weighted propagation delay is calculated using the path importance to obtain the delay threshold. The combination of the delay threshold and the cumulative impact weight is set as the trigger condition for the process warning.

[0044] For the terminal nodes, a dynamic time window is used to calculate the anomaly score sequence, and a smoothed anomaly sequence is obtained through exponentially weighted moving average processing. The anomaly warning threshold is set based on the changing trend of the smoothed anomaly sequence, and the anomaly warning threshold is set as the trigger condition for the terminal warning;

[0045] The triggering conditions of source warning, process warning and terminal warning and their corresponding warning response levels are combined to generate an early warning rule table including warning triggering conditions and warning levels.

[0046] In an optional embodiment,

[0047] When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing, including:

[0048] Acquire abnormal physiological indicators that exceed multi-level warning trigger thresholds, and determine the abnormal source nodes corresponding to the abnormal physiological indicators from the abnormal conduction topology diagram;

[0049] Extracting all associated paths from the abnormal physiological indicators to the abnormal source node from the abnormal conduction topology map, extracting the propagation delay on each associated path, calculating the influence weight of each associated path, and determining the propagation diffusion coefficient based on the branch structure of the associated path;

[0050] Performing a weighted combination of the propagation delay, the impact weight, and the propagation diffusion coefficient to obtain a cumulative impact factor, multiplying the cumulative impact factor by the attenuation coefficient of abnormal propagation to calculate a correlation degree, and determining a warning response level in a warning rule table based on the correlation degree;

[0051] generating an early warning signal based on the early warning response level, encapsulating an abnormal physiological indicator identifier, a correlation degree value, and an associated path identifier in the early warning signal, and assigning a transmission priority to the early warning signal according to the early warning response level;

[0052] The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing. The first-level warning response signal is transmitted and received for confirmation in real time, and the second-level and third-level warning response signals are transmitted in batches.

[0053] A second aspect of an embodiment of the present invention provides a remote health monitoring and abnormal state warning system for smart fiber clothing, comprising:

[0054] The first unit is used to collect the wearer's physiological signal data and preprocess it to obtain standardized physiological data;

[0055] The second unit is used to construct a multi-dimensional physiological data collaborative analysis matrix based on standardized physiological data;

[0056] The third unit is used to construct a physiological indicator causal network based on a multi-dimensional physiological data collaborative analysis matrix, calculate the influence weights and propagation delays between each physiological indicator in the physiological indicator causal network, construct an indicator association data table, identify abnormal source indicators based on the indicator association data table, use the abnormal source indicator as the starting point, determine the abnormal diffusion path based on the numerical value of the propagation delay, and generate an abnormal conduction topology map;

[0057] The fourth unit is used to set a multi-level warning trigger threshold according to the abnormal conduction topology diagram and generate a warning rule table including warning trigger conditions and warning levels;

[0058] The fifth unit is used to monitor the wearer's standardized physiological data in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

[0059] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0060] processor;

[0061] a memory for storing processor-executable instructions;

[0062] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0063] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0064] In this embodiment, by constructing a multi-dimensional physiological data collaborative analysis matrix and a physiological indicator causal network, it is possible to accurately identify abnormal source indicators and track abnormal diffusion paths, effectively improving the accuracy of early identification of health risks and the timeliness of early warning. By establishing an indicator association data table and an abnormal conduction topology map, systematic tracking and analysis of abnormal physiological indicator states are achieved, providing a more comprehensive assessment basis for health monitoring, reducing the misjudgment rate, and enhancing the reliability of the health monitoring system. The multi-level warning trigger threshold and warning rule table set therein realize graded warnings based on the severity of the abnormality. Combined with the wireless communication function of the smart fiber clothing, a complete remote health monitoring and warning system is constructed, realizing real-time, accurate, and personalized health status monitoring and risk warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 Schematic diagram of the process of remote health monitoring and abnormal state early warning method of smart fiber clothing according to an embodiment of the present invention;

[0066] Figure 2 This is a logical flow chart of an intelligent early warning system based on multi-dimensional physiological data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0068] The following specific embodiments are used to describe the technical solution of the present invention in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0069] Figure 1 FIG. 1 is a flow chart of a method for remote health monitoring and abnormal state warning of smart fiber clothing according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0070] Collect the wearer's physiological signal data and preprocess it to obtain standardized physiological data;

[0071] Construct a multidimensional physiological data collaborative analysis matrix based on standardized physiological data;

[0072] Based on the multi-dimensional physiological data collaborative analysis matrix, a physiological indicator causal network is constructed. The influence weights and propagation delays between each physiological indicator in the physiological indicator causal network are calculated. An indicator association data table is constructed. Based on the indicator association data table, abnormal source indicators are identified. Starting from the abnormal source indicator, the abnormal diffusion path is determined according to the numerical value of the propagation delay, and an abnormal conduction topology map is generated.

[0073] Set multi-level warning trigger thresholds based on the abnormal conduction topology diagram and generate a warning rule table containing warning trigger conditions and warning levels;

[0074] The wearer's standardized physiological data is monitored in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

[0075] In an optional embodiment, collecting physiological signal data of the wearer and preprocessing to obtain standardized physiological data includes:

[0076] Physiological signal data is collected through a multimodal sensor array in the smart fiber clothing. The multimodal sensor array is provided with sensing units at the nodes of the conductive fiber mesh woven structure to collect the original signal waveform output by the sensing units;

[0077] A wavelet basis function is selected based on the frequency distribution of the original signal waveform, the original signal waveform is decomposed using the selected wavelet basis function, an adaptive threshold is calculated based on the local variance of the coefficients of each layer after decomposition, a soft threshold processing is performed on the signal coefficients using the adaptive threshold, and the signal is reconstructed using the selected wavelet basis function to obtain a denoised signal;

[0078] The signal segment length is determined according to the fluctuation characteristics of the denoised signal, the denoised signal is divided into multiple signal segments, the mean and variance of each signal segment are calculated, each signal segment is normalized using the mean and variance, and the normalized adjacent signal segments are smoothly connected to obtain standardized physiological data.

[0079] In this embodiment, the first step is to collect physiological signal data through a multimodal sensor array in smart fiber clothing. The smart fiber clothing adopts a conductive fiber grid weaving structure, and a variety of sensing units are set at the intersection nodes of the grid, including electrocardiogram signal sensing units, electromyography signal sensing units, body temperature sensing units and humidity sensing units. Each sensing unit is made of a composite weaving of silver fiber and polyester fiber to form a detection point with stable conductive performance. The sensing unit transmits the signal to a miniature signal processing module through the conductive yarn woven into the clothing. The module is located at the side waist of the clothing. The sampling frequency can be adjusted according to different signal types. The sampling frequency of electrocardiogram signals is set to 500Hz, the sampling frequency of electromyography signals is set to 1000Hz, and the sampling frequency of body temperature and humidity signals is set to 10Hz. The collected original signal waveform contains the wearer's physiological information, but is also mixed with environmental interference, motion artifacts and electronic equipment noise.

[0080] The collected original signal waveform requires effective denoising. This embodiment employs an adaptive threshold denoising method based on wavelet transform. Based on the frequency characteristics of the original signal waveform, the db6 wavelet basis function is selected for the ECG signal, the sym8 wavelet basis function is selected for the EMG signal, and the coif3 wavelet basis function is selected for the body temperature and humidity signal. Taking the ECG signal as an example, the collected original ECG signal is subjected to a five-layer wavelet decomposition to obtain detailed coefficients and approximate coefficients for different frequency ranges. For each layer of wavelet coefficients, the local variance is calculated using a sliding window method, with the window size set to 1 / 32 of the signal length. For the third layer of detailed coefficients of the ECG signal, the window size is 128 points, and the window sliding step is 32 points. Based on the local variance value, an adaptive threshold is calculated. The threshold calculation takes into account the ratio of signal strength to noise estimate, using a smaller threshold for strong signal areas and a larger threshold for weak signal areas. For example, when the local variance is greater than 0.8, the threshold is set at 0.6 times the noise standard deviation; when the local variance is between 0.3 and 0.8, the threshold is set at 1.2 times the noise standard deviation; and when the local variance is less than 0.3, the threshold is set at 2.0 times the noise standard deviation. The noise standard deviation is estimated by the median absolute deviation of the first-layer detailed coefficients, and its measured value is approximately 0.05 mV. A soft thresholding process is performed on the wavelet coefficients of each layer using the calculated adaptive threshold. Coefficients below the threshold are set to zero, while coefficients above the threshold are proportionally reduced. The processed wavelet coefficients are reconstructed using an inverse wavelet transform to obtain the denoised physiological signal. Experimental measurements show that this method effectively reduces signal noise, improving the signal-to-noise ratio of the ECG signal from the original 8.4 dB to 15.7 dB, while preserving the morphological features of the P wave, QRS complex, and T wave.

[0081] After obtaining the denoised signal, it needs to be standardized to eliminate the influence of individual differences and changes in acquisition conditions. The appropriate signal segment length is determined according to the fluctuation characteristics of the denoised signal. For the ECG signal, the segment length is set to 10 seconds (5000 sampling points); for the electromyography signal, the segment length is set to 5 seconds (5000 sampling points); for the body temperature and humidity signal, the segment length is set to 60 seconds (600 sampling points). Taking the ECG signal as an example, the 10-minute denoised ECG signal is divided into 60 signal segments, each segment contains 5000 sampling points. The mean and variance are calculated for each signal segment. The mean range of the measured ECG signal segment is between -0.2mV and 0.3mV, and the variance range is 0.04mV. 2 to 0.12mV 2 Each signal segment is normalized by subtracting the mean of the segment from each sampling point in the signal segment and then dividing it by the standard deviation (square root of the variance) of the segment. The normalized signal segment has the characteristics of zero mean and unit variance, which effectively eliminates the individual differences in signal amplitude. In order to avoid mutations at the junction of adjacent signal segments, an overlapping connection method is used to achieve a smooth transition. The specific approach is to overlap adjacent segments by 500 sampling points (about 1 second), and use weighted averaging to fuse the signal values ​​in the overlapping area. The weight coefficient changes linearly, and the weight of the previous segment decreases linearly from 1 to 0, and the weight of the next segment increases linearly from 0 to 1. Through this smooth connection method, continuous standardized physiological data is obtained, eliminating the data discontinuity caused by segmented processing. The standardized physiological data after the above processing retains the key features and morphological characteristics of the original physiological signal, while significantly improving the data quality and consistency.

[0082] In this embodiment, by setting a multimodal sensor array at the nodes of the conductive fiber mesh woven structure, multi-point synchronous acquisition of physiological signals is achieved, which improves the spatial resolution and signal integrity of data acquisition. By adopting an adaptive wavelet denoising method based on signal frequency characteristics, the optimal wavelet basis function can be selected according to the frequency distribution characteristics of different types of physiological signals. The adaptive threshold calculated by local variance realizes accurate identification and removal of signal noise, effectively retaining the effective components of the signal. The segment length is adaptively determined by the signal fluctuation characteristics, which avoids the loss of signal features that may be caused by fixed segmentation. The segment normalization processing based on mean and variance eliminates the influence of signal amplitude and baseline drift, and the smooth connection of adjacent segments ensures the continuity of standardized data. The overall solution not only ensures the acquisition quality of physiological signals, but also realizes effective noise reduction and standardization processing of signals, providing a reliable data foundation for subsequent multidimensional data analysis.

[0083] In an optional embodiment, constructing a multidimensional physiological data collaborative analysis matrix based on standardized physiological data includes:

[0084] Acquiring a local extreme value sequence of standardized physiological data, determining a characteristic period based on a time interval distribution of the local extreme value sequence, and constructing an adaptive decomposition basis function according to the characteristic period;

[0085] Adaptive decomposition basis functions are used to decompose the standardized physiological data to obtain multiple frequency components;

[0086] Calculating the phase correlation between the frequency component and the standardized physiological data, and selecting the frequency component whose phase correlation is greater than a preset correlation threshold as the periodic component;

[0087] Extracting the instantaneous phase from the periodic component as a periodic feature, and determining a time warping constraint according to a temporal variation law of the periodic feature;

[0088] Introducing the time warping constraint into a time warping cost function to limit the deformation range of adjacent sampling points, and calculating an optimal deformation path based on the time warping cost function;

[0089] The standardized physiological data are time-aligned according to the optimal deformation path to obtain time-series aligned physiological data, and the time-series aligned physiological data are reorganized into a multi-dimensional physiological data collaborative analysis matrix.

[0090] For example, when constructing a multi-dimensional physiological data collaborative analysis matrix, the local extreme value sequence of the standardized physiological data is first obtained. The standardized physiological data can be an electrocardiogram signal, a respiratory signal, or a blood pressure signal that has been denoised and normalized. By performing local maximum and minimum detection on the standardized signal, a series of local extreme value points can be obtained. For example, for an electrocardiogram signal with a sampling frequency of 1000 Hz, the sliding window method can be used to detect the R wave peak, and the window width is set to 0.2 seconds. When the amplitude of a point in the window is greater than all other points in the window and exceeds the set threshold (such as 3 times the standard deviation of the signal), it is marked as a local maximum value point. For a 30-second electrocardiogram signal, about 35-40 R wave peak points may be detected, forming a local extreme value sequence.

[0091] Based on the obtained local extreme value sequence, the time intervals between adjacent extreme value points are calculated to form a time interval distribution. By performing statistical analysis on the time interval sequence, such as calculating a histogram or a probability density function, the time interval range with the highest frequency of occurrence is determined as the characteristic period. For example, for adult ECG signals, the characteristic period is usually between 0.6 and 1.0 seconds. Taking the characteristic period of 0.8 seconds as an example, an adaptive decomposition basis function is constructed based on this characteristic period. The adaptive decomposition basis function can use a Gaussian waveform or a wavelet function, whose time scale matches the characteristic period. For example, a wavelet basis function with a center frequency of 1 / 0.8 Hz is constructed, and the function width is 1.5 times the characteristic period, that is, 1.2 seconds.

[0092] The standardized physiological data is decomposed using the constructed adaptive decomposition basis functions. Specifically, the basis functions are convolved with the physiological data at different time scales to extract multiple frequency components. For an ECG signal, 8-10 frequency components may be obtained, including respiratory modulation (approximately 0.2-0.3Hz), heart rate fundamental (approximately 1-1.2Hz), and ECG T waves (approximately 3-5Hz). Each component represents the fluctuation characteristics of the original signal within a specific frequency range.

[0093] The phase correlation between the calculated frequency components and the standardized physiological data is calculated. The phase correlation calculation can obtain the instantaneous phase of the signal through Hilbert transform, and then calculate the degree of synchronization between the phase of the frequency component and the phase of the original signal. In the specific implementation, the original signal and the frequency component can be Hilbert transformed to obtain the analytical signal, the phase information can be extracted, and then the cyclic variance of the phase difference between the two can be calculated and converted into a correlation value between 0 and 1. When the phase is highly synchronized, the correlation is close to 1; when the phase is irrelevant, the correlation is close to 0. Set the preset correlation threshold to 0.7, and select the frequency component with a correlation greater than 0.7 as the periodic component. For example, for the decomposition of the ECG signal, 3-4 frequency components are usually retained as periodic components.

[0094] The instantaneous phase is extracted from the retained periodic components as the periodic feature. Hilbert transform is performed on each periodic component to obtain its instantaneous phase change curve. The instantaneous phase is usually manifested as an envelope that increases monotonically with time, and the phase change between adjacent cardiac cycles is approximately 2π. By analyzing the temporal variation of the instantaneous phase, the time regularization constraint can be determined. The time regularization constraint is mainly reflected in the continuity and monotonicity of the phase change, that is, the phase change after deformation should remain smooth and not reverse. The specific constraint parameters can be set as follows: the maximum deformation range of adjacent sampling points does not exceed ±20% of the original sampling interval, and the phase change rate is maintained between 0.8-1.2 times the original change rate.

[0095] A time warping constraint is introduced into the time warping cost function to limit the deformation range of adjacent sampling points. The cost function consists of two parts: a signal difference metric and a time warping constraint. The signal difference metric uses Euclidean distance to calculate the difference between corresponding points in the original signal and the target signal; the time warping constraint penalizes excessive time distortion to prevent unreasonable alignment. In actual calculations, for a signal of length N, an N×N cost matrix is ​​constructed, where each element represents the cost of aligning corresponding points. For example, for two ECG signals of length 1000 points each, the cost matrix size is 1000×1000, and the calculation of a single element includes a comprehensive evaluation of the signal amplitude difference and phase change constraints.

[0096] Based on the constructed time warping cost function, a dynamic programming algorithm is used to calculate the optimal warping path. Starting from the upper left corner of the cost matrix, the cumulative minimum cost path to the lower right corner is calculated step by step. At each step, the lowest cost of three possible directions (right, down, and right-down) is selected. Ultimately, an optimal path from (0, 0) to (N, N) is obtained, representing the best correspondence between the two signals. For a signal with a length of 1000 points, the optimal path typically consists of approximately 1200-1500 point pairs, describing the nonlinear time correspondence.

[0097] Based on the calculated optimal deformation path, the standardized physiological data are time-aligned. The source signal is mapped to the time axis of the target signal according to the deformation path, and interpolation processing may be required to ensure signal continuity. After alignment, the key feature points of different physiological data (such as ECG R wave and blood pressure peak) tend to be synchronized in time, which is conducive to subsequent collaborative analysis. For example, before alignment, the ECG R wave and blood pressure peak may have a time delay of 50-150ms. After alignment, this difference is minimized.

[0098] Multiple physiological data sets aligned in time are reorganized into a multidimensional physiological data collaborative analysis matrix. Each row of the matrix represents a physiological signal, and each column represents multiple physiological indicators at the same time point. For collaborative analysis involving three signals, namely electrocardiogram (ECG), blood pressure, and respiration, the constructed matrix is ​​3×N dimensional, where N is the number of aligned sampling points. This matrix structure clearly visualizes the temporal relationships and interactions between different physiological signals, providing a unified data foundation for subsequent multidimensional data analysis, pattern recognition, and physiological status assessment.

[0099] In this embodiment, the characteristic period of the signal is extracted through local extreme value sequence analysis, and an adaptive decomposition basis function matching the signal characteristics is constructed to achieve accurate decomposition of different types of physiological signals. The periodic component screening method based on phase correlation effectively identifies signal components with significant periodic characteristics and avoids the influence of non-periodic interference components. By extracting periodic characteristics to determine the time regularization constraints, the constraints are introduced into the cost function to limit the deformation range, which not only ensures the accuracy of signal alignment, but also avoids signal distortion caused by excessive deformation. The calculation of the optimal deformation path realizes the time alignment of multidimensional physiological data and eliminates the influence of time delay and sampling asynchrony between different physiological indicators. By reorganizing the time-aligned data into a collaborative analysis matrix, a unified data structure is provided for the subsequent analysis of the correlation between multidimensional physiological indicators, thereby improving the accuracy and reliability of collaborative analysis.

[0100] In an optional embodiment, a physiological indicator causal network is constructed based on a multi-dimensional physiological data collaborative analysis matrix, and the influence weights and propagation delays between the physiological indicators in the physiological indicator causal network are calculated. The indicator association data table is constructed including:

[0101] Divide the physiological indicator data in the multi-dimensional physiological data collaborative analysis matrix into multiple time windows according to the fluctuation period;

[0102] Perform amplitude analysis on the physiological indicator data within each time window, identify the transition time point based on the fluctuation threshold, and divide the physiological indicator data into a stable data interval and a fluctuating data interval;

[0103] Calculate the information entropy value of the physiological indicators in the stable data interval and the information gain value of the physiological indicators in the fluctuating data interval, and determine the information transmission direction between the physiological indicators according to the information entropy value and the information gain value;

[0104] The initial topological structure of the physiological indicator causal network is constructed based on the information transmission direction, and the transmission correlation strength between physiological indicators is obtained by forward recursive verification.

[0105] Classifying the physiological indicators in the physiological indicator causal network according to the transfer association strength, determining the physiological indicators with a transfer association strength higher than a preset strength threshold as source indicators, and determining the physiological indicators with a transfer association strength lower than the preset strength threshold as target indicators;

[0106] Extract the time series data of adjacent indicators in the physiological indicator causal network, calculate the synchronization coefficient of indicator fluctuations in adjacent time windows as the influence weight, and calculate the time difference of the peak value of indicator fluctuations as the propagation delay;

[0107] The corresponding relationship between the source indicator and the target indicator, the impact weight and the propagation delay are written into the indicator association data table to complete the construction of the indicator association data table.

[0108] During implementation, a multidimensional physiological data collaborative analysis matrix is ​​first acquired. This matrix contains time series data for various physiological indicators, including heart rate, blood pressure, blood oxygen saturation, and body temperature. This data is collected through wearable devices and has high temporal resolution and continuity. For example, for a subject, heart rate data was collected for 24 consecutive hours, with a sampling frequency of one point every 10 seconds, generating 8,640 data points; blood pressure data was also recorded at a sampling frequency of one point every 30 seconds, generating 2,880 data points.

[0109] The collected multidimensional physiological data is divided into fluctuation cycles, and the time window size is determined based on the natural variation patterns of physiological indicators. For heart rate data, a 30-minute window can be set, and the full-day data can be divided into 48 windows. For blood pressure data, considering its relatively slow fluctuations, a 60-minute window can be set, and the full-day data can be divided into 24 windows. In actual applications, the window size can be adjusted according to the characteristics of different physiological indicators to ensure that the indicator's complete fluctuation cycle is captured.

[0110] Within each time window, the physiological indicator data is analyzed for amplitude, and a fluctuation threshold is set to identify transition points. For example, for heart rate data, the fluctuation threshold can be set to ±10% of the baseline value. When the heart rate value changes by exceeding this threshold, a transition point is marked. By identifying transition points, the data is divided into stable data intervals and fluctuating data intervals. For example, within a 30-minute window, if the heart rate remains within the range of 72-78 beats / minute for the first 15 minutes, it is marked as a stable interval; then, over the next 15 minutes, the heart rate rises from 75 beats / minute to 95 beats / minute, marking it as a fluctuating interval.

[0111] For each divided data interval, the information entropy of physiological indicators within the stable data interval and the information gain of physiological indicators within the fluctuating data interval were calculated. The information entropy reflects the degree of uncertainty in the data and is obtained by calculating the probability distribution of the data. Taking the heart rate data in the stable interval as an example, the range of 72-78 beats / minute was divided into six smaller intervals. The frequency of data points within each smaller interval was counted, and the information entropy value was calculated to be 1.92. The information gain value for the fluctuating interval was calculated by comparing the change in information entropy before and after the fluctuation. For example, the information gain value of the heart rate in the fluctuating interval was 0.65, indicating an increase in the orderliness of the data.

[0112] The direction of information transfer between physiological indicators is determined based on the calculated information entropy and information gain values. When fluctuations in indicator A occur before fluctuations in indicator B, and the information gain value of A is greater than the information entropy value of B, information is determined to be transferred from A to B. For example, in two consecutive time windows, if heart rate fluctuations occur before blood pressure fluctuations, and the information gain value of the heart rate fluctuation interval is 0.65, which is higher than the information entropy value of 0.58 in the stable blood pressure interval, information is determined to be transferred from heart rate to blood pressure.

[0113] Based on the determined information transmission direction, the initial topological structure of the causal network of physiological indicators is constructed. This structure is represented as a directed graph, with nodes representing physiological indicators and edges representing the information transmission relationship between indicators. The initially constructed network may contain some spurious associations, and forward recursive verification is required to obtain the accurate transmission association strength. Forward recursive verification is an iterative calculation process that gradually adjusts the weights of the edges in the network and removes the edges when the weight falls below a set threshold. For the association between heart rate and blood pressure, after 10 recursive calculations, the transmission association strength was 0.72, which exceeded the preset threshold of 0.5, so the association was retained.

[0114] Physiological indicators are graded based on their transfer correlation strength. Indicators with transfer correlation strengths above a preset strength threshold are identified as source indicators, while those below the threshold are identified as target indicators. The preset strength threshold can be set to 0.6. In this example, the transfer correlation strength of heart rate is 0.72, which is above the threshold and is identified as a source indicator. The transfer correlation strength of blood oxygen saturation is 0.45, which is below the threshold and is identified as a target indicator.

[0115] Time series data of adjacent indicators in the physiological indicator causal network were extracted, and the synchronization coefficient of indicator fluctuations within adjacent time windows was calculated as the influence weight. The synchronization coefficient is obtained by calculating the similarity of the fluctuation patterns of two indicators and ranges from 0 to 1, with larger values ​​indicating greater synchronization. For the association between heart rate and blood pressure, the calculated synchronization coefficient is 0.68, which is used as the influence weight of heart rate on blood pressure. The time difference between the peak fluctuations of the indicators is calculated as the propagation delay. For example, if the peak of the heart rate fluctuation occurs at 10:15 and the peak of the blood pressure fluctuation occurs at 10:18, the propagation delay is 3 minutes.

[0116] The corresponding relationship between the source and target indicators, their impact weights, and propagation delays are recorded in the indicator association data table. This data table contains fields for source indicator, target indicator, impact weight, and propagation delay. For example, for the relationship between heart rate and blood pressure, add a record to the data table: Source indicator: "heart rate," Target indicator: "blood pressure," Impact weight: 0.68, Propagation delay: 3 minutes. Similarly, record all identified indicator associations to complete the construction of the indicator association data table, providing data support for subsequent physiological status assessment and abnormality warnings.

[0117] In this embodiment, the information transmission direction is determined based on the information entropy value and information gain value analysis, which accurately reflects the causal relationship between physiological indicators. The forward recursive verification method is used to obtain the transmission association strength, which effectively avoids the interference of false associations. By grading the transmission association strength to identify the source indicators and target indicators, the hierarchical relationship of physiological indicators in the causal network is clarified. Based on the time series data, the synchronization coefficient is calculated as the influence weight, and the statistical fluctuation peak time difference is used as the propagation delay, which not only quantifies the degree of influence between indicators, but also determines the propagation law of abnormal conditions. The various correlation features are written into the indicator correlation data table, and a complete indicator correlation feature set is constructed, which provides reliable data support for the subsequent identification of abnormal conduction paths and the formulation of early warning rules.

[0118] like Figure 2 As shown, the logical flow of the intelligent early warning system based on multi-dimensional physiological data of this embodiment is demonstrated.

[0119] In an optional embodiment, identifying an abnormal source indicator based on an indicator association data table, taking the abnormal source indicator as a starting point, determining an abnormal diffusion path according to a numerical value of a propagation delay, and generating an abnormal conduction topology map includes:

[0120] Extract the association paths of abnormal indicators, as well as the impact weights and propagation delays corresponding to the association paths, from the indicator association data table. Construct a nonlinear attenuation function based on the impact weights. Calculate the cumulative impact weight of each association path. Determine the starting indicator of the association path with the largest cumulative impact weight as the abnormal source indicator.

[0121] Taking the anomaly source indicator as the starting point, constructing an anomaly diffusion path based on the propagation delay, sorting the anomaly diffusion path in layers according to the numerical value of the propagation delay, and generating a time-series progressive anomaly diffusion branch tree based on a recursive iterative algorithm;

[0122] Perform dynamic propagation modeling on each branch in the abnormal diffusion branch tree, calculate the propagation attenuation coefficient and spatial superposition coefficient of the abnormal signal, determine the signal propagation reachability probability based on the propagation attenuation coefficient and spatial superposition coefficient, and select the abnormal diffusion branch with a propagation reachability probability higher than the preset probability threshold as a valid branch;

[0123] An abnormal conduction topology diagram is generated based on effective branches, in which nodes represent indicators, directed connecting lines represent the direction of abnormal diffusion, the influence weight is mapped to the width parameter of the connecting line, the propagation delay is mapped to the length parameter of the connecting line, and the topological hierarchy of the nodes is set according to the timing characteristics of abnormal propagation.

[0124] In this implementation, the indicator association data table contains relationship data between multiple indicators. Each record includes a source indicator, a target indicator, an impact weight, and a propagation delay. The impact weight indicates the degree of influence of the source indicator on the target indicator, and the propagation delay indicates the time required for an anomaly to propagate from the source indicator to the target indicator.

[0125] During the identification of abnormal source indicators, all associated paths containing abnormal indicators are first extracted from the indicator association data table. For example, the associated path extracted from the indicator association data table is: A→B→C→D, where D is the detected abnormal indicator. The impact weights of the associated paths are 0.8, 0.7, and 0.9, respectively, and the propagation delays are 2 minutes, 3 minutes, and 1 minute, respectively. A nonlinear decay function is constructed based on these impact weights. This function uses an exponential decay formula, calculated by multiplying the impact weights of each segment in the path and adjusting it with a path length factor. For the above path, the cumulative impact weight is calculated as 0.8 × 0.7 × 0.9 × (1 - 0.1 × 3) = 0.3402, where 3 is the path length and 0.1 is the decay adjustment factor. All associated paths that could potentially lead to abnormal indicator D are calculated, and the path with the highest cumulative impact weight is found. Assuming that the cumulative impact weight of another path E→F→D is 0.2856, while the cumulative impact weight of path A→B→C→D is 0.3402, A is determined to be the abnormal source indicator.

[0126] After determining the anomaly source indicator, the anomaly diffusion path is constructed, starting with anomaly source indicator A. Based on the propagation delay, the anomaly diffusion paths are hierarchically sorted by delay size. For example, the propagation delay from A to B is 2 minutes, the propagation delay from A to G is 3 minutes, and the propagation delay from A to H is 1 minute. The system sorts these paths from smallest to largest delay: A to H (1 minute), A to B (2 minutes), and A to G (3 minutes). A recursive iterative algorithm is used to generate a time-sequential anomaly diffusion branch tree. Starting from A, the first-level branches include H, B, and G; continuing from node H, there may be H to I (2 minutes) and H to J (4 minutes); continuing from node B, there is B to C (3 minutes); and continuing from node G, there is G to K (1 minute). These branches are integrated to generate a complete anomaly diffusion branch tree: A is the root node, the first level is H, B, and G, the second level is I, J, C, and K, and the third level is D (spreading from C).

[0127] Dynamic propagation modeling is performed for each branch in the anomaly diffusion tree, and the propagation attenuation coefficient and spatial superposition coefficient of the anomaly signal are calculated. The propagation attenuation coefficient is related to the path length and influence weight. For example, the propagation attenuation coefficient of the path A→B→C is 0.8×0.7×(1-0.05×2)=0.532, where 0.05 is the length adjustment factor and 2 is the number of path segments. The spatial superposition coefficient considers the cumulative effect of multiple paths on the same indicator and is calculated as the weighted sum of the attenuation coefficients of each path. For example, assuming that C can be reached via both A→B→C and A→G→C, and the propagation attenuation coefficients of the two paths are 0.532 and 0.475, respectively, then the spatial superposition coefficient of C is 0.532+0.475=1.007.

[0128] The reachability probability of the abnormal signal is calculated based on the propagation attenuation coefficient and the spatial stacking coefficient. The reachability probability is calculated as the product of the propagation attenuation coefficient and the spatial stacking coefficient, and then normalized. For example, the reachability probability of the path A→B→C→D is 0.532 × 1.007 × 0.9 × (1 - 0.05 × 3) × 100% = 43.15%. With a preset probability threshold of 30%, this path is considered a valid branch.

[0129] Finally, an anomaly transmission topology map is generated based on the valid branches. In the topology map, nodes represent indicators, and directed lines represent the direction of anomaly diffusion. The impact weight is mapped to the width parameter of the connection line. For example, a weight of 0.9 is mapped to a line width of 3 pixels, 0.7 is mapped to a line width of 2 pixels, and 0.5 is mapped to a line width of 1 pixel. The propagation delay is mapped to the length parameter of the connection line. For example, a delay of 1 minute is mapped to a length of 30 pixels, and a delay of 2 minutes is mapped to a length of 60 pixels. The topological hierarchy of the nodes is set according to the temporal characteristics of anomaly propagation. The anomaly source indicator A is located at the top layer (0th layer), the direct impact indicator H with a propagation delay of 1 minute is located at the first layer, the indicator B with a propagation delay of 2 minutes is located at the second layer, and so on. The generated anomaly transmission topology map shows the complete path of the anomaly starting from the source indicator A and propagating to each related indicator in a temporal sequence. This determines the anomaly transmission process and impact range, providing effective guidance for anomaly location and troubleshooting.

[0130] The cumulative impact weight is calculated by a nonlinear attenuation function, and the source indicators of the abnormal state are accurately identified. Based on the propagation delay, the abnormal diffusion path is constructed and hierarchically sorted, and a time-series progressive branch tree is generated by combining a recursive iterative algorithm to achieve a dynamic characterization of the abnormal conduction process. By establishing a dynamic propagation model to calculate the propagation attenuation coefficient and spatial superposition coefficient, and introducing the propagation reachability probability to screen effective branches, the interference of invalid propagation paths is avoided. In the abnormal conduction topology diagram, nodes are used to represent indicators, directed connecting lines are used to represent the diffusion direction, and the impact weight and propagation delay are mapped to the width and length parameters of the connecting line respectively, which intuitively shows the spatial structure and temporal characteristics of abnormal conduction. By setting the topological hierarchy of the nodes, the conduction order of the abnormal state between different physiological indicators is clearly reflected, providing visual support for the formulation of accurate early warning strategies.

[0131] In an optional embodiment, a multi-level warning trigger threshold is set according to the abnormal conduction topology diagram, and a warning rule table including warning trigger conditions and warning levels is generated, including:

[0132] Extracting the connection relationship between nodes in the abnormal transmission topology graph, determining the topological hierarchy of the nodes based on the node connection relationship, and dividing the nodes into abnormal source nodes, propagation path nodes, and terminal nodes based on the topological hierarchy;

[0133] Based on the node division results, the warning level system is determined, and the warning types are divided into source warning, process warning and terminal warning. The propagation characteristics of each node are extracted to set the warning response level including first-level warning, second-level warning and third-level warning.

[0134] For abnormal source nodes, adaptive kernel density estimation is used to calculate the probability distribution of abnormal fluctuations to obtain the fluctuation warning threshold. The influence threshold is set based on the node out-degree information. The combination of the fluctuation warning threshold and the influence threshold is set as the source warning trigger condition.

[0135] For propagation path nodes, the propagation delay and cumulative impact weight are extracted based on the propagation path position. The weighted propagation delay is calculated using the path importance to obtain the delay threshold. The combination of the delay threshold and the cumulative impact weight is set as the trigger condition for the process warning.

[0136] For the terminal nodes, a dynamic time window is used to calculate the anomaly score sequence, and a smoothed anomaly sequence is obtained through exponentially weighted moving average processing. The anomaly warning threshold is set based on the changing trend of the smoothed anomaly sequence, and the anomaly warning threshold is set as the trigger condition for the terminal warning;

[0137] The triggering conditions of source warning, process warning and terminal warning and their corresponding warning response levels are combined to generate an early warning rule table including warning triggering conditions and warning levels.

[0138] In this embodiment, the connection relationship between the nodes in the abnormal conduction topology is first extracted, and the topological hierarchy of the nodes is determined based on the node connection relationship. The physiological data collected by the smart fiber clothing constitutes the abnormal conduction topology map, in which key physiological indicators such as heart rate, blood pressure, and body temperature serve as nodes in the map, and the edges between the nodes represent the conduction relationship of the abnormal state. By analyzing the connection structure of the topology map, the nodes are divided into three categories: abnormal source nodes refer to physiological indicators that may first become abnormal, such as abnormal heart rate, which may cause a series of health problems; propagation path nodes are intermediate indicators in the process of abnormal state transmission; and end nodes are the final influencing indicators of abnormal conduction. For example, when a user is in a high temperature environment, an increase in body temperature (abnormal source) may cause an increase in heart rate (propagation path), which ultimately causes changes in blood pressure (end node).

[0139] Based on the node classification results, an early warning system is established, with early warning types divided into source warnings, process warnings, and terminal warnings. The system then extracts the propagation characteristics of each node to set early warning response levels, including level one, level two, and level three. Source warnings target abnormal source nodes, such as abnormal body temperature; process warnings target propagation path nodes, such as heart rate fluctuations; and terminal warnings target terminal nodes, such as blood pressure fluctuations. Level three warnings are the lowest level, indicating that attention is needed but no immediate risk is present. Level two warnings indicate moderate risk and recommend that users adjust their activities. Level one warnings are the highest level, indicating a potential health risk requiring immediate intervention. The corresponding early warning response level for each indicator is determined by analyzing the propagation speed, impact range, and severity of each indicator's anomalies in historical data.

[0140] For abnormal source nodes, adaptive kernel density estimation is used to calculate the probability distribution of abnormal fluctuations and determine the fluctuation warning threshold. Taking body temperature monitoring as an example, by analyzing the distribution of user temperatures in different environments and activity states, the probability density function of the normal temperature range is calculated. Temperature values ​​falling within the low-probability range are considered abnormal fluctuations. For example, normal body temperatures typically range from 36.3°C to 37.2°C, with the highest probability density range between 36.5°C and 36.9°C. A fluctuation warning is triggered when a user's temperature exceeds 37.5°C for more than 30 minutes. The influence threshold is set based on node outdegree information. The node outdegree represents the number of other indicators that a physiological indicator may affect. An outdegree of 3 for the body temperature node indicates that abnormal body temperature may affect heart rate, blood pressure, and respiratory rate. An influence warning is triggered when two or more of these three indicators are simultaneously affected. The fluctuation warning threshold and the influence threshold are combined to form the source warning trigger condition. For example, if the body temperature exceeds 37.5°C and affects two or more physiological indicators, a secondary source warning is triggered.

[0141] For the nodes on the propagation path, the propagation delay and cumulative impact weight are extracted based on the propagation path position. The propagation delay refers to the time required for the anomaly to propagate from the source node to the current node. According to historical data analysis, it usually takes 5-15 minutes for abnormal body temperature to propagate to the heart rate. The cumulative impact weight represents the cumulative effect on the abnormal propagation path and is calculated by analyzing the amplitude of the parameter changes at each node under the abnormal state. The weighted propagation delay is calculated using the importance of the path to obtain the delay threshold. The importance of the path is determined based on the centrality and connectivity of the path in the entire topology graph. For example, if the heart rate is used as a propagation path node and the body temperature becomes abnormal, if the heart rate rises by more than 20% of the normal value within 10 minutes and the cumulative impact weight reaches 0.65, a process warning is triggered.

[0142] For terminal nodes, a dynamic time window is used to calculate anomaly score sequences. The time window is dynamically adjusted based on the changing characteristics of different physiological indicators. For example, a 30-minute sliding window is used for blood pressure indicators. Within each window, the degree of deviation of the current value from the historical normal range is calculated to form an anomaly score sequence. A smoothed anomaly sequence is obtained through exponentially weighted moving average processing, with more recent data given a higher weight to reduce interference caused by short-term fluctuations. A smoothing factor of 0.3 is used for the blood pressure indicator to effectively filter out noise from physiological fluctuations. Anomaly warning thresholds are set based on the changing trend of the smoothed anomaly sequence. When the smoothed anomaly score exceeds 0.75 for three consecutive time windows, a terminal warning is triggered.

[0143] The triggering conditions for source warnings, process warnings, and terminal warnings, along with their corresponding warning response levels, are combined to generate a warning rule table containing warning trigger conditions and warning levels. The rule table includes three dimensions: trigger conditions, warning types, and response levels. For example, when a body temperature exceeds 37.8°C and affects more than three indicators, a level one source warning is triggered; when the propagation delay of an abnormal heart rate is less than 5 minutes and the cumulative impact weight exceeds 0.8, a level two process warning is triggered; and when the abnormal blood pressure score exceeds 0.85 for four consecutive windows, a level one terminal warning is triggered. The warning rule table is stored in the control module of the smart fiber clothing. The system compares the monitoring data with the warning rules in real time. When the triggering conditions are met, a health warning of the corresponding level is immediately generated, sent to the user via the smart terminal, and corresponding health intervention recommendations are provided based on the warning level.

[0144] In this embodiment, by constructing an abnormal conduction topology map and classifying nodes, the system can identify the source, propagation path, and ultimate impact of health anomalies, improving the accuracy and pertinence of early warnings. Multi-level early warning trigger threshold settings enable more refined early warnings, reducing false alarm rates while not missing important health risks. Adaptive kernel density estimation technology enables automatic adjustment of fluctuation warning thresholds based on individual differences, improving personalized monitoring effectiveness. Dynamic time windowing and exponentially weighted moving average processing effectively filter short-term fluctuations and noise interference in physiological indicators, enhancing system stability. By analyzing abnormal conduction characteristics, the system can predict the development trend of health risks, achieving a transition from passive response to active prevention. The generation of early warning rule tables makes the system highly interpretable, allowing users and medical personnel to understand the cause and severity of early warning triggers. This method is particularly suitable for daily monitoring of patients with cardiovascular and cerebrovascular diseases and chronic diseases, as well as health risk management for the elderly and people in special occupations, significantly enhancing the practical value of remote health monitoring.

[0145] In an optional embodiment, when it is detected that a physiological indicator exceeds a multi-level warning trigger threshold, a warning level is determined according to the degree of correlation between the physiological indicator exceeding the threshold and the abnormal source indicator, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing, including:

[0146] Acquire abnormal physiological indicators that exceed multi-level warning trigger thresholds, and determine the abnormal source nodes corresponding to the abnormal physiological indicators from the abnormal conduction topology diagram;

[0147] Extracting all associated paths from the abnormal physiological indicators to the abnormal source node from the abnormal conduction topology map, extracting the propagation delay on each associated path, calculating the influence weight of each associated path, and determining the propagation diffusion coefficient based on the branch structure of the associated path;

[0148] Performing a weighted combination of the propagation delay, the impact weight, and the propagation diffusion coefficient to obtain a cumulative impact factor, multiplying the cumulative impact factor by the attenuation coefficient of abnormal propagation to calculate a correlation degree, and determining a warning response level in a warning rule table based on the correlation degree;

[0149] generating an early warning signal based on the early warning response level, encapsulating an abnormal physiological indicator identifier, a correlation degree value, and an associated path identifier in the early warning signal, and assigning a transmission priority to the early warning signal according to the early warning response level;

[0150] The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing. The first-level warning response signal is transmitted and received for confirmation in real time, and the second-level and third-level warning response signals are transmitted in batches.

[0151] In this embodiment, a smart fiber garment serves as a monitoring platform. It integrates multiple physiological sensors, including heart rate, blood pressure, temperature, and oxygen saturation sensors. Each sensor collects data on a specific physiological indicator at a specific frequency, which is then processed by a microprocessor within the garment. The microprocessor stores a table of multi-level warning trigger thresholds, which it uses to determine whether the collected physiological indicators exceed normal ranges.

[0152] When a physiological metric is detected exceeding the warning trigger threshold, the system first obtains the abnormal physiological metric exceeding the threshold, such as a heart rate exceeding 110 beats / minute. By querying the abnormal conduction topology map, the abnormal source node corresponding to the abnormal physiological metric is determined. For example, if an abnormal heart rate, elevated blood pressure, or increased respiratory rate is detected, "cardiac function abnormality" may be determined as the abnormal source node.

[0153] Extract all associated paths from abnormal physiological indicators to the abnormal source node from the abnormal conduction topology. For example, there may be a direct path from "abnormal heart rate" to "abnormal cardiac function" or an indirect path through the "autonomic nervous system." For each associated path, extract its propagation delay value. Propagation delay represents the time difference between the abnormal source and the manifestation of the abnormal physiological indicator, recorded in seconds. For example, the propagation delay from abnormal cardiac function to abnormal heart rate may be 5 seconds, and the propagation delay from abnormal cardiac function to abnormal blood pressure may be 30 seconds.

[0154] Calculate the impact weight of each association path. The impact weight reflects the degree of influence of the abnormal source on the physiological indicator, and the value range is 0 to 1. The impact weight is pre-set based on historical data and physiological knowledge. For example, the impact weight of abnormal cardiac function on heart rate is 0.9, and the impact weight on respiratory rate is 0.6. In addition, the propagation diffusion coefficient is determined based on the branching structure of the association path. When the abnormal source affects multiple physiological indicators, the propagation diffusion coefficient is large; when it affects only a small number of physiological indicators, the propagation diffusion coefficient is small. For example, the diffusion coefficient for affecting less than 3 physiological indicators is 1.0, the diffusion coefficient for affecting 4-6 physiological indicators is 1.5, and the diffusion coefficient for affecting more than 7 physiological indicators is 2.0.

[0155] The cumulative impact factor is calculated by weighting the propagation delay, impact weight, and propagation diffusion coefficient. Specifically, the inverse of the propagation delay is multiplied by the impact weight, and then multiplied by the propagation diffusion coefficient to obtain the cumulative impact factor. For example, if the propagation delay of a certain associated path is 10 seconds, the impact weight is 0.8, and the propagation diffusion coefficient is 1.5, then the cumulative impact factor is 0.8×(1 / 10)×1.5=0.12. The degree of association is calculated by multiplying the cumulative impact factor by the attenuation coefficient of the abnormal propagation. The attenuation coefficient represents the degree to which the abnormal signal attenuates over time and the length of the propagation path, and its value ranges from 0 to 1. For example, if the attenuation coefficient is set to 0.9, the degree of association is 0.12×0.9=0.108.

[0156] Based on a pre-defined warning rule table, the warning response level is determined based on the calculated correlation level. The warning rule table defines the correspondence between correlation levels and warning levels. For example, a correlation level greater than 0.5 is a Level 1 warning (the most urgent), a correlation level between 0.2 and 0.5 is a Level 2 warning, a correlation level between 0.05 and 0.2 is a Level 3 warning, and a correlation level less than 0.05 does not trigger an alert. In the above example, the correlation level is 0.108, which corresponds to a Level 3 warning response.

[0157] An early warning signal is generated based on the early warning response level. This signal encapsulates the abnormal physiological indicator identifier (e.g., "abnormal heart rate"), the correlation degree value (e.g., 0.108), and the correlation path identifier (e.g., "abnormal cardiac function -> autonomic nervous system -> heart rate regulation"). Different transmission priorities are assigned to early warning signals based on the early warning response level: Level 1 has the highest transmission priority, while Level 3 has the lowest.

[0158] The smart fiber clothing transmits warning signals to a remote monitoring terminal via a built-in wireless communication module. For level-one warning signals, real-time transmission is used, ensuring immediate signal transmission and requiring confirmation from the recipient. If no confirmation is received, the signal is resent every 30 seconds until confirmation is received. For level-two and level-three warning signals, batch transmission is used, bundling multiple warning signals into a single packet and sending them periodically. Level-two warnings are sent every five minutes, and level-three warnings every 15 minutes, reducing communication overhead and extending device battery life. Upon receiving the warning signal, the remote monitoring terminal takes different response measures based on the warning level. For example, a level-one warning triggers an emergency alert and notifies medical staff; a level-two warning records the anomaly and prompts the monitoring staff's attention; and a level-three warning simply records the data for subsequent analysis. The entire system enables intelligent monitoring of the wearer's physiological state and multi-level warnings, improving the accuracy and effectiveness of remote health monitoring.

[0159] By constructing an abnormal conduction topology map and classifying nodes, the source, propagation path, and ultimate impact of health anomalies can be identified, improving the accuracy and relevance of early warnings. Multi-level warning trigger thresholds enable more refined alerts, reducing false alarms while not missing important health risks. Adaptive kernel density estimation technology automatically adjusts fluctuation warning thresholds based on individual differences, enhancing personalized monitoring. Dynamic time windowing and exponentially weighted moving average processing effectively filter short-term fluctuations and noise in physiological indicators, enhancing system stability. By analyzing abnormal conduction characteristics, the development trend of health risks can be predicted, enabling a shift from passive response to proactive prevention. The generation of warning rule tables makes the system highly interpretable, enabling users and medical staff to understand the cause and severity of warning triggers. This method is particularly suitable for routine monitoring of patients with cardiovascular and cerebrovascular diseases and chronic diseases, as well as for health risk management for the elderly and those in special occupations, significantly enhancing the practical value of remote health monitoring.

[0160] In this embodiment,

[0161] A second aspect of an embodiment of the present invention provides a remote health monitoring and abnormal state warning system for smart fiber clothing, the system comprising:

[0162] The first unit is used to collect the wearer's physiological signal data and pre-process it to obtain standardized physiological data;

[0163] The second unit is used to construct a multi-dimensional physiological data collaborative analysis matrix based on standardized physiological data;

[0164] The third unit is used to construct a physiological indicator causal network based on a multi-dimensional physiological data collaborative analysis matrix, calculate the influence weights and propagation delays between each physiological indicator in the physiological indicator causal network, construct an indicator association data table, identify abnormal source indicators based on the indicator association data table, use the abnormal source indicator as the starting point, determine the abnormal diffusion path based on the numerical value of the propagation delay, and generate an abnormal conduction topology map;

[0165] The fourth unit is used to set a multi-level warning trigger threshold according to the abnormal conduction topology diagram and generate a warning rule table including warning trigger conditions and warning levels;

[0166] The fifth unit is used to monitor the wearer's standardized physiological data in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

[0167] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0168] processor;

[0169] a memory for storing processor-executable instructions;

[0170] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0171] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0172] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote health monitoring and abnormal state warning method for smart fiber clothing, characterized in that: include: Collect the wearer's physiological signal data and preprocess it to obtain standardized physiological data; Construct a multidimensional physiological data collaborative analysis matrix based on standardized physiological data; Based on the multi-dimensional physiological data collaborative analysis matrix, a physiological indicator causal network is constructed. The influence weights and propagation delays between each physiological indicator in the physiological indicator causal network are calculated. An indicator association data table is constructed. Based on the indicator association data table, abnormal source indicators are identified. Starting from the abnormal source indicator, the abnormal diffusion path is determined according to the numerical value of the propagation delay, and an abnormal conduction topology map is generated. Set multi-level warning trigger thresholds based on the abnormal conduction topology diagram and generate a warning rule table containing warning trigger conditions and warning levels; The wearer's standardized physiological data is monitored in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

2. The method according to claim 1, characterized in that Collecting the wearer's physiological signal data and preprocessing it to obtain standardized physiological data includes: Physiological signal data is collected through a multimodal sensor array in the smart fiber clothing. The multimodal sensor array is provided with sensing units at the nodes of the conductive fiber mesh woven structure to collect the original signal waveform output by the sensing units; A wavelet basis function is selected based on the frequency distribution of the original signal waveform, the original signal waveform is decomposed using the selected wavelet basis function, an adaptive threshold is calculated based on the local variance of the coefficients of each layer after decomposition, a soft threshold processing is performed on the signal coefficients using the adaptive threshold, and the signal is reconstructed using the selected wavelet basis function to obtain a denoised signal; The signal segment length is determined according to the fluctuation characteristics of the denoised signal, the denoised signal is divided into multiple signal segments, the mean and variance of each signal segment are calculated, each signal segment is normalized using the mean and variance, and the normalized adjacent signal segments are smoothly connected to obtain standardized physiological data.

3. The method according to claim 1, characterized in that Constructing a multidimensional physiological data collaborative analysis matrix based on standardized physiological data includes: Acquiring a local extreme value sequence of standardized physiological data, determining a characteristic period based on a time interval distribution of the local extreme value sequence, and constructing an adaptive decomposition basis function according to the characteristic period; Adaptive decomposition basis functions are used to decompose the standardized physiological data to obtain multiple frequency components; Calculating the phase correlation between the frequency component and the standardized physiological data, and selecting the frequency component whose phase correlation is greater than a preset correlation threshold as the periodic component; Extracting the instantaneous phase from the periodic component as a periodic feature, and determining a time warping constraint according to a temporal variation law of the periodic feature; Introducing the time warping constraint into a time warping cost function to limit the deformation range of adjacent sampling points, and calculating an optimal deformation path based on the time warping cost function; The standardized physiological data are time-aligned according to the optimal deformation path to obtain time-series aligned physiological data, and the time-series aligned physiological data are reorganized into a multi-dimensional physiological data collaborative analysis matrix.

4. The method according to claim 1, wherein Based on the multi-dimensional physiological data collaborative analysis matrix, a physiological indicator causal network is constructed, and the influence weights and propagation delays between the physiological indicators in the physiological indicator causal network are calculated. The indicator association data table is constructed, including: Divide the physiological indicator data in the multi-dimensional physiological data collaborative analysis matrix into multiple time windows according to the fluctuation period; Perform amplitude analysis on the physiological indicator data within each time window, identify the transition time point based on the fluctuation threshold, and divide the physiological indicator data into a stable data interval and a fluctuating data interval; Calculate the information entropy value of the physiological indicators in the stable data interval and the information gain value of the physiological indicators in the fluctuating data interval, and determine the information transmission direction between the physiological indicators according to the information entropy value and the information gain value; The initial topological structure of the causal network of physiological indicators is constructed based on the direction of information transmission, and the transmission correlation strength between physiological indicators is obtained by forward recursive verification. Classifying the physiological indicators in the physiological indicator causal network according to the transfer association strength, determining the physiological indicators with a transfer association strength higher than a preset strength threshold as source indicators, and determining the physiological indicators with a transfer association strength lower than the preset strength threshold as target indicators; Extract the time series data of adjacent indicators in the physiological indicator causal network, calculate the synchronization coefficient of indicator fluctuations in adjacent time windows as the influence weight, and calculate the time difference of the peak value of indicator fluctuations as the propagation delay; The corresponding relationship between the source indicator and the target indicator, the impact weight and the propagation delay are written into the indicator association data table to complete the construction of the indicator association data table.

5. The method according to claim 1, wherein Based on the indicator association data table, the abnormal source indicator is identified. Taking the abnormal source indicator as the starting point, the abnormal diffusion path is determined according to the value of the propagation delay. The abnormal conduction topology diagram is generated, including: Extract the association paths of abnormal indicators, as well as the impact weights and propagation delays corresponding to the association paths, from the indicator association data table. Construct a nonlinear attenuation function based on the impact weights. Calculate the cumulative impact weight of each association path. Determine the starting indicator of the association path with the largest cumulative impact weight as the abnormal source indicator. Taking the anomaly source indicator as the starting point, constructing an anomaly diffusion path based on the propagation delay, sorting the anomaly diffusion path in layers according to the numerical value of the propagation delay, and generating a time-series progressive anomaly diffusion branch tree based on a recursive iterative algorithm; Perform dynamic propagation modeling on each branch in the abnormal diffusion branch tree, calculate the propagation attenuation coefficient and spatial superposition coefficient of the abnormal signal, determine the signal propagation reachability probability based on the propagation attenuation coefficient and spatial superposition coefficient, and select the abnormal diffusion branch with a propagation reachability probability higher than the preset probability threshold as a valid branch; An abnormal conduction topology diagram is generated based on effective branches, in which nodes represent indicators, directed connecting lines represent the direction of abnormal diffusion, the influence weight is mapped to the width parameter of the connecting line, the propagation delay is mapped to the length parameter of the connecting line, and the topological hierarchy of the nodes is set according to the timing characteristics of abnormal propagation.

6. The method according to claim 1, characterized in that According to the abnormal conduction topology, a multi-level warning trigger threshold is set, and a warning rule table containing warning trigger conditions and warning levels is generated, including: Extracting the connection relationship between nodes in the abnormal transmission topology graph, determining the topological hierarchy of the nodes based on the node connection relationship, and dividing the nodes into abnormal source nodes, propagation path nodes, and terminal nodes based on the topological hierarchy; Based on the node division results, the warning level system is determined, and the warning types are divided into source warning, process warning and terminal warning. The propagation characteristics of each node are extracted to set the warning response level including first-level warning, second-level warning and third-level warning. For abnormal source nodes, adaptive kernel density estimation is used to calculate the probability distribution of abnormal fluctuations to obtain the fluctuation warning threshold. The influence threshold is set based on the node out-degree information. The combination of the fluctuation warning threshold and the influence threshold is set as the source warning trigger condition. For propagation path nodes, the propagation delay and cumulative impact weight are extracted based on the propagation path position. The weighted propagation delay is calculated using the path importance to obtain the delay threshold. The combination of the delay threshold and the cumulative impact weight is set as the trigger condition for the process warning. For the terminal nodes, a dynamic time window is used to calculate the anomaly score sequence, and a smoothed anomaly sequence is obtained through exponentially weighted moving average processing. The anomaly warning threshold is set based on the changing trend of the smoothed anomaly sequence, and the anomaly warning threshold is set as the trigger condition for the terminal warning; The triggering conditions of source warning, process warning and terminal warning and their corresponding warning response levels are combined to generate an early warning rule table including warning triggering conditions and warning levels.

7. The method according to claim 1, characterized in that When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing, including: Acquire abnormal physiological indicators that exceed multi-level warning trigger thresholds, and determine the abnormal source nodes corresponding to the abnormal physiological indicators from the abnormal conduction topology diagram; Extracting all associated paths from the abnormal physiological indicators to the abnormal source node from the abnormal conduction topology map, extracting the propagation delay on each associated path, calculating the influence weight of each associated path, and determining the propagation diffusion coefficient based on the branch structure of the associated path; Performing a weighted combination of the propagation delay, the impact weight, and the propagation diffusion coefficient to obtain a cumulative impact factor, multiplying the cumulative impact factor by the attenuation coefficient of abnormal propagation to calculate a correlation degree, and determining a warning response level in a warning rule table based on the correlation degree; generating an early warning signal based on the early warning response level, encapsulating an abnormal physiological indicator identifier, a correlation degree value, and an associated path identifier in the early warning signal, and assigning a transmission priority to the early warning signal according to the early warning response level; The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing. The first-level warning response signal is transmitted and received for confirmation in real time, and the second-level and third-level warning response signals are transmitted in batches.

8. A remote health monitoring and abnormal state warning system for smart fiber clothing, used to implement the method according to any one of claims 1 to 7, characterized in that: include: The first unit is used to collect the wearer's physiological signal data and pre-process it to obtain standardized physiological data; The second unit is used to construct a multi-dimensional physiological data collaborative analysis matrix based on standardized physiological data; The third unit is used to construct a physiological indicator causal network based on a multi-dimensional physiological data collaborative analysis matrix, calculate the influence weights and propagation delays between each physiological indicator in the physiological indicator causal network, construct an indicator association data table, identify abnormal source indicators based on the indicator association data table, use the abnormal source indicator as the starting point, determine the abnormal diffusion path based on the numerical value of the propagation delay, and generate an abnormal conduction topology map; The fourth unit is used to set a multi-level warning trigger threshold according to the abnormal conduction topology diagram and generate a warning rule table including warning trigger conditions and warning levels; The fifth unit is used to monitor the wearer's standardized physiological data in real time. When it is detected that the physiological indicators exceed the multi-level warning trigger threshold, the warning level is determined according to the degree of correlation between the physiological indicators exceeding the threshold and the abnormal source indicators, and a warning signal is generated according to the warning rule table. The warning signal is sent to the remote monitoring terminal through the wireless communication module of the smart fiber clothing.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images