Integrated pediatric vital sign intelligent monitoring and early warning method and system

Abstract

The invention discloses an integrated pediatric vital sign intelligent monitoring and early warning method and system, and relates to the technical field of biomedical signal processing. The method is used for solving the problems of large monitoring interference and early warning lag in a dynamic environment. The method comprises the following steps: firstly, synchronously collecting a multi-modal signal to generate a motion intensity envelope signal and an R wave peak value interval sequence, carrying out orthogonal projection denoising on a photoelectric volume pulse wave signal by utilizing motion envelope based on a recursive least square algorithm, and outputting a reconstructed blood oxygen waveform; then, calculating a cross correlation coefficient of a heart rate trend and a motion envelope, quantifying a physiological coupling degree to generate an independent heart rate anomaly index, and accurately identifying a non-motion related risk; finally, the real-time blood oxygen value is analyzed, the index is used for dynamically increasing the low oxygen threshold value so as to tighten the alarm boundary, physiological parameter linkage grading early warning is achieved, and the recognition sensitivity of the compensation period risk is improved while interference resistance is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical signal processing, in particular to an integrated pediatric vital sign intelligent monitoring and early warning method and system. BACKGROUND

[0002] In modern pediatric clinical nursing and family rehabilitation scenarios, continuous monitoring of vital signs is a key means to assess the evolution of the child's condition and prevent the occurrence of critical illness. Because the physiological functions of children and infants have not yet matured, the disease often has the characteristics of acute onset and rapid progression, so real-time monitoring of their heart rate, blood oxygen and movement state at all times is of great importance to ensure medical safety. Traditional monitoring modes rely on complex lead wire connections or high-frequency manual measurements by medical staff, while under the trend of intelligent medical care, using flexible wearable devices for unobtrusive monitoring has become the main development direction to reduce the burden on medical staff and improve the comfort of sick children.

[0003] However, existing pediatric monitoring technologies have significant drawbacks in actual application. Existing devices mainly use general linear filtering algorithms, which are difficult to effectively distinguish non-stationary motion noise that overlaps with the physiological signal spectrum. When the child cries or moves violently, the photoplethysmogram signal is easily contaminated by motion artifacts, resulting in false drops in blood oxygen readings or waveform distortion. In addition, existing early warning mechanisms mostly use single static threshold decision logic, without correlating physiological parameters with behavioral actions for analysis, and are unable to identify physiological heart rate increases caused by body movement, nor can they issue warnings in advance during the compensatory period when the heart rate is abnormal but the blood oxygen has not yet fallen below the conventional threshold, which results in a very high false alarm rate (alarm fatigue) and potential missed diagnosis risk in clinical practice. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides an integrated pediatric vital sign intelligent monitoring and early warning method and system, which solves the problems of the above background technology.

[0005] To achieve the above object, the present application is implemented by the following technical solutions: an integrated pediatric vital sign intelligent monitoring and early warning method, comprising the following steps: S1. synchronously collecting the electrocardiosignal, the photoplethysmographic signal and the three-axis acceleration signal of a child, performing vector synthesis on the three-axis acceleration signal and calculating the root mean square envelope to generate a motion intensity envelope signal, while extracting the peak feature points of the electrocardiosignal to construct an R-wave peak interval sequence; S2. calling the motion intensity envelope signal as a non-stationary noise reference source, projecting and denoising the photoplethysmographic signal based on the recursive least squares adaptive filtering algorithm, removing the motion artifacts in the photoplethysmographic signal through orthogonal projection of the fluctuation characteristics of the motion intensity envelope signal, and outputting the reconstructed blood oxygen waveform; S3. interpolating the R-wave peak interval sequence, matching the sampling rate of the motion intensity envelope signal, calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R-wave peak interval sequence within a preset physiological reaction lag window, comparing the maximum cross-correlation coefficient with a preset physiological coupling threshold value, and generating an independent heart rate abnormality index when the R-wave peak interval sequence shows a rising trend of heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold value; S4. analyzing the reconstructed blood oxygen waveform to calculate the real-time blood oxygen saturation value, taking the independent heart rate abnormality index as an adjustment factor of the blood oxygen alarm threshold value, raising the preset hypoxia threshold value to tighten the alarm boundary when the independent heart rate abnormality index is generated, and triggering an alarm instruction if the real-time blood oxygen saturation value is lower than the raised hypoxia threshold value.

[0006] Further, the specific process of synchronously collecting the electrocardiosignal, the photoplethysmographic signal and the three-axis acceleration signal of a child, performing vector synthesis on the three-axis acceleration signal and calculating the root mean square envelope to generate a motion intensity envelope signal is as follows: taking the sampling time of the photoplethysmographic signal as the reference time stamp, linearly interpolating and resampling the electrocardiosignal and the three-axis acceleration signal, extracting the instantaneous components of the resampled three-axis acceleration signal in the X-axis, Y-axis and Z-axis, calculating the arithmetic square root of the square sum of the three components to obtain the instantaneous combined acceleration vector; performing root mean square operation on the instantaneous combined acceleration vector within a set sliding time window, extracting the low-frequency trend term and filtering out the high-frequency dithering noise to generate a motion intensity envelope signal reflecting the overall kinetic energy change of the child.

[0007] Further, the specific process of extracting the peak feature points of the electrocardio signal and constructing the R-wave peak interval sequence is as follows: sequentially performing band-pass filtering, first-order differential processing and squaring operation on the synchronously collected electrocardio signal to generate a preprocessed signal with enhanced energy; setting an adaptive floating threshold dynamically changing with the amplitude of the preprocessed signal, comparing the amplitudes of the preprocessed signal and the adaptive floating threshold, and locating the maximum value points exceeding the adaptive floating threshold as R-wave candidate points; performing timing logic verification on the R-wave candidate points according to a preset refractory period time window, retaining the effective R-wave peak time, calculating the time difference between adjacent effective R-wave peak times, and constructing the R-wave peak interval sequence in time sequence.

[0008] Further, the specific process of projecting denoising processing on the photoplethysmogram signal based on the recursive least square adaptive filtering algorithm by calling the motion intensity envelope signal as a non-stationary noise reference source is as follows: defining the motion intensity envelope signal as the reference input vector of the filter and the photoplethysmogram signal as the expected response signal, initializing the weight coefficient vector of the filter and the inverse matrix of the autocorrelation matrix, and calculating the Kalman gain vector through the reference input vector at the current time; calculating the prior estimation error according to the difference between the photoplethysmogram signal and the filter output value, and iteratively updating the weight coefficient vector combined with the Kalman gain vector to generate the converged filter weight coefficient.

[0009] Further, the specific process of projecting denoising processing on the photoplethysmogram signal based on the recursive least square adaptive filtering algorithm by calling the motion intensity envelope signal as a non-stationary noise reference source is as follows: defining the motion intensity envelope signal as the reference input vector of the filter and the photoplethysmogram signal as the expected response signal, initializing the weight coefficient vector of the filter and the inverse matrix of the autocorrelation matrix, and calculating the Kalman gain vector through the reference input vector at the current time; calculating the prior estimation error according to the difference between the photoplethysmogram signal and the filter output value, and iteratively updating the weight coefficient vector combined with the Kalman gain vector to generate the converged filter weight coefficient.

[0010] Further, the R-wave peak interval sequence is interpolated to match the sampling rate of the motion intensity envelope signal, and the specific process of calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R-wave peak interval sequence within the preset physiological response lag window is as follows: the discrete R-wave peak interval sequence is reconstructed continuously by a cubic spline interpolation algorithm, and resampled according to the time axis of the motion intensity envelope signal to generate a time-aligned heart rate trend sequence; a physiological response lag window containing the heart rate trend sequence and the motion intensity envelope signal is constructed by intercepting a preset length of double-channel data segment forward from the current time as the end point; the motion intensity envelope signal is kept stationary within the window, and the heart rate trend sequence is time-domain sliding translation, the normalized cross-correlation function value under different time lags is calculated, and the global maximum value of the function value is searched and extracted as the maximum cross-correlation coefficient.

[0011] Further, the maximum cross-correlation coefficient is compared with the preset physiological coupling threshold, and when the R-wave peak interval sequence shows a heart rate rising trend and the maximum cross-correlation coefficient is lower than the physiological coupling threshold, the specific process of generating an independent heart rate abnormality index is as follows: the interpolated R-wave peak interval sequence is subjected to a first-order differential operation to obtain an instantaneous change rate, and the cumulative sum of the change rate within a preset window is calculated, and when the cumulative sum is positive and exceeds a preset trend judgment threshold, the heart rate rising trend is confirmed; a logical AND operation is performed, and only when the heart rate rising trend is confirmed and the absolute value of the calculated maximum cross-correlation coefficient is less than the preset physiological coupling threshold, the abnormality quantification logic is activated; the absolute value of the difference between the maximum cross-correlation coefficient and the physiological coupling threshold is calculated, which is mapped to the preset risk classification scale as a weight factor, and the quantified value is output as an independent heart rate abnormality index.

[0012] Further, the reconstructed blood oxygen waveform is analyzed to calculate the real-time blood oxygen saturation value, and the logic process of taking the independent heart rate abnormality index as the adjustment factor of the blood oxygen alarm threshold is as follows: the peak and trough values of the reconstructed blood oxygen waveform within a cardiac cycle are extracted, the alternating current pulsation component and the direct current baseline component are separated, the light intensity ratio under double wavelengths is calculated and substituted into the calibration curve to obtain the real-time blood oxygen saturation value; a linear threshold compensation function is established, the independent heart rate abnormality index is input into the function as an independent variable, the corresponding threshold floating compensation amount is calculated, and the threshold floating compensation amount is added to the system default preset hypoxia threshold to construct a dynamically changing current execution alarm threshold.

[0013] Further, when the independent heart rate abnormality index is generated, the preset hypoxia threshold is raised to tighten the alarm boundary, and if the real-time blood oxygen saturation value is lower than the raised hypoxia threshold, the specific process of triggering the alarm instruction is as follows: the generation state of the independent heart rate abnormality index is monitored in real time, once a non-zero independent heart rate abnormality index is received, the threshold dynamic migration mechanism is immediately started; the preset hypoxia threshold is positively offset corrected by the threshold floating compensation amount, the raised high-sensitivity hypoxia threshold is obtained, the real-time blood oxygen saturation value is compared with the high-sensitivity hypoxia threshold, and if the real-time blood oxygen saturation value is lower than the high-sensitivity hypoxia threshold, the hierarchical alarm instruction is immediately generated and the sound and light early warning output is executed.

[0014] An integrated pediatric vital sign intelligent monitoring and early warning system comprises the following modules: a synchronous acquisition preprocessing module for synchronously acquiring electrocardiosignal, photoplethysmogram signal and three-axis acceleration signal of a sick child, performing vector synthesis on the three-axis acceleration signal and calculating the root mean square envelope to generate a motion intensity envelope signal, and extracting the peak feature points of the electrocardiosignal to construct an R wave peak interval sequence; a projection denoising reconstruction module for calling the motion intensity envelope signal as a non-stationary noise reference source, performing projection denoising processing on the photoplethysmogram signal based on a recursive least square adaptive filtering algorithm, performing orthogonal projection elimination on the motion artifacts in the photoplethysmogram signal through the fluctuation characteristics of the motion intensity envelope signal, and outputting a reconstructed blood oxygen waveform; a physiological coupling analysis module for performing interpolation processing on the R wave peak interval sequence, matching the sampling rate of the motion intensity envelope signal, calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R wave peak interval sequence within a preset physiological reaction lag window, comparing the maximum cross-correlation coefficient with a preset physiological coupling threshold, and generating an independent heart rate abnormality index when the R wave peak interval sequence shows a heart rate rising trend and the maximum cross-correlation coefficient is lower than the physiological coupling threshold; and a dynamic early warning decision module for analyzing the reconstructed blood oxygen waveform to calculate a real-time blood oxygen saturation value, taking the independent heart rate abnormality index as an adjustment factor of the blood oxygen alarm threshold, raising the preset hypoxia threshold to tighten the alarm boundary when the independent heart rate abnormality index is generated, and triggering an alarm instruction if the real-time blood oxygen saturation value is lower than the raised hypoxia threshold.

[0015] The present application has the following beneficial effects:

[0016] (1) An integrated pediatric vital sign intelligent monitoring and early warning method, by introducing a motion intensity envelope signal as a reference source for adaptive filtering, high-fidelity reconstruction of physiological signals in dynamic scenarios is achieved. The present application aims at the problem of photoplethysmogram signal distortion caused by crying and limb restlessness of children in pediatric clinical practice. The motion intensity envelope signal calculated from the synchronous three-axis acceleration can accurately quantify the body movement energy characteristics of children. In the signal processing link, the recursive least squares adaptive filtering algorithm is used, and the motion intensity envelope signal is used as a noise reference for orthogonal projection denoising of the photoplethysmogram signal. This method can accurately strip the artifact components highly related to body movement characteristics from the mixed signal, and retain the pure physiological blood flow component, so as to ensure that even in the case of severe movement of children, the reconstructed blood oxygen waveform output by the system is still baseline smooth and the characteristics are clear, avoiding false oxygen value drop caused by motion interference.

[0017] (2) An integrated pediatric vital sign intelligent monitoring and early warning system, by calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the R-wave peak interval sequence within the physiological response lag window, the system can accurately identify whether the heart rate increase is caused by motion. When determining that it is an independent heart rate abnormality unrelated to motion, the system does not directly alarm, but generates an independent heart rate abnormality index and raises the preset hypoxia threshold based on it, and actively tightens the alarm boundary. This mechanism, on the one hand, maintains a high threshold during normal activities of children to suppress false alarms, and on the other hand, automatically increases the sensitivity to hypoxia when potential pathological risks are detected, ensuring that the alarm can be triggered in advance when the oxygen saturation has not yet dropped to the traditional critical value but an abnormal trend has appeared, and valuable time is gained for clinical intervention.

[0018] Of course, implementing any product of the present application does not necessarily require all the advantages described above to be achieved at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flowchart of an integrated pediatric vital sign intelligent monitoring and early warning method of the present application.

[0020] Figure 2 A flowchart of an integrated pediatric vital sign intelligent monitoring and early warning system of the present application. DETAILED DESCRIPTION

[0021] The embodiments of the present application solve the problems of poor anti-interference ability of existing pediatric monitoring devices in dynamic environments, high false alarm rate and early warning lag caused by lack of physiological-behavioral correlation analysis through an integrated pediatric vital sign intelligent monitoring and early warning method and system.

[0022] The scheme in the embodiments of the present application has the following general idea: first, time-aligned motion and physiological data streams are constructed through synchronous acquisition of multiple sensors; then, motion envelope is used as a reference source, and adaptive filtering algorithm is used to remove motion artifacts in the blood oxygen signal to obtain pure physiological waveform; subsequently, cross-correlation analysis technology is used to determine whether the heart rate change is caused by motion, so as to accurately identify non-motion-related pathological abnormalities; finally, the abnormality index is used to dynamically adjust the blood oxygen alarm threshold, realize the linkage warning between physiological parameters, and ensure that the motion interference is excluded while the potential pathological risk is intervened early.

[0023] Please refer to Figure 1 The embodiments of the present application provide a technical scheme: an integrated pediatric vital sign intelligent monitoring and early warning method, comprising the following steps: S1. synchronously acquiring the electrocardiosignal, the photoplethysmogram signal and the three-axis acceleration signal of a child, performing vector synthesis on the three-axis acceleration signal and calculating the root mean square envelope to generate a motion intensity envelope signal, and extracting the peak feature points of the electrocardiosignal to construct an R-wave peak interval sequence; S2. calling the motion intensity envelope signal as a non-stationary noise reference source, projecting and denoising the photoplethysmogram signal based on the recursive least squares adaptive filtering algorithm, removing the motion artifacts in the photoplethysmogram signal through orthogonal projection of the fluctuation characteristics of the motion intensity envelope signal, and outputting the reconstructed blood oxygen waveform; S3. interpolating the R-wave peak interval sequence to match the sampling rate of the motion intensity envelope signal, calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R-wave peak interval sequence within a preset physiological reaction lag window, comparing the maximum cross-correlation coefficient with a preset physiological coupling threshold, and generating an independent heart rate abnormality index when the R-wave peak interval sequence shows a heart rate rising trend and the maximum cross-correlation coefficient is lower than the physiological coupling threshold; S4. analyzing the reconstructed blood oxygen waveform to calculate the real-time blood oxygen saturation value, using the independent heart rate abnormality index as an adjustment factor of the blood oxygen alarm threshold, raising the preset hypoxia threshold to tighten the alarm boundary when the independent heart rate abnormality index is generated, and triggering an alarm instruction if the real-time blood oxygen saturation value is lower than the raised hypoxia threshold.

[0024] In the present embodiment, step S1 is mainly to perform the synchronous acquisition and feature preprocessing of multi-source physiological physical signals. The step utilizes the sensor group in the flexible patch to collect data under the same clock reference, so as to eliminate the time deviation between different signals. Among them, the vector synthesis of three-axis acceleration signal and the calculation of root mean square envelope refer to the square sum and square root operation of the acceleration components in X-axis, Y-axis and Z-axis directions, to obtain the resultant acceleration vector, and to calculate the effective value in the preset time window, so as to generate the motion intensity envelope signal which can represent the overall body kinetic energy of the child. At the same time, the peak feature points of the electrocardiogram signal are extracted and the R wave peak interval sequence is constructed, which refers to identifying the time of the R wave with the highest amplitude in the electrocardiogram waveform by algorithm, and calculating the time difference between the adjacent two R waves to form a discrete time sequence reflecting the rhythm change of heart beat, providing the basis data for subsequent analysis of the correlation between heart rate change and limb movement. The core of step S2 is to use adaptive signal processing technology to eliminate measurement interference in dynamic environment. The motion intensity envelope signal generated in step S1 is called as a non-stationary noise reference source, that is, it is regarded as a known interference mode, and is input into the recursive least squares adaptive filter. The recursive least squares algorithm is an algorithm that can adjust the filter weight coefficient in real time according to the statistical characteristics of the input signal. It uses the fluctuation characteristics of the motion intensity envelope signal to simulate the artifact component in the photoplethysmogram signal which is highly related to the body movement characteristics, and performs orthogonal projection removal, that is, subtracts the fitted artifact component from the original mixed signal. This process can effectively remove the baseline drift and waveform distortion caused by the limb shaking or patting of the child, and output the reconstructed blood oxygen waveform which retains the true blood flow characteristics, ensuring the accuracy of subsequent blood oxygen saturation calculation. Step S3 aims to distinguish physiological compensation from pathological abnormalities through multi-modal coupling analysis. First, the discrete R wave peak interval sequence is interpolated to convert it into a continuous signal with the same sampling rate as the motion intensity envelope signal. The preset physiological reaction lag window refers to considering the physiological delay of several seconds of human heart rate change relative to limb movement. The system calculates the maximum cross-correlation coefficient of the similarity of the two signal patterns within this specific time delay range. The cross-correlation coefficient is a statistical indicator to measure the linear correlation degree of two time series. When it is monitored that the R wave peak interval sequence shows an upward trend in heart rate, but the maximum cross-correlation coefficient calculated is lower than the preset physiological coupling threshold, it indicates that the current heart rate acceleration is not induced by limb movement, which is an independent abnormality unrelated to movement. The system generates an independent heart rate abnormality index accordingly, which is used to quantify the degree of abnormality of the heart autonomic regulation function. Step S4 performs dynamic threshold regulation and final warning decision based on risk superposition. The system analyzes the reconstructed blood oxygen waveform output by step S2, extracts the ratio of the alternating component to the direct current component of the waveform to calculate the real-time blood oxygen saturation value.Meanwhile, an independent heart rate abnormality index is introduced as an adjusting factor for the blood oxygen alarm threshold, and an alarm boundary tightening strategy is implemented: when the system generates an independent heart rate abnormality index, it means that the child is in a high cardiac load or a potential pathological state, at which time the system automatically raises the preset low oxygen threshold, for example, the alarm trigger line is raised from the regular standard. If the measured real-time blood oxygen saturation value is lower than the raised low oxygen threshold at this time, even if the value may be within the normal range, the system still determines that it is a high-risk state and triggers the alarm instruction. This mechanism realizes earlier intervention in hypoxic state when the pathological risk increases, effectively solving the problem of traditional fixed threshold alarm lag in pediatric patients with rapid changes in disease.

[0025] Specifically, the ECG signal, the photoplethysmogram signal and the three-axis acceleration signal of the patient are synchronously collected, the three-axis acceleration signal is vector-synthesized and the root mean square envelope is calculated, and the specific process of generating the motion intensity envelope signal is as follows: taking the sampling time point of the photoplethysmogram signal as the reference time stamp, the ECG signal and the three-axis acceleration signal are linearly interpolated and resampled, the instantaneous components of the resampled three-axis acceleration signal in the X-axis, Y-axis and Z-axis are extracted, and the arithmetic square root of the square sum of the three components is calculated to obtain the instantaneous combined acceleration vector; the instantaneous combined acceleration vector is subjected to root mean square operation within a set sliding time window, the low-frequency trend term is extracted and the high-frequency dithering noise is filtered out, and the motion intensity envelope signal reflecting the overall body kinetic energy change of the patient is generated.

[0026] In the embodiment, first, the system performs a time axis alignment operation of multi-sensor data. Due to the difference in the underlying sampling frequency of different sensors, direct calculation will cause phase misalignment, therefore, taking the sampling time point of the photoplethysmogram signal as the reference time stamp, the ECG signal and the three-axis acceleration signal are resampled by using a linear interpolation algorithm to ensure that all data points strictly correspond in the time dimension. Next, for the resampled three-axis acceleration signal, the system extracts its three components in the spatial rectangular coordinate system, eliminates the influence of the sensor wearing direction on the measurement result by vector synthesis, and obtains the instantaneous combined acceleration vector which can represent the instantaneous combined external force. In order to extract the smooth features reflecting the macroscopic body kinetic energy of the human body from the original high-frequency dithering acceleration data, the system sets a sliding time window, performs root mean square envelope operation within the window, and the calculation process is based on the following discrete mathematical model: ; in the formula, : the motion intensity envelope signal value output at the kth moment; : the sample point length of the sliding time window, which is usually valued at 0.5 to 1 times the sampling rate to cover a complete human motion period; : the summation index variable within the sliding window : the component of the resampled acceleration signal in the X-axis direction at the k-jth moment; : the component of the resampled acceleration signal in the Y-axis direction at the k-j moment; : the component of the resampled acceleration signal in the Z-axis direction at the k-j moment. Through the above operation, the high-frequency mechanical vibration noise is effectively filtered out, and the output signal mainly retains the low-frequency body movement trend item, providing an accurate noise reference model for subsequent adaptive filtering.

[0027] Specifically, the peak feature points of the electrocardiogram signal are extracted, and the specific process of constructing the R-wave peak interval sequence is as follows: the synchronously collected electrocardiogram signal is sequentially subjected to band-pass filtering, first-order differential processing and square operation to generate a preprocessed signal with enhanced energy; an adaptive floating threshold that dynamically changes with the amplitude of the preprocessed signal is set, the preprocessed signal is compared with the adaptive floating threshold in amplitude, and the maximum value point that exceeds the adaptive floating threshold is located as the R-wave candidate point; the R-wave candidate point is subjected to timing logic verification according to a pre-set refractory period time window, the effective R-wave peak moment is retained, the time difference value of adjacent effective R-wave peak moments is calculated, and the R-wave peak interval sequence is constructed in time sequence.

[0028] In the embodiment, the process of constructing the R-wave peak interval sequence is described in detail. This step aims to accurately locate the moment of cardiac pulsation from the original electrocardiogram signal containing electromyographic interference and baseline drift. First, the synchronously collected electrocardiogram signal is sequentially subjected to band-pass filtering to isolate the QRS complex frequency band, first-order differential processing to highlight the steep slope feature of the R-wave, and nonlinear square operation to enhance high-frequency energy, generating an energy-enhanced preprocessed signal with significantly improved signal-to-noise ratio. Subsequently, to adapt to individual differences in the amplitude of electrocardiogram signals of different children, the system discards the fixed threshold and adopts an adaptive floating threshold strategy that adjusts in real time with the signal strength. This strategy dynamically tracks the signal background noise and signal peak level through a recursive update algorithm, and the specific adaptive floating threshold calculation is as follows: ; in the formula, : adaptive floating threshold at the n-th update; : historical threshold at the n-1-th update; : local maximum amplitude of the preprocessed signal detected at present; The threshold updating forgetting factor is used to control the following speed of the threshold to the signal change, and the determination method of the factor is: according to the signal-to-noise ratio level of the preprocessed signal, an empirical value is selected in the interval of 0.1 to 0.3, and the higher the signal-to-noise ratio is, the larger the value is. The system compares the preprocessed signal with the adaptive floating threshold in real time, locates the maximum value point exceeding the threshold as the R wave candidate point. Finally, the refractory period time window logic is introduced, the candidate point is checked for physiological reasonableness, the pseudo-wave peak with a spacing less than the physiological limit of the human body (such as 200 milliseconds) is removed, the effective R wave peak time is confirmed, the time difference of adjacent time is calculated, and the accurate R wave peak interval sequence is constructed in sequence, thereby providing a reliable basis for subsequent heart rate variability analysis and physiological-physical coupling judgment.

[0029] Specifically, the specific process of projecting denoising processing of the photoplethysmogram signal based on the recursive least square adaptive filtering algorithm is as follows: the motion intensity envelope signal is defined as the reference input vector of the filter, the photoplethysmogram signal is defined as the expected response signal, the weighting coefficient vector of the filter and the inverse matrix of the autocorrelation matrix are initialized, and the Kalman gain vector is calculated through the reference input vector at the current time; the prior estimation error is calculated according to the difference between the photoplethysmogram signal and the filter output value, the weighting coefficient vector is iteratively updated combined with the Kalman gain vector, and the converged filter weight coefficient is generated.

[0030] In the embodiment, first, the system constructs an input signal vector, and the motion intensity envelope signal generated in the foregoing step is used as the reference input vector because the signal contains only body motion characteristics; meanwhile, the photoplethysmogram signal containing noise is used as the expected response signal because the signal superimposes the physiological signal that the system tries to retain and the motion interference that the system tries to eliminate. The initialization process includes setting the weighting coefficient vector of the filter to zero and initializing the inverse matrix of the autocorrelation matrix as a large-value diagonal matrix. Subsequently, the system enters an iterative loop, first calculates the Kalman gain vector using the input vector at the current time, and the vector determines the influence weight of the new input data on the filter parameter update. In order to adapt to the non-stationary signal (i.e., the motion state changes over time), an exponential forgetting factor is introduced, and the determination method of the factor is: according to the signal sampling frequency and the speed of motion change, a value is usually selected between 0.98 and 0.999, and the closer the value is to 1, the longer the memory length is. Based on this, the update calculation model of the Kalman gain vector and the filter weight coefficient is as follows: ; wherein, : the Kalman gain vector calculated at the mth iteration time; : the inverse matrix of the input signal autocorrelation matrix at the m-1th iteration time; ​: the reference input vector composed of the motion intensity envelope signal at the mth moment; : exponential forgetting factor; : the updated filter weight coefficient vector at the mth moment, which is the converged filter weight coefficient and represents the transfer function of the motion interference at the current moment; : the original photoplethysmogram signal value collected at the mth moment; : the priori estimation value of the interference at the current moment based on the weight coefficient at the previous moment; the term in the square brackets is the priori estimation error. Through the above iteration, the filter can quickly converge and lock the dynamic characteristics of the motion interference.

[0031] Specifically, the specific process of outputting the reconstructed blood oxygen waveform by orthogonal projection removing the motion artifact in the photoplethysmogram signal through the fluctuation characteristics of the motion intensity envelope signal is as follows: performing convolution operation on the converged filter weight coefficient and the current motion intensity envelope signal, estimating the fitting motion artifact component in the photoplethysmogram signal highly related to the body motion characteristics; performing orthogonal projection subtraction operation, subtracting the fitting motion artifact component from the original photoplethysmogram signal point by point, retaining the physiological blood flow component orthogonal to the motion intensity envelope signal, performing smoothing filtering processing on the residual signal after the subtraction operation, repairing the waveform burr caused by phase lag, and outputting the reconstructed blood oxygen waveform with stable waveform baseline and retained re-pulse notch characteristics.

[0032] In the embodiment, after obtaining the converged filter weight coefficient, the system has mastered the current motion artifact generation mechanism. At this time, the system performs orthogonal projection removing operation, which is actually subtracting the component parallel to the motion vector space from the original signal space, thereby retaining the physiological component orthogonal thereto. Specifically, first, convolution operation is performed on the converged filter weight coefficient and the current motion intensity envelope signal, which is actually generating an artificial fitting motion artifact component using the trained model, which is highly similar in morphology to the real noise superimposed in the blood oxygen signal. Subsequently, subtraction operation is performed to subtract the fitting component from the original photoplethysmogram signal point by point. In order to eliminate the slight phase lag caused by digital filtering and the edge burr generated by subtraction operation, the system further performs smoothing processing on the residual signal, and finally outputs the reconstructed blood oxygen waveform. The calculation of the reconstruction and repair process is as follows: ; wherein, : the final reconstructed blood oxygen waveform value output at the mth moment; : the original noisy photoplethysmogram signal at the mth moment; : the order of the filter, representing the time span of the motion interference considered by the system; : the rth component in the filter weight coefficient vector at the mth moment; : reference the motion intensity envelope signal value corresponding to the m-r moment in the input vector; the summation term is the estimated fitting motion artifact component; : smoothing filter operator, Savitzky-Golay filter algorithm is adopted here, which can smooth the noise while keeping the width and height characteristics of the waveform unchanged, ensuring that the reconstructed waveform retains the double-notch characteristics reflecting the vessel elasticity, providing a high-fidelity waveform basis for subsequent blood oxygen saturation extraction.

[0033] Specifically, the R-wave peak interval sequence is interpolated to match the sampling rate of the motion intensity envelope signal, and the specific process of calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R-wave peak interval sequence within the preset physiological response lag window is as follows: the discrete R-wave peak interval sequence is reconstructed continuously by a cubic spline interpolation algorithm, and resampled according to the time axis of the motion intensity envelope signal to generate a time-aligned heart rate trend sequence; a physiological response lag window containing the heart rate trend sequence and the motion intensity envelope signal is constructed by intercepting a segment of double-channel data of a preset duration forward from the current time as the endpoint; the motion intensity envelope signal is kept stationary within the window, and the heart rate trend sequence is time-domain sliding translation, the normalized cross-correlation function value under different time lags is calculated, and the global maximum value of the function value is searched and extracted as the maximum cross-correlation coefficient.

[0034] In this embodiment, first, the system performs sequence continuous reconstruction, since the R-wave peak interval sequence is a non-uniform discrete signal based on heartbeat events, the system uses a cubic spline interpolation algorithm to fit it. The cubic spline interpolation constructs a cubic polynomial between each two data points, ensuring that the first and second derivatives of the reconstructed curve are continuous at the connection points, thereby generating a smooth and time-aligned heart rate trend sequence with a sampling rate strictly consistent with the motion intensity envelope signal. Subsequently, the system constructs a physiological response lag window, considering that the autonomic nervous system regulates heart rate relative to limb movement usually exists a delay of several seconds to more than ten seconds, the system intercepts a segment of double-channel data before the current time. Within this window, in order to quantify the similarity of the two forms and find the best time matching point, the system keeps the motion data stationary and performs time-domain sliding on the heart rate data to calculate the normalized cross-correlation function. ; in the formula, : the normalized cross-correlation coefficient value when the time lag is , the value range is between negative 1 and 1, the closer the value is to 1, the stronger the positive correlation is; the time sample index within the physiological response lag window; the total number of data samples contained in the window; The time sliding offset of the heart rate trend sequence relative to the exercise intensity envelope signal, the search range is usually set to cover the normal physiological response delay time of the human body; : the heart rate trend sequence value at the offset time; : the arithmetic mean of the heart rate trend sequence within the window; : the value of the exercise intensity envelope signal within the window; : the arithmetic mean of the exercise intensity envelope signal within the window. The system traverses all possible values, searches for the global maximum of the cross-correlation coefficient, and extracts it as the maximum cross-correlation coefficient. This step plays a key role in discrimination: if the coefficient is high, it means that the change in heart rate occurs after exercise, which is a normal physiological compensation; if the coefficient is very low, it means that the two do not match in time sequence pattern.

[0035] Specifically, when the R-wave peak interval sequence shows a rising trend in heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold, the specific process of generating the independent heart rate abnormality index is as follows: perform a first-order differential operation on the interpolated R-wave peak interval sequence to obtain the instantaneous change rate, calculate the cumulative sum of the change rate within a preset window, and when the cumulative sum is positive and exceeds a preset trend judgment threshold, confirm the rising trend of heart rate; perform a logical AND operation, and only when it is confirmed that there is a rising trend in heart rate and the absolute value of the calculated maximum cross-correlation coefficient is less than the preset physiological coupling threshold, activate the abnormality quantification logic; calculate the absolute value of the difference between the maximum cross-correlation coefficient and the physiological coupling threshold, map it to the preset risk classification scale as a weight factor, and output the quantified value as the independent heart rate abnormality index.

[0036] In this embodiment, this step aims to convert mathematical correlation analysis into a clinically useful risk quantification indicator. First, the system needs to confirm whether the heart rate is in a rising state, and the first-order derivative of the interpolated R-wave peak interval sequence is integrated. If it is confirmed that the heart rate is continuously rising, the abnormality discrimination logic is entered. The system performs a logical AND operation: only when the heart rate is rising and the maximum cross-correlation coefficient calculated above is lower than the preset physiological coupling threshold, it is determined to be abnormal. The determination method of the physiological coupling threshold is: based on the heart rate-exercise data of a large number of healthy children in exercise state, an empirical value between 0.6 and 0.8 is usually set. Once the abnormality quantification logic is activated, the system will calculate the deviation of the correlation coefficient from the threshold, and map it to the risk index. The calculation of this quantification process is as follows: ; in the formula, the independent heart rate abnormality index output at the pth determination time, the higher the index, the greater the pathological risk; : Scaling factor of the risk classification scale, used to normalize the calculation result to a numerical interval that is convenient for the system to process; : Preset physiological coupling threshold, i.e. the dividing line for determining whether the heart rate is related to exercise; : Absolute value of the maximum cross-correlation coefficient obtained by the current window calculation, which takes the absolute value in order to take into account the strong coupling of positive and negative correlations; : Slope adjustment parameter of the Sigmoid function, used to control the sensitivity of the index to the heart rate rising rate; : Slope cumulative value of the current detected heart rate rising trend; : Minimum heart rate change slope limit for triggering an alarm. This formula not only considers the degree of irrelevance between heart rate and exercise (the first factor), but also introduces the heart rate rising rate as a weighting (the second Sigmoid function factor), ensuring that only when the heart rate is severely irrelevant to exercise and rapidly rising, a high value of the independent heart rate anomaly index is output, thereby accurately capturing sudden heart rate anomalies caused by non-exercise factors.

[0037] Specifically, the logic process of analyzing and reconstructing the blood oxygen waveform to calculate the real-time blood oxygen saturation value and using the independent heart rate anomaly index as the adjustment factor of the blood oxygen alarm threshold is as follows: the peak value and the trough value of the reconstructed blood oxygen waveform within a cardiac cycle are extracted, the alternating current pulsatile component and the direct current baseline component are separated, the light intensity ratio under the dual wavelength is calculated and substituted into the calibration curve, and the real-time blood oxygen saturation value is obtained; a linear threshold compensation function is established, the independent heart rate anomaly index is input into the function as the independent variable, the corresponding threshold floating compensation amount is calculated, and the threshold floating compensation amount is added to the system default preset hypoxia threshold to build a dynamically changing current execution alarm threshold.

[0038] In this embodiment, first, the system scans the reconstructed blood oxygen waveform within a complete cardiac cycle, and extracts the maximum value and the minimum value of the waveform. The alternating current pulsatile component reflecting the arterial blood flow pulsatile component is obtained by subtracting the minimum value from the maximum value, and the direct current baseline component reflecting the venous blood and tissue background absorption is obtained by using the minimum value. The system performs the above operation on the red light channel and the infrared light channel respectively, calculates the absorbance ratio under the dual wavelength, and calculates the real-time blood oxygen saturation value. The photoelectric conversion and blood oxygen calculation process is based on the following mathematical model: ; ; wherein, : Relative light intensity absorption ratio under the dual wavelength, which eliminates the influence of light intensity fluctuations and tissue thickness differences; : Amplitude of the alternating current pulsatile component of the red light channel; : Amplitude of the direct current baseline component of the red light channel; : Amplitude of the alternating current pulsatile component of the infrared light channel; : Amplitude of the direct current baseline component of the infrared light channel; : the real-time blood oxygen saturation value calculated at the qth moment; : the intercept coefficient of the empirical calibration curve; : the slope coefficient of the empirical calibration curve. The determination method of the two calibration coefficients is: in clinical experiments, a large number of pediatric sample blood gas analysis data and photoelectric signal ratio are collected to obtain a least squares fitting, which is usually stored in the firmware lookup table of the system. Subsequently, in order to realize the linkage warning between physiological parameters, the system establishes a linear threshold compensation function. The purpose of the function is to convert the abstract independent heart rate abnormality index into a specific blood oxygen threshold offset. The logical basis is that when the child has non-exercise related heart rate abnormalities, the body's oxygen consumption increases and the compensatory reserve decreases, so a more stringent blood oxygen standard is needed. The model for calculating the corresponding threshold floating compensation is as follows: ; wherein, : the threshold floating compensation value calculated at the qth moment, which represents how many percentage points the alarm line needs to be raised; : the independent heart rate abnormality index generated in the previous step; : the threshold adjustment gain coefficient, used to control the sensitivity of the system to heart rate abnormalities. The larger the coefficient, the higher the alarm threshold will be raised for the same abnormality index; : the maximum compensation upper limit set, for example, limiting the threshold to float by 5 percentage points at most, to prevent the threshold from exceeding 100 due to excessive compensation.

[0039] Specifically, when the independent heart rate abnormality index is generated, the preset hypoxia threshold is raised to tighten the alarm boundary. If the real-time blood oxygen saturation value is lower than the raised hypoxia threshold, the specific process of triggering the alarm instruction is as follows: the generation state of the independent heart rate abnormality index is monitored in real time. Once a non-zero independent heart rate abnormality index is received, the threshold dynamic migration mechanism is immediately started. The preset hypoxia threshold is positively offset corrected by the threshold floating compensation value to obtain a high-sensitivity hypoxia threshold. The real-time blood oxygen saturation value is compared with the high-sensitivity hypoxia threshold. If the real-time blood oxygen saturation value is lower than the high-sensitivity hypoxia threshold, a graded alarm instruction is immediately generated and the sound and light warning output is executed.

[0040] In this embodiment, the system monitors the generation state of the independent heart rate abnormality index in real time. Once a non-zero independent heart rate abnormality index is received, it indicates that the child currently has unexplained heart rate elevation. The system immediately starts the threshold dynamic migration mechanism. The mechanism uses the aforementioned calculated compensation value to positively offset correct the system's default preset hypoxia threshold, thereby constructing a high-sensitivity hypoxia threshold that dynamically changes with the risk of the disease. The dynamic comparison and decision-making process is based on the following logic: ; ; wherein, The high-sensitivity hypoxia threshold actually executed at time q; The system's preset default low oxygen threshold, usually set to 90 or 92, represents the alarm threshold under normal circumstances. The current threshold upward compensation amount; The final generated alarm trigger command status: 1 represents triggering an alarm, and 0 represents silence. The currently measured real-time blood oxygen saturation value. Through the above calculations, the system implements a risk-overlay warning logic: if the independent heart rate abnormality index is 0 (i.e., normal or purely exercise-compensated), then... Equal to the default value The system operates according to standard procedures; however, if the independent heart rate abnormality index is high, it may lead to… The threshold is raised from the default 90 to 94. If the real-time blood oxygen saturation is measured at 93, although the value is higher than the standard baseline of 90, it is lower than the current high-sensitivity threshold of 94. Therefore, the system determines that the child is in a state of relative hypoxia, immediately generates a graded alarm command, and executes an audible and visual warning. This mechanism ensures that when the pathological risk is high, the system can protect the child's safety with a more stringent standard, effectively preventing missed diagnoses during the compensatory period.

[0041] Please see Figure 2 An integrated intelligent monitoring and early warning system for pediatric vital signs includes the following modules: a synchronous acquisition and preprocessing module, used to synchronously acquire the child's electrocardiogram (ECG), photoplethysmography (PPG) wave signal, and triaxial acceleration signal; to perform vector synthesis of the triaxial acceleration signal and calculate the root mean square envelope to generate a motion intensity envelope signal; and to extract peak feature points of the ECG signal to construct an R-wave peak interval sequence; a projection denoising and reconstruction module, used to call the motion intensity envelope signal as a non-stationary noise reference source; to perform projection denoising processing on the PPG wave signal based on a recursive least squares adaptive filtering algorithm; and to perform orthogonal projection removal of motion artifacts in the PPG wave signal based on the fluctuation characteristics of the motion intensity envelope signal, outputting a reconstructed blood oxygen waveform; and a physiological coupling analysis module, using... The R-wave peak interval sequence is interpolated to match the sampling rate of the exercise intensity envelope signal. Within a preset physiological response lag window, the maximum cross-correlation coefficient between the exercise intensity envelope signal and the interpolated R-wave peak interval sequence is calculated. The maximum cross-correlation coefficient is compared with a preset physiological coupling threshold. When the R-wave peak interval sequence shows an upward trend in heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold, an independent heart rate abnormality index is generated. The dynamic early warning decision module is used to analyze and reconstruct the blood oxygen waveform to calculate the real-time blood oxygen saturation value. The independent heart rate abnormality index is used as an adjustment factor for the blood oxygen alarm threshold. When the independent heart rate abnormality index is generated, the preset hypoxia threshold is increased to tighten the alarm boundary. If the real-time blood oxygen saturation value is lower than the increased hypoxia threshold, an alarm command is triggered.

[0042] In this embodiment, the synchronous acquisition preprocessing module serves as the data perception and primary processing unit of the system, responsible for establishing a unified space-time reference. Its core role is to control the multi-channel sensors at the front end to operate under the drive of the same clock source, ensuring that the physiological electrical signals and physical motion signals obtained are strictly aligned on the time axis. At the same time, this module undertakes the task of data dimension reduction and feature extraction, on the one hand converting complex acceleration data in three-dimensional space into single-dimensional motion energy representation, and on the other hand discretizing and extracting the timing features of heartbeats from continuous electrocardiogram waveforms, providing standardized input data streams for subsequent modules, solving the problem of mismatched feature dimensions and different sampling rates of original multi-modal signals. The projection denoising reconstruction module serves as the signal purification unit of the system, responsible for restoring the true form of the photoelectric signal in a dynamic environment. This module uses the motion energy representation generated by the previous module as a known interference reference to drive an adaptive filtering algorithm for real-time processing of the photoplethysmogram signal. It accurately identifies and strips the highly correlated artifact components in the signal through mathematical projection, eliminating waveform distortion and baseline drift caused by the child's crying or patting. The output of this module provides a high-fidelity waveform basis for subsequent blood oxygen calculation, ensuring that even under strong motion interference, the system can still obtain pure signals reflecting the true blood perfusion. The physiological coupling analysis module serves as the logical decision core of the system, responsible for evaluating the causal relationship between heart rate changes and limb motion. This module first solves the frequency matching problem between different signals, and then quantitatively analyzes the similarity of heart rate trends and motion intensity in the time sequence form under the consideration of human physiological response delay. Its key role lies in distinguishing compensation from pathology: when it identifies a heart rate increase without a corresponding motion trigger, the module can determine the non-motion-related risk caused by autonomic nervous regulation abnormalities or increased cardiac load and quantify this qualitative risk into a specific abnormality index, thereby avoiding the misjudgment of traditional systems that treat all heart rate increases as motion interference or pathological abnormalities. The dynamic early warning decision module serves as the execution control terminal of the system, responsible for implementing the alarm strategy based on risk superposition. This module calculates the accurate blood oxygen value based on the purified waveform, and dynamically adjusts the safety boundary according to the degree of heart rate abnormality output by the previous module. When receiving a non-motion-related heart rate abnormality signal, the module will automatically tighten the tolerance for hypoxia and raise the threshold value for alarm triggering. This mechanism ensures that when the child is in potential pathological risk, the system can trigger the alarm instruction with a more sensitive response speed than the conventional standard, realizing early intervention under physiological parameter linkage analysis and solving the clinical pain point of single parameter alarm lag.

[0043] In summary, the present application has at least the following effects:

[0044] The application discloses an integrated pediatric vital sign intelligent monitoring and early warning method and system, which is characterized in that: a motion intensity envelope signal synthesized by three-axis acceleration is used as an adaptive filtering reference source, and a recursive least square algorithm is used to realize accurate orthogonal removal of motion artifacts in a photoplethysmogram signal in a dynamic scene, so as to ensure the accuracy of blood oxygen monitoring when a child cries; meanwhile, a time series cross-correlation analysis technology is innovatively introduced to quantize the physiological coupling degree of heart rate and motion, to accurately distinguish physiological compensation and pathological abnormality, and to establish a dynamic linkage mechanism of an independent heart rate abnormality index and a blood oxygen alarm threshold, so that the early warning boundary is automatically tightened when a non-motion related risk is identified, thereby greatly reducing the false positive rate caused by body movement, and significantly improving the early warning sensitivity to potential critical compensation periods.

[0045] Those skilled in the art will appreciate that embodiments of the application can be provided as methods, systems or computer program products. Accordingly, the application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can be embodied in the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage media, etc.) having computer usable program code embodied therein.

[0046] The application is described in reference to flowcharts and / or block diagrams of systems, apparatuses (systems), and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, a special purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate an apparatus that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or flows and / or blocks Figure 1 The functions specified in the flow or flows and / or blocks

[0047] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or flows and / or blocks Figure 1 The functions specified in the flow or flows and / or blocks

[0048] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The flowchart blocks Figure 1 The flowchart blocks

[0049] Although preferred embodiments of the application have been described herein, it will be apparent to those skilled in the art that various modifications can be made within the scope of the application without departing from the spirit of the application. Accordingly, it is intended that all such possible modifications be included within the scope of the application as defined in the following claims in which the use of the singular is deemed to include the plural, unless specifically stated otherwise.

[0050] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described herein.

Claims

1. An integrated intelligent monitoring and early warning method for pediatric vital signs, characterized in that, Includes the following steps: S1. Simultaneously acquire the child's electrocardiogram (ECG), photoplethysmography (PPG), and triaxial acceleration signals. Perform vector synthesis on the triaxial acceleration signals and calculate the root mean square envelope to generate the motion intensity envelope signal. At the same time, extract the peak feature points of the ECG signals to construct the R-wave peak interval sequence. S2. The motion intensity envelope signal is used as a non-stationary noise reference source. The photoplethysmography (PPG) signal is denoised by projection based on the recursive least squares adaptive filtering algorithm. The motion artifacts in the PPG signal are orthogonally projected and removed by the fluctuation characteristics of the motion intensity envelope signal. The reconstructed blood oxygen waveform is then output. S3. Interpolate the R-wave peak interval sequence to match the sampling rate of the exercise intensity envelope signal. Calculate the maximum cross-correlation coefficient between the exercise intensity envelope signal and the interpolated R-wave peak interval sequence within a preset physiological response lag window. Compare the maximum cross-correlation coefficient with a preset physiological coupling threshold. When the R-wave peak interval sequence shows an upward trend in heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold, generate an independent heart rate abnormality index. S4. Analyze and reconstruct the blood oxygen waveform to calculate the real-time blood oxygen saturation value. Use the independent heart rate abnormality index as the adjustment factor for the blood oxygen alarm threshold. When the independent heart rate abnormality index is generated, increase the preset hypoxia threshold to tighten the alarm boundary. If the real-time blood oxygen saturation value is lower than the increased hypoxia threshold, an alarm command is triggered.

2. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 1, characterized in that: The specific process of simultaneously acquiring the child's electrocardiogram (ECG), photoplethysmography (PPG) pulse wave, and triaxial acceleration signals, vector synthesizing the triaxial acceleration signals, calculating the root mean square envelope, and generating the motion intensity envelope signal is as follows: Using the sampling time of the photoplethysmography pulse wave signal as the reference timestamp, the electrocardiogram signal and the triaxial acceleration signal are linearly interpolated and resampled. The instantaneous components of the resampled triaxial acceleration signal in the X, Y and Z axes are extracted, and the arithmetic square root of the sum of squares of the three components is calculated to obtain the instantaneous resultant acceleration vector. Within a set sliding time window, the root mean square operation is performed on the instantaneous resultant acceleration vector to extract the low-frequency trend term and filter out high-frequency jitter noise, generating a motion intensity envelope signal that reflects the overall change in the child's kinetic energy.

3. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 1, characterized in that: Simultaneously, the specific process of extracting peak feature points from the electrocardiogram signal and constructing the R-wave peak interval sequence is as follows: The synchronously acquired electrocardiogram signals are sequentially subjected to bandpass filtering, first-order differentiation processing, and squaring to generate a pre-processed signal with enhanced energy. An adaptive floating threshold is set that dynamically changes with the amplitude of the preprocessed signal. The amplitude of the preprocessed signal is compared with that adaptive floating threshold, and the maximum value point exceeding the adaptive floating threshold is located as the R-wave candidate point. Based on the preset refractory period time window, the candidate points of the R-wave are checked by time sequence logic, the effective R-wave peak time is retained, the time difference between adjacent effective R-wave peak times is calculated, and the R-wave peak interval sequence is constructed by arranging them in time sequence.

4. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 1, characterized in that: The specific process of projection denoising of the photoplethysmography pulse wave signal, based on the recursive least squares adaptive filtering algorithm and using the motion intensity envelope signal as a non-stationary noise reference source, is as follows: The motion intensity envelope signal is defined as the reference input vector of the filter, the photoplethysmography pulse wave signal is defined as the desired response signal, the weighting coefficient vector and the inverse of the autocorrelation matrix of the filter are initialized, and the Kalman gain vector is calculated using the reference input vector at the current time. The prior estimation error is calculated based on the difference between the photoplethysmography pulse wave signal and the filter output value. The weighting coefficient vector is then iteratively updated using the Kalman gain vector to generate the converged filter weighting coefficients.

5. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 4, characterized in that: The specific process of orthogonally projecting and removing motion artifacts in the photoplethysmography (PPG) wave signal based on the fluctuation characteristics of the motion intensity envelope signal, and outputting the reconstructed blood oxygen waveform, is as follows: The converged filter weight coefficients are convolved with the current motion intensity envelope signal to estimate the fitting motion artifact component in the photoplethysmography pulse wave signal that is highly correlated with the body motion characteristics. The orthogonal projection subtraction operation is performed to subtract the fitted motion artifact component from the original photoplethysmography pulse wave signal point by point, retaining the physiological blood flow component orthogonal to the motion intensity envelope signal. The residual signal after the subtraction operation is smoothed and filtered to repair the waveform spikes caused by phase lag, and the output waveform has a stable baseline and retains the dicrotic notch characteristics.

6. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 1, characterized in that: The specific process of interpolating the R-wave peak interval sequence, matching the sampling rate of the motion intensity envelope signal, and calculating the maximum cross-correlation coefficient between the motion intensity envelope signal and the interpolated R-wave peak interval sequence within a preset physiological response lag window is as follows: The discrete R-wave peak interval sequence is reconstructed into a continuous sequence using a cubic spline interpolation algorithm, and then resampled according to the time axis of the motion intensity envelope signal to generate a time-aligned heart rate trend sequence. Using the current moment as the endpoint, a pre-defined dual-channel data segment is extracted to construct a physiological response lag window containing the heart rate trend sequence and the exercise intensity envelope signal; Within this window, the exercise intensity envelope signal is kept still, and the heart rate trend sequence is time-domain sliding shifted. The normalized cross-correlation function value under different time lags is calculated, and the global maximum value of the function value is searched through and extracted as the maximum cross-correlation coefficient.

7. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 6, characterized in that: The process of generating an independent heart rate abnormality index is as follows: The maximum cross-correlation coefficient is compared with a preset physiological coupling threshold. When the R-wave peak interval sequence shows an upward trend in heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold, the following steps are taken: Perform first-order differential operation on the interpolated R-wave peak interval sequence to obtain the instantaneous rate of change, calculate the cumulative sum of the rates of change within a preset window, and confirm the upward trend of heart rate when the cumulative sum is positive and exceeds the preset trend judgment threshold. Perform logical AND operation, and activate the abnormal quantification logic if and only if it is confirmed that there is an upward trend in heart rate and the absolute value of the calculated maximum cross-correlation coefficient is less than the preset physiological coupling threshold. Calculate the absolute value of the difference between the maximum cross-correlation coefficient and the physiological coupling threshold, and use it as a weighting factor to map to a preset risk grading scale. Output the quantified value as an independent heart rate abnormality index.

8. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 1, characterized in that: The logical process of analyzing and reconstructing the blood oxygen waveform to calculate the real-time blood oxygen saturation value, and using the independent heart rate abnormality index as the adjustment factor for the blood oxygen alarm threshold, is as follows: The peak and trough values ​​of the reconstructed blood oxygen waveform within one cardiac cycle are extracted, the AC pulsation component and the DC baseline component are separated, the light intensity ratio under dual wavelengths is calculated and substituted into the calibration curve to obtain the real-time blood oxygen saturation value. A linear threshold compensation function is established, and the independent heart rate abnormality index is input into the function as the independent variable. The corresponding threshold upward compensation amount is calculated, and the threshold upward compensation amount is superimposed on the system's default preset hypoxia threshold to construct a dynamically changing current execution alarm threshold.

9. The integrated intelligent monitoring and early warning method for pediatric vital signs according to claim 8, characterized in that: When an independent heart rate abnormality index is generated, the preset hypoxia threshold is increased to tighten the alarm boundary. If the real-time blood oxygen saturation value is lower than the increased hypoxia threshold, the alarm command is triggered. The specific process is as follows: The generation status of the independent heart rate abnormality index is monitored in real time. Once a non-zero independent heart rate abnormality index is received, the threshold dynamic migration mechanism is immediately activated. By adjusting the threshold upward compensation amount, the preset hypoxia threshold is positively offset to obtain the improved high-sensitivity hypoxia threshold. The real-time blood oxygen saturation value is compared with the high-sensitivity hypoxia threshold. If the real-time blood oxygen saturation value is lower than the high-sensitivity hypoxia threshold, a graded alarm command is immediately generated and an audible and visual warning is output.

10. An integrated intelligent monitoring and early warning system for pediatric vital signs, applied to the integrated intelligent monitoring and early warning system for pediatric vital signs as described in any one of claims 1-9, characterized in that, Includes the following modules: The synchronous acquisition and preprocessing module is used to simultaneously acquire the child's electrocardiogram (ECG), photoplethysmography (PPG) pulse wave, and triaxial acceleration signals. It performs vector synthesis of the triaxial acceleration signals and calculates the root mean square envelope to generate a motion intensity envelope signal. At the same time, it extracts the peak feature points of the ECG signals and constructs the R-wave peak interval sequence. The projection denoising and reconstruction module is used to call the motion intensity envelope signal as a non-stationary noise reference source. Based on the recursive least squares adaptive filtering algorithm, it performs projection denoising on the photoplethysmography (PPG) signal. It also performs orthogonal projection to remove motion artifacts in the PPG signal by using the fluctuation characteristics of the motion intensity envelope signal, and outputs the reconstructed blood oxygen waveform. The physiological coupling analysis module is used to interpolate the R-wave peak interval sequence, match the sampling rate of the exercise intensity envelope signal, calculate the maximum cross-correlation coefficient between the exercise intensity envelope signal and the interpolated R-wave peak interval sequence within a preset physiological response lag window, compare the maximum cross-correlation coefficient with a preset physiological coupling threshold, and generate an independent heart rate abnormality index when the R-wave peak interval sequence shows an upward trend in heart rate and the maximum cross-correlation coefficient is lower than the physiological coupling threshold. The dynamic early warning decision module is used to analyze and reconstruct the blood oxygen waveform to calculate the real-time blood oxygen saturation value. The independent heart rate abnormality index is used as the adjustment factor for the blood oxygen alarm threshold. When the independent heart rate abnormality index is generated, the preset hypoxia threshold is increased to tighten the alarm boundary. If the real-time blood oxygen saturation value is lower than the increased hypoxia threshold, an alarm command is triggered.

Images