A method and system for monitoring and early warning of vital signs based on multi-source data fusion

Abstract

This invention provides a method and system for monitoring and warning vital signs based on multi-source data fusion, belonging to the field of vital sign monitoring and warning technology. The method includes simultaneously acquiring multi-dimensional physiological signals of the target subject, such as electrocardiogram, photoplethysmography pulse wave, heart sounds, and chest wall impedance respiration, as well as environmental data such as temperature and humidity; combining preset individualized patient configurations and real-time environmental data, performing parallel preprocessing and context grouping; extracting multi-dimensional characteristic parameters reflecting cardiac electrical activity from the preprocessed and grouped data, and analyzing long-term trends; constructing a three-level physiological control logic fusion model and outputting qualitative labels; and executing multi-level warning decisions. The system includes modules such as data acquisition and synchronization. By simulating the inherent regulatory logic of the physiological system and integrating multi-dimensional information from individuals and the environment, this invention can achieve early, accurate, personalized, and interpretable risk warnings, significantly reducing false alarms and missed alarms.

Description

Technical Field

[0001] This invention relates to the field of vital sign monitoring and early warning technology, and in particular to a method and system for vital sign monitoring and early warning based on multi-source data fusion. Background Technology

[0002] In the field of vital sign monitoring, the evolution from traditional single-parameter bedside monitoring to the current ubiquitous sensing based on multi-source information fusion has become the mainstream direction for improving clinical reliability and achieving early warning. Wearable devices, environmental sensors, and medical instruments constitute a dense data network, continuously collecting and aggregating multi-dimensional information such as heart rate, blood oxygen, respiration, and body temperature. Mainstream methods typically collect data from multiple sensors, perform preprocessing such as filtering and alignment, and then use statistical or deep learning models to fuse features or decisions, ultimately outputting assessments and alerts. However, this approach has significant bottlenecks: First, data quality and consistency issues are prominent. Due to differences in principles, manufacturing processes, and wearing methods, different sensors contain baseline drift, motion artifacts, and random noise in their signals. Simple data alignment and weighted averaging are insufficient to bridge the gaps in physical meaning and may even introduce new errors. Secondly, rigid early warning logic leads to a high false alarm rate. Rule engines relying on fixed thresholds or simple trend judgments cannot adapt to individual differences and physiological dynamic fluctuations, resulting in alarm fatigue. While black-box models trained on large amounts of data have some adaptive capabilities, they lack interpretability and are unreliable in complex scenarios involving slow deterioration of vital signs or multiple intertwined complications. Thirdly, and more fundamentally, existing methods mostly stop at statistical correlations at the data level, seriously ignoring the physiological and pathophysiological principles underlying changes in vital signs. For example, an increase in heart rate may stem from exercise, emotions, or insufficient blood volume, and its relationship with other parameters follows different regulatory logics. Data correlation alone cannot distinguish between physiological compensation and pathological deterioration, resulting in poor predictability of early warnings. Existing systems generally lack the ability to model and understand the internal logical chain of dynamic evolution of physiological systems. Systems are good at processing discrete data points but are unable to interpret the continuous process described by these data—the sliding of a life system from a steady state to imbalance. In real-world scenarios, critical events often involve a dynamic process where multiple systems and parameters evolve step-by-step according to specific physiological and pathological logic. For example, the progression from infection to septic shock involves multiple stages, including temperature regulation, inflammatory release, compensatory increase in cardiac output, and ultimately, compensatory failure. Each step leaves a temporal imprint on vital signs. Existing systems, even with access to a large amount of data, still have fragmented analysis units, and their warning logic remains based on the assumption that an alarm is triggered when both A and B are abnormal. This fails to address crucial questions such as where the current abnormal combination falls within the logical chain, the causal relationship between different parameters, and the next step the system will take. Therefore, there is an urgent need for a novel method and system for monitoring and warning vital signs based on multi-source data fusion. Summary of the Invention

[0003] The purpose of this invention is to provide a vital sign monitoring and early warning method and system based on multi-source data fusion, in order to solve the technical bottlenecks in the prior art, such as low fusion quality due to the fragmentation of the physical meaning of multi-source data, high false alarm rate due to rigid early warning logic and black box model, inability to distinguish between compensated and decompensated states due to lack of embedded modeling of physiological and pathophysiological response principles, and early warning lag and attribution difficulties due to the inability to model the internal logical chain of the dynamic evolution of physiological systems, as well as insufficient system robustness due to data missingness and noise interference. The specific technical solution is as follows: This invention provides a method for monitoring and early warning of vital signs based on multi-source data fusion, comprising: S10 is equipped with a multi-source physiological data acquisition module, which simultaneously acquires the target object's surface electrocardiogram signal, photoplethysmography pulse wave signal, apical heart sound signal and chest wall impedance respiratory signal, as well as the temperature, humidity, atmospheric pressure and ultraviolet intensity data of the surrounding environment. S20 performs parallel preprocessing and group analysis on the collected multi-source raw signals according to the preset individual and environmental context configuration; S30 extracts multi-dimensional feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from the preprocessed signal, and analyzes the evolution trend of feature parameters under different environmental and individual factors based on historical data and individual risk profiles. S40 constructs a three-level physiological control logic fusion model, which deeply integrates multi-dimensional feature parameters and their trend information and performs state inference; the model includes a signal-cell function layer, an organ-regulatory circuit layer, and a system-compensatory state layer; S50, based on the quantitative indices, qualitative labels, and trend analysis results output by the system-compensation state layer, executes multi-level early warning decisions; S60 outputs the early warning decision results, including the early warning level, the currently determined dominant pathophysiological mechanism, the current logical chain stage, the associated environmental and individual risk factors, a list of key abnormal characteristic parameters, and the degree of deviation from the normal reference range.

[0004] Furthermore, parallel preprocessing of the acquired multi-source raw signals includes bandpass filtering of the electrocardiogram signal from 0.05 Hz to 150 Hz, bandpass filtering of the photoplethysmography (PPG) signal from 0.1 Hz to 20 Hz and simultaneous motion artifact elimination based on adaptive filtering using triaxial accelerometer data, bandpass filtering of the heart sound signal from 25 Hz to 400 Hz, and bandpass filtering of the chest wall impedance signal from 0.05 Hz to 2 Hz.

[0005] Furthermore, S10 also includes: S101 is equipped with a differential amplifier for acquiring surface electrocardiogram signals, and the lead configuration is a bipolar chest lead. S102 is equipped with dual-wavelength light-emitting diodes with emission wavelengths of 660 nm and 905 nm and a photodetector for collecting photoplethysmography pulse waves. The sensor is driven by a constant current source and a temperature compensation circuit to ensure stable light source output. It is equipped with an integrated environmental sensor module to synchronously collect the temperature, relative humidity, atmospheric pressure and ultraviolet index of the microenvironment in which the target object is located, with a sampling rate of not less than 1Hz, and synchronized with the physiological signal acquisition system.

[0006] Furthermore, S20 also includes: S201, the preprocessing of the electrocardiogram signal also includes the use of a threshold denoising method based on wavelet transform, specifically using the sym4 wavelet basis for 5-level decomposition, and using a soft threshold function to process the detail coefficients of the first to third levels. S202 eliminates motion artifacts in photoplethysmography pulse waves by using a normalized minimum mean square adaptive filter with the vector sum of the triaxial accelerometer signals as the reference noise input. The filter order is set to 32 and the step size parameter is set to 0.01.

[0007] The individual and environmental context configuration includes a patient individual profile, which includes age, gender, history of underlying diseases, history of long-term medication, and dietary habit tags; an environmental profile, which is used to identify normal and abnormal environments (such as high altitude, high temperature and humidity, low temperature, etc.); and a risk profile, which is used to define the baseline of characteristic parameters and risk warning threshold adjustment strategies for a specific population in a specific environment.

[0008] The grouping analysis refers to the preprocessing workflow automatically selecting or adjusting filtering parameters, feature extraction priorities, and baseline reference ranges based on contextual configuration. For example, different blood oxygen saturation trend analysis models and warning thresholds are used for "long-term residents of high-altitude areas" and "short-term residents from plains entering high-altitude areas."

[0009] Furthermore, the multi-dimensional feature parameters in S30 include: Extract the QT interval, T wave peak-end interval, T wave amplitude and morphological asymmetry coefficient, and beat-by-beat RR interval sequence from the electrocardiogram signal and calculate the low-frequency power to high-frequency power ratio of heart rate variability. Calculate pulse wave conduction time from the synchronous ECG R wave peak and pulse wave trough; Locate the peak point of the first heart sound S1 from the heart sound signal and calculate the electromechanical delay time between it and the peak value of the ECG R wave; The ratio of the diastolic notch height to the peak value of the main wave is extracted from the photoplethysmography (PPG) signal and defined as the reflection enhancement index. Establish short-term correlation models between characteristic parameters and environmental parameters (such as temperature and air pressure), as well as statistical association models between characteristic parameters and individual long-term risk factors (such as high-carbohydrate diet markers); For specific risk populations (such as patients labeled as having a "long-term high-carbohydrate diet"), the long-term deviation trend of their cardiovascular function-related characteristic parameters (such as enhanced reflex index and heart rate variability) compared to the baseline of healthy individuals of the same age was calculated, and the acute risk superposition effect was assessed in conjunction with real-time environmental stress (such as high altitude and hypoxia).

[0010] Furthermore, the S30 also includes: S301, The calculation method for the T-wave morphological asymmetry coefficient is as follows: within the time limit from the start to the end of the T-wave, taking the peak value of the T-wave as the boundary, calculate the absolute value of the difference between the area of ​​the waveform before the peak value and the area of ​​the waveform after the peak value, and then divide it by the total area of ​​the entire T-wave waveform. S302, The calculation of the electromechanical delay time is as follows: In the synchronized electrocardiogram signal and heart sound signal, the peak point time t_R of the electrocardiogram R wave and the peak point time t_S1 of the principal component of the first heart sound S1 are accurately detected. The delay time EMD = t_S1- t_R is calculated and the average value is taken over 5 consecutive stable heartbeat cycles. S303 calculates the frequency domain index of heart rate variability from beat-by-beat RR interval sequences. Specifically, it uses the Lomb-Scargle periodogram estimation method to adapt to non-uniformly sampled RR interval sequences. The power in the 0.04 Hz to 0.15 Hz band is calculated as the low-frequency power, and the power in the 0.15 Hz to 0.4 Hz band is calculated as the high-frequency power. The ratio LF / HF is then obtained.

[0011] Furthermore, the S40 also includes: S401, the signal-cell function layer, is used to assess the functional status of cardiomyocyte ion channels and excitation-contraction coupling. Inputs include QT interval, T-wave end-peak interval, T-wave morphology asymmetry coefficient, and electromechanical delay time. It incorporates a computational network based on cellular electrophysiology. By performing Fridricia correction on the QT interval and heart rate, and substituting the corrected QTc and T-wave end-peak interval into an empirical formula, it estimates the relative intensity of transmembrane potassium ion current during phase 3 repolarization of the ventricular myocyte action potential. Regression correction is then performed on the electromechanical delay time and heart rate to obtain a heart rate-independent electromechanical coupling index. S402, the organ-regulation loop layer, is used to assess the stability of cardiac autonomic regulation and hemodynamic feedback. The inputs are the heart rate variability (LF / HF ratio), pulse wave conduction time, reflex enhancement index, and electromechanical coupling index. It simulates cardiovascular baroreflex and mechano-electro-feedback loops, maps the LF / HF ratio to sympathetic-vagal balance tension, establishes a nonlinear mapping relationship between the reflex enhancement index and peripheral vascular tension, and calculates the quantification value of ventricular-vascular coupling efficiency by integrating the electromechanical coupling index and pulse wave conduction time. S403, the system-compensation state layer, is used to comprehensively assess the overall compensatory capacity and decompensation risk of the circulatory system. The inputs are the ion current intensity index and electromechanical coupling index output from layer S401, and the autonomic nervous balance tension and ventricular-vascular coupling efficiency values ​​output from layer S402. The layer adopts a fuzzy inference system and outputs the control logic stability index and the qualitative label of the current dominant imbalance mode. The labels include electroremodeling-dominant, mechanical remodeling-dominant, autonomic nervous disorder, and mixed types.

[0012] Furthermore, the S50 also includes: S501, the early warning decision is divided into attention level early warning, advanced early warning and emergency early warning; the decision input is the control logic stability index and the current dominant imbalance mode label, as well as the real-time trend of T-wave morphology asymmetry coefficient and electro-mechanical delay time; S502, Establish an early warning rule base, which contains a set of logical judgments; S503 monitors parameters in real time and triggers alerts.

[0013] Furthermore, the S60 also includes: S601, the warning output includes the warning level code, trigger timestamp, current control logic stability index value and current dominant imbalance mode label; S602, and outputs a list of key evidence chain parameters that led to the warning. The list includes the three parameters that deviated most significantly from the baseline and their percentage deviation, as well as a description of the logical chain stage in which the judgment is made. The S603 output methods include displaying on a local monitoring terminal, generating voice prompts, and sending encrypted data packets to a remote monitoring center server via a wireless network.

[0014] This invention also provides a vital sign monitoring and early warning system based on multi-source data fusion, used to implement the method described above. The system includes: The data acquisition and synchronization module is used to control the integrated multi-physiological sensors and environmental sensors to synchronously acquire surface electrocardiogram signals, photoplethysmography pulse wave signals, apical heart sound signals, chest wall impedance respiratory signals, and ambient temperature, humidity, air pressure and ultraviolet intensity data of the target object according to the preset monitoring mode. The signal preprocessing and grouping analysis module is used to store and retrieve the patient's individualized profile, environmental profile, and risk profile. It is used to perform parallel preprocessing and context grouping of multi-source raw data according to individual and environmental context configurations, generating clean physiological waveform data after denoising and artifact elimination, as well as aligned environmental data. The multi-dimensional feature extraction and trend analysis engine is used to calculate a set of feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from preprocessed and grouped data, and analyze the dynamic evolution trend of this set under different individual and environmental dimensions. The hierarchical physiological control logic fusion engine is used to receive the set of feature parameters and their trend information, perform state deduction through a three-level fusion model, and output quantitative indices and qualitative labels that characterize the stability of the system. The three-level fusion model includes the signal-cell function layer, the organ-regulatory circuit layer, and the system-compensation state layer. A multi-level early warning decision engine is used to perform graded early warning judgments based on the control logic stability index and the current dominant imbalance mode label, combined with the real-time trend and trend analysis results of characteristic parameters. The results output and communication module is used to generate and transmit structured early warning decision results that contain information on the correlation between environmental and individual risk factors.

[0015] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described herein.

[0016] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method.

[0017] The beneficial effects of this invention are as follows: By simultaneously acquiring electrocardiogram, photoplethysmography (PPG), heart sounds, and chest wall impedance respiratory signals, and constructing a three-tiered physiological control logic fusion model—signal-cell function layer, organ-regulatory circuit layer, and system-compensatory state layer—this invention achieves the interpretation of the complete logical chain of the evolution of a living organism from homeostasis to decompensation from multi-source physiological data. Its core beneficial effect lies in transforming the traditional lagging, high-false-alarm warning based on a single parameter threshold into an early, accurate, and interpretable risk prediction based on the deduction of the physiological regulatory logic chain. It can capture the intrinsic linkages between deep parameters such as myocardial cell ion channel function, electromechanical coupling efficiency, autonomic nerve tension, and vascular coupling state, thereby identifying the specific stage of the system in strong compensation, weak compensation, or decompensation before significant deterioration of vital signs, and determining the dominant imbalance mode (such as electrical remodeling, mechanical disorder, etc.). This not only greatly reduces false alarms and missed alarms caused by individual differences and physiological fluctuations, but also provides clear pathophysiological basis and key time windows for clinical intervention, realizing a qualitative change in intelligent monitoring from phenomenon alarm to mechanism early warning.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram illustrating the steps of a vital sign monitoring and early warning method based on multi-source data fusion according to the present invention; Figure 2 This is a schematic diagram of the structure of a vital signs monitoring and early warning system based on multi-source data fusion according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0022] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0023] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0024] In an embodiment of the present invention, a method for monitoring and early warning of vital signs based on multi-source data fusion is provided. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps: S10 is equipped with a multi-source physiological data acquisition module, which simultaneously acquires the target object's surface electrocardiogram signal, photoplethysmography pulse wave signal, apical heart sound signal and chest wall impedance respiratory signal, as well as the temperature, humidity, atmospheric pressure and ultraviolet intensity data of the surrounding environment. S20, based on preset individual and environmental context configurations, performs parallel preprocessing and group analysis on the acquired multi-source raw signals, including bandpass filtering of ECG signals from 0.05 Hz to 150 Hz to retain complete QRS wave and ST-T segment information, bandpass filtering of photoplethysmography signals from 0.1 Hz to 20 Hz and simultaneously using triaxial accelerometer data for motion artifact elimination based on adaptive filtering, bandpass filtering of heart sound signals from 25 Hz to 400 Hz to highlight the components of the first heart sound S1 and the second heart sound S2, and bandpass filtering of chest wall impedance signals from 0.05 Hz to 2 Hz to extract respiratory rhythm; S30 extracts multi-dimensional feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from the preprocessed signal, and analyzes the evolution trend of feature parameters under different environmental and individual factors based on historical data and individual risk profiles. S40, construct a three-level physiological control logic fusion model, and deeply fuse and extrapolate the multi-dimensional feature parameters and their trend information extracted in step S30; the model includes a signal-cell function layer, an organ-regulatory circuit layer and a system-compensatory state layer; S50, based on the quantitative index, qualitative label and trend analysis results output by the system-compensation state layer in step S40, execute multi-level early warning decisions; the decision logic is based on preset thresholds and rule bases or those corresponding to different pathophysiological evolution chains. S60 outputs the early warning decision results, including the early warning level, the currently determined dominant pathophysiological mechanism, the current logical chain stage, the associated environmental and individual risk factors, a list of key abnormal characteristic parameters and their degree of deviation from the normal reference range.

[0025] The principle and effects of the above technical solution are as follows: By simultaneously acquiring multi-source signals such as electrocardiogram, pulse wave, heart sound, and impedance respiration, and extracting deep-level characteristics from electrical activity to mechanical contraction and from central conduction to peripheral reflexes, it breaks through the limitations of traditional methods that only focus on heart rate, blood pressure, and blood oxygen. The core lies in constructing a three-layer fusion model that simulates the internal regulatory logic of a physiological system. This model links the microscopic cellular functional state, the mesoscopic organ regulatory circuit, and the macroscopic system's compensatory capacity, achieving a qualitative leap from "data correlation" to "logical deduction." Its effect is that it can identify the core precursor mechanisms of sudden cardiac death, such as cardiac electromechanical coupling disorders and autonomic nervous system regulatory imbalances, earlier, and quantitatively assess the system's compensatory reserves and decompensation risks, thereby achieving accurate and forward-looking early warning and greatly reducing false alarms and missed alarms.

[0026] In one specific embodiment, the sampling rate of the S10 center electrical signal is 1000 Hz, the sampling rate of the photoplethysmography pulse wave signal is 500 Hz, the sampling rate of the heart sound signal is 4000 Hz, and the sampling rate of the chest wall impedance signal is 100 Hz.

[0027] In one embodiment, S10 further includes: The S101 is equipped with a differential amplifier with high input impedance and a common-mode rejection ratio greater than 120 dB for acquiring surface electrocardiogram signals. The lead configuration is a modified bipolar chest lead to maximize R-wave amplitude and reduce electromyographic interference. S102 is equipped with a dual-wavelength light-emitting diode with emission wavelengths of 660 nm and 905 nm and a corresponding photodetector for collecting photoplethysmography pulse waves. The sensor is attached to the fingertip of the left index finger of the target object, and the light source output is ensured to be stable through constant current source drive and temperature compensation circuit.

[0028] The principle and effect of the above technical solution are as follows: through hardware-level optimization design, the signal-to-noise ratio and stability of key bioelectric and optical signals are ensured from the source, laying a solid physical data foundation for subsequent in-depth feature extraction and accurate analysis, and avoiding the problem of failure of subsequent advanced analysis due to poor signal quality.

[0029] In one embodiment, S20 further includes: S201, the preprocessing of the electrocardiogram signal also includes the use of a threshold denoising method based on wavelet transform. Specifically, the 'sym4' wavelet basis is used for 5-level decomposition, and the detail coefficients of the first to third levels are processed using a soft threshold function to suppress high-frequency electromyographic noise without causing QRS wave distortion. S202 eliminates motion artifacts in photoplethysmography pulse waves by using a normalized minimum mean square adaptive filter with the vector sum of the triaxial accelerometer signals as the reference noise input. The filter order is set to 32 and the step size parameter is set to 0.01.

[0030] The principle and effect of the above technical solution are as follows: Advanced signal processing technology is used to specifically remove the most significant interference components in various signals. Wavelet denoising can better preserve the subtle morphological changes in the ECG waveform (such as T-wave alternans), while adaptive filtering can dynamically track and eliminate pulse wave baseline drift and morphological changes caused by motion, thereby obtaining a purer core waveform that reflects the true physiological state and significantly improving the measurement accuracy of subsequent characteristic parameters.

[0031] In one embodiment, the multidimensional feature parameters in S30 specifically include extracting the QT interval, T wave peak-end interval, T wave amplitude and morphological asymmetry coefficient, and beat-by-beat RR interval sequence from the electrocardiogram signal and calculating the low-frequency power to high-frequency power ratio of heart rate variability; calculating the pulse wave conduction time from the synchronous electrocardiogram R wave peak and pulse wave trough; locating the peak point of the first heart sound S1 from the heart sound signal and calculating the electromechanical delay time between it and the peak value of the electrocardiogram R wave; and extracting the ratio of the diastolic notch height to the peak value of the main wave from the photoplethysmography pulse wave signal, which is defined as the reflection enhancement index.

[0032] The S30 also includes: S301, the calculation method of the T-wave morphological asymmetry coefficient is as follows: within the time limit from the start point to the end point of the T-wave, taking the peak value of the T-wave as the boundary, calculate the absolute value of the difference between the area of ​​the waveform before the peak value and the area of ​​the waveform after the peak value, and then divide it by the total area of ​​the entire T-wave waveform. S302, the calculation of the electromechanical delay time is as follows: In the synchronized electrocardiogram signal and heart sound signal, the peak point time t_R of the electrocardiogram R wave and the peak point time t_S1 of the principal component of the subsequent first heart sound S1 are accurately detected. Then the delay time EMD = t_S1 - t_R. This calculation is taken as the average value within 5 consecutive stable heartbeat cycles. S303 calculates the frequency domain index of heart rate variability from the beat-by-beat RR interval sequence. Specifically, the Lomb-Scargle periodogram estimation method is used to adapt to the non-uniformly sampled RR interval sequence. The power in the 0.04 Hz to 0.15 Hz frequency band is calculated as the low-frequency power, and the power in the 0.15 Hz to 0.4 Hz frequency band is calculated as the high-frequency power. The ratio LF / HF is then obtained.

[0033] The principle and effects of the above technical solution are as follows: the extracted feature parameters directly point to the deep pathophysiological state of the heart. T-wave morphological asymmetry may reflect ventricular repolarization dispersion; electromechanical delay directly measures the coupling efficiency from electrical excitation to mechanical contraction and is a sensitive indicator of cardiac function; the Lomb-Scargle method for calculating HRV can effectively handle non-uniform interval sequences caused by arrhythmias or noise interference, resulting in a more reliable assessment of autonomic tension. These deep features provide rich and valuable input for subsequent logical fusion.

[0034] In one embodiment, S40 further includes: S401, the signal-cell function layer, is used to assess the state of excitation-contraction coupling function of cardiomyocyte ion channels: its inputs are QT interval, T wave end-peak interval, T wave morphological asymmetry coefficient, and electromechanical delay time; this layer has a built-in computational network based on cellular electrophysiology knowledge, which estimates the relative intensity of transmembrane potassium ion current during phase 3 repolarization of ventricular myocyte action potential by performing Fridricia correction on QT interval and heart rate, and substituting the corrected QTc and T wave end-peak interval into empirical formulas; at the same time, the electromechanical delay time is regressed and corrected with heart rate to obtain a heart rate-independent electromechanical coupling index, which is related to the intracellular calcium ion transient rate and sarcoplasmic reticulum functional state; S402, the organ-regulation loop layer, is used to evaluate the stability of cardiac autonomic regulation and hemodynamic feedback. Its inputs are the heart rate variability LF / HF ratio, pulse wave conduction time, reflex enhancement index, and electromechanical coupling index calculated by layer S401. This layer simulates cardiovascular pressure reflex and mechano-electro-feedback loops, maps the LF / HF ratio to sympathetic-vagal balance tension, establishes a nonlinear mapping relationship between the reflex enhancement index and peripheral vascular tension, and calculates a quantitative value of "ventricular-vascular coupling efficiency" by integrating the electromechanical coupling index and pulse wave conduction time. S403, the system-compensation state layer, is used to comprehensively assess the overall compensatory capacity and decompensation risk of the circulatory system. Its inputs are the ion current intensity index and electromechanical coupling index output from layer S401, and the autonomic nervous balance tension and ventricular-vascular coupling efficiency values ​​output from layer S402. This layer uses a fuzzy inference system trained with clinical data, taking the above four quantitative indicators as input variables and outputting two results: one is the control logic stability index ranging from 0 to 1, and the other is the qualitative label of the current dominant imbalance mode, which includes electroremodeling-dominant, mechanical remodeling-dominant, autonomic nervous disorder, and mixed types.

[0035] The principle and effects of the above technical solution are as follows: This step is the core innovation of this invention, establishing a coherent logical explanation chain from microscopic cellular function to macroscopic system manifestations. The signal-cell function layer reverse-engineers surface waveform characteristics to cellular ion channels and calcium circulation levels, revealing the abnormal electrophysiological or contractile material basis. The organ-regulatory circuit layer integrates neural regulation and mechanical coupling, assessing the efficiency of interactions between organs. Finally, the system-compensatory state layer provides a comprehensive evaluation of overall stability. This hierarchical, physiologically logical fusion approach enables the system to understand the "cause" behind abnormal signs rather than just the "phenomenon," providing a pathophysiological basis for accurate early warning.

[0036] In one embodiment, S50 further includes: S501, the early warning decision is divided into three levels: attention level early warning, advanced early warning, and emergency early warning; the decision input is the "control logic stability index" and "current dominant imbalance mode" label output by the S403 layer, as well as the real-time trend of the T-wave morphology asymmetry coefficient and electro-mechanical delay time extracted by S30. S502, Establish an early warning rule base, which contains a set of logical judgments for the "precursor state of malignant empty arrhythmia"; S503 monitors the above parameters in real time, and triggers an alert of the corresponding level when any combination of conditions in the rule set is met.

[0037] In one embodiment, S60 further includes: S601, the warning output content is presented in the form of a structured data packet, which includes at least: warning level code, trigger timestamp, current "control logic stability index" value, and "current dominant imbalance mode" label; S602, and output a list of key evidence chain parameters that led to this warning. The list should include at least three parameters that deviate most significantly from the baseline and their percentage deviation, as well as a description of the logical chain stage in which the judgment is made. The S603's output methods include displaying on a local monitoring terminal, generating voice prompts, and sending encrypted data packets to a remote monitoring center server via a wireless network.

[0038] In one specific embodiment, in S50, if the current dominant imbalance mode label is electric reconfiguration dominant and the control logic stability index is between 0.7 and 0.85, then the electric reconfiguration risk assessment sub-process is adopted, and the specific steps are as follows: SA51: Examine the trend of the T-wave morphological asymmetry coefficient over the past 10 minutes and calculate its linear regression slope; SA52: Check whether the estimated relative intensity of transmembrane potassium ion current during phase 3 repolarization of ventricular myocyte action potential is lower than 65% of the historical baseline. SA53: Check whether the high-frequency power in heart rate variability shows a progressive decreasing trend, with a decrease of more than 30% of its average value over the previous 15 minutes in the past 5 minutes; SA54: If the slope of SA51 is positive and greater than 0.01 per minute, the SA52 condition, and the SA53 condition are met simultaneously, it is judged as "electrical remodeling progress, moderate risk of malignant arrhythmia", triggering a high-level warning; if only two of them are met, a level-of-concern warning is triggered.

[0039] The principle of the described electrical remodeling risk assessment sub-process is to make a comprehensive judgment by combining three key logical links of electrophysiological instability: increased repolarization dispersion, weakened repolarization ion current, and decreased vagal nerve protective tone.

[0040] In S50, if the current dominant imbalance mode is labeled as mechanical reconfiguration dominant and the control logic stability index is less than 0.7, then a pump failure compensation risk assessment sub-process is adopted. The specific steps of the sub-process are as follows: SB51: Calculate the percentage increase in current electromechanical delay time relative to the individualized baseline; SB52: Calculate the "ventricular-vascular coupling efficiency" value and check if it has decreased by more than 20% in the past 30 minutes; SB53: Monitor the reflection enhancement index to determine if it shows a rapid upward trend (an increase of more than 10% within 5 minutes), indicating a sudden increase in afterload; SB54: If the extension of SB51 by more than 15%, SB52, and SB53 are all met simultaneously, it is determined to be an "acute pump function decompensation critical state" and an emergency warning is triggered; if SB51 and SB52 are met simultaneously, but SB53 is not met, an advanced warning is triggered.

[0041] The principle of the pump failure compensation risk assessment sub-process is to capture the progressive deterioration logic chain from decreased myocardial contractility to ventricular-vascular mismatch to inappropriate afterload, and provide early warning before hemodynamic collapse.

[0042] In S50, if the "control logic stability index" is lower than 0.6, regardless of the dominant imbalance mode, an emergency assessment sub-process for system instability is adopted. The specific steps of the sub-process are as follows: SC51: Continuously monitor the stability index of the control logic. If it decreases linearly within 3 minutes and the slope is less than -0.1 per minute; SC52: Simultaneously check whether the heart rate variability LF / HF ratio shows dramatic fluctuations (standard deviation increases by approximately 200% in the past 2 minutes) or a sharp decrease; SC53: Check for irregular fluctuations in pulse wave conduction time that are unrelated to the respiratory cycle; SC54: If conditions SC51, SC52 and SC53 are met simultaneously, it is determined that "the autonomic nervous and mechanical regulation system is on the verge of collapse", immediately triggering the highest level of emergency warning and initiating continuous real-time transmission of vital signs waveforms.

[0043] In one embodiment, S10 further includes configuring an integrated environmental sensor module, which includes a high-precision digital temperature and humidity sensor, a barometer, and an ultraviolet sensor, connected to the physiological signal acquisition motherboard via an I2C bus, and is triggered by a unified high-precision real-time clock for synchronous sampling, with an environmental data sampling rate of 1Hz.

[0044] In one embodiment, S20 further includes loading or creating the patient's "individual and environmental context configuration" during system initialization or via the cloud. This configuration is a structured profile, such as: {"Patient ID": "P001", "Age": 55, "Dietary Habits": "High Carbohydrate", "Residential Environment": "High Altitude (3000m)", "Risk Markers": ["Hypertension", "Improper Glucose Tolerance"]}. The preprocessing module and subsequent analysis engine will call this profile to contextualize the data. For example, for patients marked as "High Altitude" and "High Carbohydrate", the system will automatically load a targeted analysis model, focusing on monitoring the changes in their reflex enhancement index and low-frequency power of heart rate variability during the day and after meals, and comparing them with a high-altitude hypoxia adaptation model.

[0045] In one embodiment, the trend analysis in S30 specifically includes: the system maintaining a database of long-term characteristic parameters for the patient. When analyzing real-time data, the engine not only calculates the instantaneous values ​​of the parameters, but also calculates their deviations relative to the individual's historical baseline and relative to the statistical baseline of a population of the same age, dietary habits, and environment. For example, for a patient from a high-altitude region with a long-term high-carbohydrate diet, the system will specifically analyze the flattening trend of the "pulse wave transit time" in the diurnal rhythm, as well as the magnitude and duration of the abnormal increase in the "reflex enhancement index" after meals, quantifying these trends into an "arterial stiffness progression risk index" and a "postprandial endothelial function damage risk index," as inputs for subsequent fusion and early warning.

[0046] In one embodiment, the multi-level early warning decision engine of S50 integrates the trend risk index from S30 on the basis of the original physiological logic judgment. For example, when the system determines that the current dominant imbalance mode is "mechanical remodeling dominant type", if it also detects that the patient is in a "high altitude hypoxia" environment and that the "arterial stiffness progression risk index" is at a high level, the warning level will be upgraded from "advanced warning" to "emergency warning", and the warning evidence chain will clearly indicate that "the combination of high altitude hypoxia environment and long-term arteriosclerosis leads to a sharp increase in the risk of acute heart failure".

[0047] The principle of the system instability emergency assessment sub-process is to identify an extreme dangerous state in which multiple control loops are simultaneously disordered and the system feedback control is about to fail or has already failed. This state is a direct prelude to sudden cardiac death and requires the fastest possible intervention.

[0048] This invention provides a vital sign monitoring and early warning method based on multi-source data fusion. By constructing a three-level physiological control logic fusion model of signal-organ-system, it realizes the interpretation of the complete narrative of the evolution of a living organism from homeostasis to decompensation from multi-source physiological data. Its core beneficial effect is to transform the traditional threshold-based lagging and high false alarm early warning into an early, accurate, and interpretable risk prospect based on the deduction of physiological regulation logic chain. This method can capture the logical linkage between deep parameters such as myocardial cell ion channel function, electromechanical coupling efficiency, autonomic nerve tension, and vascular coupling status. Thus, before the vital signs deteriorate significantly, it can identify the specific stage of the system in strong compensation, weak compensation, or decompensation threshold, and determine the dominant imbalance mode (such as electrical remodeling, mechanical disorder, etc.). This not only greatly reduces false alarms caused by individual differences and physiological fluctuations, but also provides clear pathophysiological basis and key time windows for clinical intervention, realizing a leap in intelligent monitoring capabilities.

[0049] Accordingly, such as Figure 2 As shown, based on a vital sign monitoring and early warning method based on multi-source data fusion, this embodiment of the invention also provides a vital sign monitoring and early warning system based on multi-source data fusion, realizing the vital sign monitoring and early warning method based on multi-source data fusion of this embodiment of the invention. The system includes: The data acquisition and synchronization module is used to control the integrated multi-physiological sensors and environmental sensors to synchronously acquire surface electrocardiogram signals, photoplethysmography pulse wave signals, apical heart sound signals, chest wall impedance respiratory signals, and ambient temperature, humidity, air pressure and ultraviolet intensity data of the target object according to the preset monitoring mode. The preset monitoring mode is the data collection and processing intensity level selected by the system according to the application scenario; The multiple physiological sensors are integrated into a wearable device or bedside monitoring terminal. The synchronous acquisition is triggered by a built-in high-precision real-time clock, which simultaneously starts the analog-to-digital conversion circuits of each sensor channel. The sampling rate of the surface electrocardiogram signal is 1000 Hz, the sampling rate of the photoplethysmography pulse wave signal is 500 Hz, the sampling rate of the heart sound signal is 4000 Hz, and the sampling rate of the chest wall impedance signal is 100 Hz. The data acquisition and synchronization module includes: an electrocardiogram acquisition unit, used to acquire surface electrocardiogram signals through a differential amplifier with high input impedance and a common-mode rejection ratio greater than 120 dB, in a modified chest bipolar lead configuration; The photoplethysmography (PPG) acquisition unit is used to acquire the blood volume change signal at the fingertip by emitting dual-wavelength light-emitting diodes with wavelengths of 660 nm and 905 nm and corresponding photodetectors. The sensor is attached to the fingertip of the left index finger of the target object, and the light source output is ensured to be stable by constant current source drive and temperature compensation circuit.

[0050] The signal preprocessing and grouping analysis module is used to store and retrieve patients' individualized files, environmental files and risk files. Based on the individual and environmental context configuration, it performs parallel preprocessing and context grouping on multi-source raw data to generate clean physiological waveform data after denoising and artifact elimination and aligned environmental data. The parallel preprocessing involves simultaneously executing the unique filtering and enhancement algorithms for each channel signal. The denoising and artifact elimination includes suppressing power frequency interference, electromyographic noise, motion artifacts, and baseline drift. The signal preprocessing and grouping analysis module includes: an electrocardiogram (ECG) signal preprocessing unit, which performs bandpass filtering from 0.05 Hz to 150 Hz and threshold denoising based on wavelet transform on the ECG signal in sequence. Specifically, it uses the 'sym4' wavelet basis to perform 5-level decomposition and uses a soft thresholding function to process the detail coefficients of the first to third levels. The pulse wave signal preprocessing unit is used to sequentially perform bandpass filtering from 0.1 Hz to 20 Hz on the photoplethysmography pulse wave signal, and to use a normalized minimum mean square adaptive filter to eliminate motion artifacts. The vector sum of the triaxial accelerometer signal is used as the reference noise input, the filter order is set to 32, and the step size parameter is 0.01.

[0051] The multi-dimensional feature extraction and trend analysis engine is used to calculate a set of feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from preprocessed and grouped data, and analyze the dynamic evolution trend of this set under different individual and environmental dimensions. The set of feature parameters includes time domain, frequency domain, and time-frequency domain indices; The multi-dimensional feature extraction and trend analysis engine includes: an electrophysiological feature calculation unit, used to extract QT interval, T wave peak-end interval, T wave amplitude and morphological asymmetry coefficient, and beat-by-beat RR interval sequence from the electrocardiogram signal; the calculation method of the T wave morphological asymmetry coefficient is as follows: within the time limit from the beginning to the end of the T wave, taking the T wave peak value as the boundary, calculate the absolute value of the difference between the area of ​​the waveform before the peak value and the area of ​​the waveform after the peak value, and then divide it by the total area of ​​the entire T wave waveform; The mechanical and hemodynamic characteristic calculation unit is used to calculate the electromechanical delay time, pulse wave conduction time, and reflection enhancement index. The calculation of the electromechanical delay time is specifically as follows: in the synchronized electrocardiogram and heart sound signals, the peak time t_R of the electrocardiogram R wave and the peak time t_S1 of the subsequent first heart sound S1 are detected, and EMD = t_S1 - t_R is calculated and averaged over 5 consecutive stable heartbeat cycles. The autonomic nervous system characteristic calculation unit is used to calculate heart rate variability indices from beat-by-beat RR interval sequences. It uses the Lomb-Scargle periodogram estimation method to calculate the low-frequency power in the 0.04 Hz to 0.15 Hz frequency band and the high-frequency power in the 0.15 Hz to 0.4 Hz frequency band, and obtains the ratio LF / HF.

[0052] A hierarchical physiological control logic fusion engine is used to receive the set of feature parameters, perform state deduction through a three-level fusion model, and output a quantitative index and qualitative label characterizing the stability of the system. The three-level fusion model includes a signal-cell function layer, an organ-regulatory circuit layer, and a system-compensatory state layer. The hierarchical physiological control logic fusion engine includes: a signal-cell function layer for assessing the functional state of cardiomyocyte ion channels and excitation-contraction coupling; its inputs are QT interval, T wave end-peak interval, T wave morphology asymmetry coefficient, and electromechanical delay time; this layer incorporates a computational network based on cellular electrophysiology knowledge, which estimates the relative intensity of transmembrane potassium ion current during phase 3 repolarization of ventricular myocyte action potential by performing Fridricia correction on QT interval and heart rate, and substituting the corrected QTc and T wave end-peak interval into an empirical formula; simultaneously, it performs regression correction on electromechanical delay time and heart rate to obtain a heart rate-independent electromechanical coupling index; The organ-regulatory loop layer is used to assess the stability of cardiac autonomic regulation and hemodynamic feedback. Its inputs are the heart rate variability LF / HF ratio, pulse wave conduction time, reflex enhancement index, and electromechanical coupling index calculated by the signal-cell function layer. This layer simulates the cardiovascular baroreflex and mechano-electro-feedback loop, maps the LF / HF ratio to sympathetic-vagal balance tension, establishes a nonlinear mapping relationship between the reflex enhancement index and peripheral vascular tension, and calculates the ventricular-vascular coupling efficiency value by integrating the electromechanical coupling index and pulse wave conduction time. The system-compensation state layer is used to comprehensively assess the overall compensatory capacity and decompensation risk of the circulatory system. Its inputs are the ion current intensity index, electromechanical coupling index, autonomic nervous balance tension, and ventricular-vascular coupling efficiency value. This layer uses a fuzzy inference system trained with clinical data, and outputs a control logic stability index ranging from 0 to 1, as well as a qualitative label of the current dominant imbalance mode. The label includes electroremodeling-dominant, mechanical remodeling-dominant, autonomic nervous disorder, and mixed types.

[0053] A multi-level early warning decision engine is used to perform graded early warning judgments based on the stability index of the control logic and the label of the current dominant imbalance mode, combined with the real-time trend and trend analysis results of the feature parameters. The tiered early warning system includes attention-level early warning, advanced early warning, and emergency early warning. The multi-level early warning decision engine includes: an early warning rule base, which stores a set of logical judgment rules corresponding to different pathophysiological evolution chains; The risk assessment sub-process execution unit is used to invoke specific risk assessment logic that matches the current state. The risk assessment sub-process execution unit includes: an electrical remodeling risk assessment sub-process, which is executed when the current dominant imbalance mode is labeled "electrical remodeling dominant" and the control logic stability index is between 0.7 and 0.85. The specific steps are: checking whether the slope of the T-wave morphology asymmetry coefficient change trend in the past 10 minutes is greater than 0.01 per minute; checking whether the estimated value of the relative intensity of potassium ion flow is lower than 65% of the historical baseline value; checking whether the high-frequency power of heart rate variability has decreased by more than 30% of its average value in the past 15 minutes in the past 5 minutes; if all three conditions are met, a high-level warning is triggered; if only two conditions are met, a level-of-concern warning is triggered. The pump failure compensation risk assessment sub-process is executed when the current dominant imbalance mode is labeled "mechanical remodeling dominant" and the control logic stability index is below 0.7. The specific steps are as follows: calculate the percentage increase in electromechanical delay time relative to the individualized baseline; calculate the percentage decrease in ventricular-vascular coupling efficiency value over the past 30 minutes; monitor the increase in reflex enhancement index over 5 minutes; if the delay increase exceeds 15%, the coupling efficiency decreases by more than 20%, and the reflex enhancement index increases by more than 10%, an emergency warning is triggered; if only the first two conditions are met, an advanced warning is triggered. The system instability emergency assessment sub-process is executed when the control logic stability index is below 0.6. The specific steps are as follows: monitor whether the index shows a linear decrease within 3 minutes with a slope of less than -0.1 per minute; check whether the standard deviation of the heart rate variability LF / HF ratio has increased by approximately 200% or decreased sharply in the past 2 minutes; check whether there are random fluctuations in pulse wave conduction time that are unrelated to the respiratory cycle; if all three conditions are met, an emergency warning is immediately triggered and continuous real-time transmission of vital signs waveforms is initiated.

[0054] The results output and communication module is used to generate and transmit structured early warning decision results that contain information on the correlation between environmental and individual risk factors; The structured early warning decision results include at least the early warning level code, trigger timestamp, control logic stability index value, and current dominant imbalance mode label; The transmission methods include wired and wireless communication protocols; The result output and communication module includes: an evidence chain encapsulation unit, which is used to automatically generate and associate a list of key evidence chain parameters that led to the warning when the warning is triggered. The list includes at least three parameters that deviate most significantly from the baseline and their deviation percentages, as well as a description of the logical chain stage in which the judgment is made. The multi-mode output unit is used to display the warning results and evidence chain on the local monitoring terminal screen, generate voice prompts, and send encrypted data packets to the remote monitoring center server via a wireless network.

[0055] The working principle of the above technical solution is as follows: A data acquisition and synchronization module acquires multi-source physiological signals such as electrocardiogram, pulse wave, heart sound, and respiratory impedance with high precision; a signal preprocessing and grouping analysis module performs targeted filtering and noise reduction on each signal to obtain a clean waveform; a multi-dimensional feature extraction and trend analysis engine calculates deep electrophysiological, mechanical, and neural regulatory features from the waveform; a hierarchical physiological control logic fusion engine fuses feature parameters from cell function, organ regulation to system compensation in three layers, outputting a stability index and imbalance mode; a multi-level early warning decision engine calls a targeted risk assessment subprocess for precise graded early warning based on the fusion results and real-time trends; and a result output and communication module encapsulates and outputs early warning information with a complete chain of evidence.

[0056] The beneficial effects of the above technical solution are as follows: This system ensures high quality and timeliness of multi-source data through hardware synchronous acquisition and deep signal processing. The core hierarchical logic fusion engine breaks through the limitations of traditional data stacking, realizing the simulation and deduction of the intrinsic regulatory logic of the physiological system, and can explain the causes of abnormal vital signs from a mechanistic level. The multi-level early warning decision engine, combined with specific risk assessment sub-processes, realizes the leap from "abnormality detection" to "risk classification and attribution", significantly improving the accuracy, foresight, and interpretability of early warning, effectively reducing false alarms and false negatives, and providing intelligent decision support for early intervention of cardiogenic risks.

[0057] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A vital sign monitoring and early warning method based on multi-source data fusion, characterized in that, include: S10 is equipped with a multi-source physiological data acquisition module, which simultaneously acquires the target object's surface electrocardiogram signal, photoplethysmography pulse wave signal, apical heart sound signal and chest wall impedance respiratory signal, as well as the temperature, humidity, atmospheric pressure and ultraviolet intensity data of the surrounding environment. S20 performs parallel preprocessing and group analysis on the collected multi-source raw signals according to the preset individual and environmental context configuration; S30 extracts multi-dimensional feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from the preprocessed signal, and analyzes the evolution trend of feature parameters under different environmental and individual factors based on historical data and individual risk profiles. S40 constructs a three-level physiological control logic fusion model, which deeply integrates multi-dimensional feature parameters and their trend information and performs state inference; the model includes a signal-cell function layer, an organ-regulatory circuit layer, and a system-compensatory state layer; S50, based on the quantitative indices, qualitative labels, and trend analysis results output by the system-compensation state layer, executes multi-level early warning decisions; S60 outputs the early warning decision results, including the early warning level, the currently determined dominant pathophysiological mechanism, the current logical chain stage, the associated environmental and individual risk factors, a list of key abnormal characteristic parameters, and the degree of deviation from the normal reference range.

2. The method of claim 1, wherein, Parallel preprocessing of the acquired multi-source raw signals includes bandpass filtering of ECG signals from 0.05 Hz to 150 Hz, bandpass filtering of photoplethysmography (PPG) signals from 0.1 Hz to 20 Hz and simultaneous motion artifact elimination based on adaptive filtering using triaxial accelerometer data, bandpass filtering of heart sound signals from 25 Hz to 400 Hz, and bandpass filtering of chest wall impedance signals from 0.05 Hz to 2 Hz.

3. The method of claim 1, wherein, S10 also includes: S101 is equipped with a differential amplifier for acquiring surface electrocardiogram signals, and the lead configuration is a bipolar chest lead. S102 is equipped with dual-wavelength light-emitting diodes with emission wavelengths of 660 nm and 905 nm and a photodetector for collecting photoplethysmography pulse waves. The sensor is driven by a constant current source and a temperature compensation circuit to ensure stable light source output. It is equipped with an integrated environmental sensor module to synchronously collect the temperature, relative humidity, atmospheric pressure and ultraviolet index of the microenvironment in which the target object is located, and synchronize it with the physiological signal acquisition system.

4. The method of claim 1, wherein, S20 also includes: S201, the preprocessing of the electrocardiogram signal also includes the use of a threshold denoising method based on wavelet transform, specifically using the sym4 wavelet basis for 5-level decomposition, and using a soft threshold function to process the detail coefficients of the first to third levels. S202, for eliminating motion artifacts in photoplethysmography pulse waves, adopts a normalized minimum mean square adaptive filter, using the vector sum of the triaxial accelerometer signals as the reference noise input, with the filter order set to 32 and the step size parameter set to 0.01; The individual and environmental context configuration includes a patient individual profile containing age, gender, history of underlying diseases, history of long-term medication, and dietary habit tags; an environmental profile used to identify normal and abnormal environments; and a risk profile used to define baseline characteristic parameters and risk warning threshold adjustment strategies for a specific population in a specific environment. The grouping analysis refers to the preprocessing process automatically selecting or adjusting filtering parameters, feature extraction priorities, and baseline reference ranges based on context configuration.

5. The method of claim 1, wherein, The multi-dimensional feature parameters in S30 include: Extract the QT interval, T wave peak-end interval, T wave amplitude and morphological asymmetry coefficient, and beat-by-beat RR interval sequence from the electrocardiogram signal and calculate the low-frequency power to high-frequency power ratio of heart rate variability. Calculate pulse wave conduction time from the synchronous ECG R wave peak and pulse wave trough; Locate the peak point of the first heart sound S1 from the heart sound signal and calculate the electromechanical delay time between it and the peak value of the ECG R wave; The ratio of the diastolic notch height to the peak value of the main wave is extracted from the photoplethysmography (PPG) signal and defined as the reflection enhancement index. Establish short-term correlation models between characteristic parameters and environmental parameters, as well as statistical association models between characteristic parameters and individual long-term risk factors; For specific risk populations, the long-term deviation trend of their cardiovascular function-related characteristic parameters compared with the baseline of healthy people of the same age is calculated, and the acute risk superposition effect is assessed in combination with real-time environmental stress.

6. The method of claim 1, wherein, The S30 also includes: S301, The calculation method for the T-wave morphological asymmetry coefficient is as follows: within the time limit from the start to the end of the T-wave, taking the peak value of the T-wave as the boundary, calculate the absolute value of the difference between the area of ​​the waveform before the peak value and the area of ​​the waveform after the peak value, and then divide it by the total area of ​​the entire T-wave waveform. S302, The calculation of the electromechanical delay time is as follows: In the synchronized electrocardiogram signal and heart sound signal, the peak point time t_R of the electrocardiogram R wave and the peak point time t_S1 of the principal component of the first heart sound S1 are accurately detected. The delay time EMD = t_S1 - t_R is calculated and the average value is taken over 5 consecutive stable heartbeat cycles. S303 calculates the frequency domain index of heart rate variability from beat-by-beat RR interval sequences. Specifically, it uses the Lomb-Scargle periodogram estimation method to adapt to non-uniformly sampled RR interval sequences. The power in the 0.04 Hz to 0.15 Hz band is calculated as the low-frequency power, and the power in the 0.15 Hz to 0.4 Hz band is calculated as the high-frequency power. The ratio LF / HF is then obtained.

7. The method of claim 1, wherein, The S40 also includes: S401, the signal-cell function layer, is used to assess the functional status of cardiomyocyte ion channels and excitation-contraction coupling. Inputs include QT interval, T-wave end-peak interval, T-wave morphology asymmetry coefficient, and electromechanical delay time. It incorporates a computational network based on cellular electrophysiology. By performing Fridricia correction on the QT interval and heart rate, and substituting the corrected QTc and T-wave end-peak interval into an empirical formula, it estimates the relative intensity of transmembrane potassium ion current during phase 3 repolarization of the ventricular myocyte action potential. Regression correction is then performed on the electromechanical delay time and heart rate to obtain a heart rate-independent electromechanical coupling index. S402, the organ-regulation loop layer, is used to assess the stability of cardiac autonomic regulation and hemodynamic feedback. The inputs are the heart rate variability (LF / HF ratio), pulse wave conduction time, reflex enhancement index, and electromechanical coupling index. It simulates cardiovascular baroreflex and mechano-electro-feedback loops, maps the LF / HF ratio to sympathetic-vagal balance tension, establishes a nonlinear mapping relationship between the reflex enhancement index and peripheral vascular tension, and calculates the quantification value of ventricular-vascular coupling efficiency by integrating the electromechanical coupling index and pulse wave conduction time. S403, the system-compensation state layer, is used to comprehensively assess the overall compensatory capacity and decompensation risk of the circulatory system. The inputs are the ion current intensity index and electromechanical coupling index output from layer S401, and the autonomic nervous balance tension and ventricular-vascular coupling efficiency values ​​output from layer S402. The layer adopts a fuzzy inference system and outputs the control logic stability index and the qualitative label of the current dominant imbalance mode. The labels include electroremodeling-dominant, mechanical remodeling-dominant, autonomic nervous disorder, and mixed types.

8. The method as described in claim 1, characterized in that, The S50 also includes: S501, the early warning decision is divided into attention level early warning, advanced early warning and emergency early warning; the decision input is the control logic stability index and the current dominant imbalance mode label, as well as the real-time trend of T-wave morphology asymmetry coefficient and electro-mechanical delay time; S502, Establish an early warning rule base, which contains a set of logical judgments; S503 monitors parameters in real time and triggers alerts.

9. The method as described in claim 1, characterized in that, The S60 also includes: S601, the warning output includes the warning level code, trigger timestamp, current control logic stability index value and current dominant imbalance mode label; S602, and outputs a list of key evidence chain parameters that led to the warning. The list includes the three parameters that deviated most significantly from the baseline and their percentage deviation, as well as a description of the logical chain stage in which the judgment is made. The S603 output methods include displaying on a local monitoring terminal, generating voice prompts, and sending encrypted data packets to a remote monitoring center server via a wireless network.

10. A vital signs monitoring and early warning system based on multi-source data fusion, used to implement the method described in any one of claims 1-9, characterized in that, The system includes: The data acquisition and synchronization module is used to control the integrated multi-physiological sensors and environmental sensors to synchronously acquire surface electrocardiogram signals, photoplethysmography pulse wave signals, apical heart sound signals, chest wall impedance respiratory signals, and ambient temperature, humidity, air pressure and ultraviolet intensity data of the target object according to the preset monitoring mode. The signal preprocessing and grouping analysis module is used to store and retrieve patients' individualized files, environmental files and risk files. Based on the individual and environmental context configuration, it performs parallel preprocessing and context grouping on multi-source raw data to generate clean physiological waveform data after denoising and artifact elimination and aligned environmental data. The multi-dimensional feature extraction and trend analysis engine is used to calculate a set of feature parameters reflecting cardiac electrical activity, mechanical activity, autonomic nervous regulation and hemodynamic state from preprocessed and grouped data, and analyze the dynamic evolution trend of this set under different individual and environmental dimensions. The hierarchical physiological control logic fusion engine is used to receive the set of feature parameters and their trend information, perform state deduction through a three-level fusion model, and output quantitative indices and qualitative labels that characterize the stability of the system. The three-level fusion model includes the signal-cell function layer, the organ-regulatory circuit layer, and the system-compensation state layer. A multi-level early warning decision engine is used to perform graded early warning judgments based on the control logic stability index and the current dominant imbalance mode label, combined with the real-time trend and trend analysis results of characteristic parameters. The results output and communication module is used to generate and transmit structured early warning decision results that contain information on the correlation between environmental and individual risk factors.

Images