Short-time dynamic monitoring method and system based on area and wavelet time-frequency boundary locking

Abstract

This invention provides a short-term dynamic monitoring method and system based on area and wavelet time-frequency boundary locking, relating to the field of non-invasive blood pressure monitoring technology. The method specifically includes: using a single cuff pressure scan, jointly calibrating the blood pressure boundary with weighted area morphology scoring and wavelet time-frequency scoring, and establishing an elastic template group arranged by relative pressure index. Stroke-by-stroke direct matching, linear resampling matching, and dynamic time warping matching are performed to calculate area difference and wavelet time-frequency difference, and the current blood pressure boundary and offset direction are inferred from the pressure index corresponding to the minimum joint difference. Cuff pressure is only adjusted with amplitude limitation when the minimum joint difference does not exceed an adaptive threshold, the quality is acceptable, the offset exceeds a dynamic adjustment threshold, and at least two channels are in the same direction during consecutive strokes. After re-locking the boundary, the pressure change is converted into a short-term blood pressure trend. This scheme limits ground-state measurement to the initialization stage, solving the problems of whether blood pressure has changed, the direction of change, and pressure adjustment judgment in short-term monitoring.

Description

Technical Field

[0001] This invention relates to the field of non-invasive blood pressure monitoring technology, and in particular to a short-time dynamic monitoring method and system based on area and wavelet time-frequency boundary locking. Background Technology

[0002] Cuff-based automated blood pressure measurement typically acquires an oscillatory wave superimposed on the cuff pressure during a single bolus or deboosting scan, and estimates systolic, diastolic, and mean arterial pressure based on the oscillatory wave envelope, a fixed amplitude ratio, empirical coefficients, or waveform characteristics. This type of approach is suitable for obtaining discrete blood pressure readings, but it usually releases the cuff pressure after a measurement, making it difficult to answer whether the blood pressure boundaries continue to change in the short term, in what direction they change, and whether the cuff pressure needs to be readjusted.

[0003] Current cuff blood pressure measurements typically obtain a discrete result from a single complete inflation or deflation process; to observe changes over a short period of time, pressure scans need to be performed repeatedly, which limits the timeliness, continuous observation capability, and subject comfort, and is also not conducive to capturing transient blood pressure changes.

[0004] Some monitoring schemes infer blood pressure changes by maintaining cuff pressure and observing changes in oscillation amplitude or single template similarity. However, if the current waveform is dissimilar to a reference template, it only indicates a change in waveform state and cannot reliably distinguish whether the current cuff pressure is on the low-pressure side or the high-pressure side of the new blood pressure boundary. If a decrease in similarity is directly equated with a change in blood pressure and the pump or valve is triggered, respiratory fluctuations, changes in heart rate cycle, body movement, residual cuff perturbations, and signal quality degradation are easily misinterpreted as true blood pressure changes. Some beat-by-beat or continuous blood pressure estimation schemes rely on mathematical regression, empirical mapping, pulse propagation parameters, or waveform extrapolation models to obtain continuous results, but their output lacks a direct oscillation morphology mechanism correspondence between the actual cuff pressure and blood pressure relative state.

[0005] Flexible template matching can reduce the impact of changes in heart rate and beat duration on waveform comparisons through time scaling or nonlinear registration. Direct correlation, linear resampling correlation, and dynamic time warping reflect the waveform similarity under unregistered, linearly registered, and nonlinearly registered conditions, respectively. However, if flexible template matching is only used for single-use blood pressure localization, it still does not solve the problems of what pressure state the template should correspond to in short-term dynamic monitoring, how to obtain the pressure adjustment direction, how to combine it with the cuff pressure closed loop, and when the template can be updated.

[0006] On the other hand, changes in the cuff oscillation wave are not only manifested in changes in the overall morphological similarity, but also in the redistribution of the area on the left and right sides of the main positive region, changes in area after notching, changes in energy due to increased curvature in the valley neighborhood, the appearance or disappearance of time-separated wavelet components after the main peak, and changes in the continuity of the time-frequency ridge. Area features are more sensitive to changes in integral distribution, while wavelet time-frequency features are more sensitive to local peaks and valleys, notches, and time-delay components. However, existing schemes usually only use single or local features for offline waveform analysis or event backtracking, lacking a mechanism for comprehensively determining and verifying the direction of pressure change in dynamic monitoring based on the above features. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention focuses on short-term dynamic monitoring. First, it establishes an individualized blood pressure boundary and its adjacent states using a complete oscillatory wave from a single cuff pressure scan. Then, it jointly determines whether the short-term blood pressure boundary has changed through elastic matching, area difference, and wavelet time-frequency difference, using this as a basis for blood pressure adjustment. This invention proposes a short-term dynamic monitoring method and system based on area and wavelet time-frequency boundary locking, aiming to limit ground-state measurement to the initialization phase, thereby solving the problems of whether blood pressure changes, the direction of change, and whether cuff pressure needs adjustment in short-term monitoring.

[0008] On the one hand, this invention proposes a short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking, which includes the following process: The raw cuff pressure signal of a single pressure scan is acquired and preprocessed to extract the oscillation wave signal. Based on the oscillation wave signal, several effective heartbeats and the baseline pressure corresponding to each effective heartbeat are obtained. The area features and wavelet time-frequency features of each effective heartbeat are extracted and weighted and fused to determine the initial blood pressure boundary value. Based on the initial blood pressure boundary value, an elastic boundary template group with relative pressure index is established. Select the dynamic monitoring mode, and use the initial blood pressure boundary value as the current reference blood pressure boundary to enter the control cycle loop. Perform the following operations in each control cycle: Based on the current reference blood pressure boundary, determine the current target relative pressure, adjust the cuff pressure applied during the current control cycle to the target relative pressure, wait for the preset stabilization time, acquire the original cuff pressure signal and preprocess it to obtain the effective heartbeat and the pressure value of the current effective heartbeat; Feature extraction is performed on the current effective heartbeat to obtain the joint features of the current effective heartbeat, and the current effective heartbeat is jointly matched with the elastic boundary template group to obtain the optimal matching template of the current effective heartbeat; Based on the relative pressure index of the optimal matching template, the relative pressure state corresponding to the current effective heartbeat is determined, and based on the pressure value of the current effective heartbeat, the relative pressure state and the relative pressure index of the optimal matching template, the current blood pressure boundary and offset direction are inferred. Perform quality verification on the optimal matching template for each currently valid heartbeat within the current control cycle; If the current monitoring period is continuous If a number of currently valid heartbeats that pass quality verification meet the continuous same-direction confirmation condition, then according to the continuous The system uses the current blood pressure boundary and offset direction of the current effective heartbeat to generate the blood pressure adjustment command for the next control cycle; otherwise, it performs monitoring anomaly handling. Adjust the blood pressure according to the adjustment command, calculate and output the blood pressure trend value, use the blood pressure boundary locked after adjustment as the current reference blood pressure boundary, and repeat the above operation until the preset single monitoring time is reached.

[0009] Further, the acquisition and preprocessing of the raw pressure signal from a single pressure scan to obtain several effective heartbeats and the baseline pressure of each effective heartbeat includes: Acquire the raw cuff pressure signal from a single boost or depressurization scan; The original cuff pressure signal is baseline separated and filtered to obtain the pressure baseline and oscillation wave component; The oscillating wave component is divided into several heartbeats, and the baseline pressure of each heartbeat is determined based on the pressure baseline. Each heartbeat is evaluated for quality, and those that fail the evaluation are removed to obtain a number of valid heartbeats. The normalized and non-normalized waveforms of each valid heartbeat are saved.

[0010] Further, the step of extracting the area features and wavelet time-frequency features of each effective heartbeat, performing weighted fusion to determine the initial blood pressure boundary value, and establishing an elastic boundary template group with a relative pressure index based on the initial blood pressure boundary value includes: Based on the baseline pressure of all effective heartbeats, the effective heartbeats are sorted to obtain a pressure sorting sequence; For each valid heartbeat in the pressure sorting sequence, the area feature is extracted, and the change in area feature between adjacent valid heartbeats is calculated; For each valid heartbeat in the stress sorting sequence, wavelet time-frequency features are extracted, and the changes in wavelet time-frequency features between adjacent valid heartbeats are calculated. The changes in area characteristics between adjacent valid heartbeats are robustly standardized, and the weights of area characteristics are determined according to their historical reliability and current quality. An initial area score is then calculated. Robustly standardize the changes in wavelet time-frequency features between adjacent valid heartbeats, determine the weights of wavelet time-frequency features based on their historical reliability and current quality, and calculate the initial wavelet score; The initial area and wavelet joint scores are weighted and summed to obtain the initial area and wavelet joint scores. The location where the initial area wavelet joint score transitions or peaks is determined along the pressure sorting sequence, and the baseline pressure of the effective heartbeat corresponding to the location is used as the initial blood pressure boundary value. A template consistency check is performed on all valid heartbeats. After the template consistency check is passed, effective single-beat oscillation waves in the boundary state, the low pressure side of the benchmark, the high pressure side of the benchmark, and the preset pressure state are selected from the oscillation waves of all valid heartbeats, based on the initial blood pressure boundary value. A template is created for each selected effective single-stroke oscillation wave to store the actual cuff pressure, oscillation wave waveform, area characteristics, wavelet time-frequency characteristics, and quality characteristics corresponding to that effective single-stroke oscillation wave; Calculate the relative pressure index of each template with respect to the initial blood pressure boundary value, arrange all templates in ascending order of relative pressure index, and establish an elastic boundary template group.

[0011] Furthermore, the dynamic monitoring mode includes: short-term dynamic monitoring of systolic blood pressure and short-term dynamic monitoring of diastolic blood pressure; the target relative pressure of the dynamic monitoring is: the boundary state of the current target blood pressure boundary or the preset safe side relative pressure within the template coverage area.

[0012] Further, the step of jointly matching the current effective heartbeat with the elastic boundary template group to obtain the optimal matching template for the current effective heartbeat includes: Traverse each template in the elastic boundary template group and calculate the amplitude difference between the current effective heartbeat and the currently traversed template; The normalized and non-normalized waveforms of the current valid heartbeats are directly matched, linearly resampled, and dynamically time-warped matched with the current traversal template, respectively. The similarity of the normalized waveforms under the three matching methods, the similarity of the non-normalized waveforms under the three matching methods, and the cumulative distance of the dynamic time warping are obtained. The similarity of the normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the normalized waveform. The similarity of the non-normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the non-normalized waveform. The fusion similarity of the normalized waveform and the fusion similarity of the non-normalized waveform are weighted and fused to obtain the elastic similarity of the dual waveform fusion. The cumulative distance of dynamic time warping matching is normalized according to the optimal warping path length to obtain the normalized dynamic time warping distance. After mapping the normalized dynamic time warping distance to the normalized waveform and the non-normalized waveform respectively, a weighted fusion is performed to obtain the dual-waveform fused dynamic time warping distance. After normalizing the dynamic time warping distance, elastic similarity, and amplitude difference of the dual waveform fusion, a weighted fusion is performed to obtain the elastic template difference between the current effective heartbeat and the current traversed template. Calculate the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template, respectively; We perform a weighted summation of the elastic template difference, area difference, and wavelet time-frequency difference to obtain the joint difference between the current valid heartbeat and the current traversed template. The template with the smallest joint difference is selected as the optimal matching template for the current valid heartbeat.

[0013] Further, the calculation of the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template includes: Extract the area features of the current valid heartbeat and several area features of the current traversed template respectively; The differences between the current valid heartbeat and the area features of the current traversed template are calculated and weighted and fused to obtain the area difference between the current valid heartbeat and the current traversed template. Extract the wavelet time-frequency features of the current valid heartbeat and the current traversed template respectively; The differences between the current valid heartbeat and the wavelet time-frequency features of the current traversed template are calculated and weighted and fused to obtain the wavelet time-frequency difference between the current valid heartbeat and the current traversed template.

[0014] Furthermore, the weighted summation of the elastic template difference, area difference, and wavelet time-frequency difference to obtain the joint difference between the current valid heartbeat and the current traversed template includes: The elastic template difference, area difference, and wavelet time-frequency difference are respectively mapped to the same dimension to obtain the elastic template mapping difference, area mapping difference, and wavelet time-frequency mapping difference; The weights of elastic template differences, area differences, and wavelet time-frequency differences are determined based on historical reliability and current quality. The elastic template mapping difference, area mapping difference, and wavelet time-frequency mapping difference are weighted and summed according to their respective weights, and a direction conflict penalty term is introduced to obtain the joint difference between the current effective heartbeat and the current traversed template. The method for calculating the direction conflict penalty term is as follows: determine the candidate template direction of the current traversal template; determine the independent directions of elastic template difference, area difference, and wavelet time-frequency difference respectively; calculate the proportion of each independent direction that is inconsistent with the candidate template direction to obtain the direction inconsistency ratio; calculate the degree of cross-level jump of the template index of the current traversal template relative to the optimal template index of the previous effective heartbeat to obtain the index jump penalty term; calculate the dynamic time warping path anomaly degree of the current traversal template to obtain the path anomaly penalty term; and sum the direction inconsistency ratio, index jump penalty term, and path anomaly penalty term by weight to obtain the direction conflict penalty term of the current traversal template.

[0015] Furthermore, if the current monitoring period is continuous If a number of currently valid heartbeats that pass quality verification meet the continuous same-direction confirmation condition, then according to the continuous The current blood pressure boundary and offset direction of each current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle, including: Obtain an initial stable heartbeat set, and determine the template matching quality threshold based on the median and median absolute deviation of the minimum joint differences of each heartbeat in the initial stable heartbeat set; The quality threshold is determined based on the median and median absolute deviation of the overall quality score of each heartbeat in the initial stable heartbeat set; If the minimum joint difference of the current effective heartbeat is greater than the template matching quality threshold, the current effective heartbeat does not trigger blood pressure regulation; if the minimum joint difference of the current effective heartbeat is not greater than the template matching quality threshold, and the overall quality score of the current effective heartbeat is not lower than the quality threshold, then the following operations are performed: The basic dead zone is determined based on the median and median absolute deviation of the absolute values ​​of the boundary offsets of each heartbeat in the initial stable heartbeat set. With continuous One current effective heartbeat is used as a control cycle; A valid heartbeat is defined as a qualified heartbeat if it meets the following conditions: the minimum joint difference of the current valid heartbeat is greater than the template matching quality threshold; the comprehensive quality score of the current valid heartbeat is not lower than the quality threshold; there is a valid discrimination channel; the cuff pressure is within the preset stable range; no safety anomaly is triggered; if there is no qualified heartbeat in the current control cycle, pressure adjustment is not triggered. The area feature variation, wavelet time-frequency feature variation, cuff pressure variation, cardiac interval variation, and body motion risk within the current control period are statistically analyzed, and the dynamic dead zone threshold is determined in combination with the baseline dead zone. The dynamic adjustment threshold is determined based on the dynamic dead zone threshold, the minimum interval between the dead zone and the adjustment threshold, and the overall quality of qualified heartbeats within the current control window. Based on the template matching quality threshold, quality threshold, and dynamic adjustment threshold, the single beat confirmation indicator is calculated according to the minimum joint difference of the current valid heartbeat, the comprehensive quality score, the boundary offset, the validity of the elastic template, and the directional consistency of the elastic template difference, area difference, and wavelet time-frequency difference. Based on the single-shot confirmation indicator, the consecutive same-direction confirmation rate is calculated. Only when the consecutive same-direction confirmation rate is not less than the consecutive confirmation threshold is the consecutive... The current blood pressure boundary and offset direction of the current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle.

[0016] Furthermore, the step of adjusting the blood pressure according to the pressure adjustment command and calculating the blood pressure trend value includes: Calculate continuous The target pressure for the next control period is obtained by summing the robust aggregate value of the current blood pressure boundary of the current effective heartbeat with the target relative pressure of the current monitored blood pressure boundary, and then subtracting the target pressure of the current control period. Adjust the pressure according to the target pressure of the next control cycle, wait until the pressure stability condition is met, re-execute joint matching until the boundary offset returns to the dynamic dead zone, and calculate the blood pressure trend value.

[0017] On the other hand, this invention proposes a short-time dynamic monitoring system based on area and wavelet time-frequency boundary locking, the system comprising: Cuffs or localized pressure units are used to apply adjustable pressure to the subject; The cuff pressure generating and regulating unit is connected to the cuff or local pressurization unit through an air path, and is used to control the cuff or local pressurization unit to generate or release pressure according to the drive signal; Pressure sensors are used to collect real-time pressure signals within the cuff or localized pressurization unit; The signal acquisition and synchronization module is used to transmit real-time pressure signals to the processor; The processor is configured to execute the short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking as described in any one of claims 1-9, and output voltage regulation commands; The pressure controller is used to convert the pressure adjustment command into a drive signal and send it to the cuff pressure generation and adjustment unit; The quality control and safety protection module is used to monitor signal quality and perform exception handling when an anomaly is detected. The memory is used to store real-time pressure signals, effective heartbeats and pressure values, joint features, elastic boundary template groups, blood pressure boundary values, offset directions, and pressure regulation commands. The display or communication output module is used to display the effective heart rate and pressure value, elastic boundary template group, blood pressure boundary value, and offset direction in real time.

[0018] The beneficial effects of adopting the above technical solution are as follows: This invention uses elastic template matching with relative pressure index as the main criterion for short-term dynamic tracking, and area and wavelet methods are used for initial boundary calibration, dynamic auxiliary verification and recalibration after template failure, thereby forming an interpretable blood pressure boundary closed-loop tracking chain.

[0019] When each template is bound, the actual cuff pressure and the relative pressure index relative to the calibrated target boundary can be used to infer the current blood pressure boundary and identify the offset direction by using the boundary state or other preset pressure states that have been indexed and calibrated.

[0020] In systolic pressure mode, the present invention can select a preset safety state on the low cuff pressure side when the template is covered and the safety gating is allowed, reducing the time of continuous approach to the blood flow occlusion state; when the matching fails, the pressure adjustment and template update are frozen and area-wavelet joint recalibration is performed to reduce the risk of mis-adjustment of pressure caused by noise, contact changes or template drift.

[0021] This invention balances short-term trend response, anti-interference capability, and security through reliability-quality weighting, continuous effective heartbeat confirmation, graded threshold state machine, amplitude limiting voltage regulation, and separate updating of reference / working templates. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking provided in an embodiment of the present invention; Figure 2 The flowchart for preprocessing, heartbeat segmentation, and time-frequency feature extraction provided in this embodiment of the invention; Figure 3 This is a flowchart of the initial blood pressure boundary calibration provided in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the short-term systolic pressure elastic boundary locking and cuff pressure closed-loop adjustment provided in an embodiment of the present invention. Figure 5 A flowchart of short-term diastolic pressure elastic boundary locking and cuff pressure closed-loop adjustment provided in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the establishment of the pressure index elastic boundary template and the calculation of joint differences provided in an embodiment of the present invention. Detailed Implementation

[0023] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the invention. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application. The embodiments are used to illustrate the technical concept and possible implementation methods, and are not intended to limit the scope of protection. Without departing from the core concept of the present invention, pressure scanning methods, heart rate segmentation methods, elastic registration methods, area features, wavelet types, weight calculations, threshold settings, control parameters, and hardware forms can be equivalently substituted.

[0024] In this invention, the ground-state area method and wavelet method are only used to initialize the blood pressure boundary, configuring a single ground-state blood pressure measurement as the initial boundary calibration step. After the initial boundary calibration, the system performs subsequent boundary dynamic change identification and cuff pressure closed-loop control. An elastic template group with relative pressure indexes is established on the low-pressure side, boundary state, and high-pressure side of the boundary. The current boundary change direction is determined through multiple relative pressure state templates. Direct matching, linear resampling, and dynamic time warping are combined to reduce the impact of changes in heartbeat duration on matching, and area difference and wavelet time-frequency characteristics are used to assist in verifying the direction determination. The current boundary state is inferred from the combined difference of the elastic template, area, and wavelet. A graded state machine is constructed by using matching quality threshold, dead zone, observation zone, adjustment threshold, safety threshold, and continuous effective heartbeat confirmation. Combined with pressure regulation limiting and template update freezing mechanisms, a comprehensive decision is made to maintain, observe, regulate, recalibrate, or exit, reducing the risk of erroneous pressure regulation caused by respiratory fluctuations, random noise, abnormal pulses, and residual pump disturbances. This forms a complete control closed loop of "whether there is a change—direction of change—whether to regulate pressure—whether to relock after pressure regulation."

[0025] Example 1: This embodiment presents a short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking, such as... Figure 1 As shown, the method includes the following steps: The raw cuff pressure signal from a single pressure scan is acquired and preprocessed. The oscillation wave signal is extracted, and based on this signal, several valid heartbeats and their corresponding baseline pressures are obtained, including: Acquire the raw cuff pressure signal from one boost or depressurization scan.

[0026] The original cuff pressure signal is subjected to baseline separation and filtering to obtain the pressure baseline and oscillation wave component.

[0027] In this embodiment, after the subject wears the arm cuff, the device is first tested for leaks, the pressure sensor is checked for zero points, and low-pressure static noise is collected. The cuff is then controlled to complete one pressure boosting or depressurizing scan, and the pressure sensor collects the raw cuff pressure signal, which includes the cuff pressure baseline and pulsating oscillation components. , This is the sampling time. For example... Figure 2 As shown, the original cuff pressure signal Baseline separation is performed, and the cuff pressure baseline is obtained through low-pass filtering, sliding polynomial fitting, spline fitting, morphological filtering, or adaptive trend estimation, and the oscillatory wave component is obtained: ; in, This serves as the baseline for cuff pressure. It is a pulsating oscillatory wave component.

[0028] This embodiment performs Fourier spectrum analysis on the original oscillation wave to determine the approximate distribution range of the effective pulsation component, baseline drift, and high-frequency noise. Subsequently, low-pass, high-pass, or band-pass filters are used to remove high-frequency interference and slow-varying drift, and the spectrum quality is checked again. This embodiment can also construct an empirical wavelet filter bank based on the spectrum segmentation results to perform empirical wavelet decomposition and artifact suppression on the single-beat oscillation wave. The aforementioned Fourier analysis and empirical wavelet processing are also used to improve the quality of heartbeat segmentation and elastic template matching, and are not limited to the only preprocessing method.

[0029] The oscillatory wave component is segmented into several heartbeats, and the baseline pressure of each heartbeat is determined based on the pressure baseline. After heartbeat segmentation, with Indicates the first A single beat oscillation wave, with Indicates the first Each heartbeat corresponds to a cuff pressure.

[0030] This embodiment can also perform at least one of the following on the oscillating wave components: detrending, adaptive filtering, empirical mode decomposition, variational mode decomposition, wavelet denoising, or template-constrained reconstruction. Based on the local extrema of the oscillating wave, the ECG R wave, the photoplethysmography pulse wave origin, period prediction, or a combination thereof, the single-beat origin, origin, peak, trough, post-peak segment, descending limb, and candidate boundary sensitive characteristic bands are determined.

[0031] Each heartbeat is evaluated for quality, and those that fail the evaluation are removed to obtain a number of valid heartbeats. The normalized and non-normalized waveforms of each valid heartbeat are saved.

[0032] In this embodiment, heartbeats with unstable pressure, incomplete segmentation boundaries, unclear main peak attribution, or obvious motion artifacts are not considered valid heartbeats for initial template construction or dynamic pressure regulation. Each heartbeat retains both normalized and unnormalized waveforms. Normalized waveforms are used to compare morphological changes independent of absolute amplitude; unnormalized waveforms, in addition to calculating correlation similarity, also calculate peak-to-peak value, root mean square value, energy, or overall positive area amplitude differences to actually retain amplitude information and prevent two waveforms from being considered completely identical simply because of the same correlation coefficient. The two types of waveforms are matched separately within the same candidate template before fusion; selecting different templates separately and then merging is not allowed. For cases with significant differences in heartbeat cycles, the phase is first mapped to a unified interval, and then limited local time scaling is allowed during the elastic matching process.

[0033] like Figure 3 As shown, during the quality assessment process, if the safety status of the heartbeat is abnormal or the quality is unqualified, the boundary judgment is suspended and additional data collection is performed; if there are continuous low-quality heartbeats, the measurement is repeated; when there is a safety abnormality, the scan is stopped and the air is released.

[0034] Area features and wavelet time-frequency features of each effective heartbeat are extracted separately. Robust standardization is performed on these features, and they are then weighted and fused based on a unified weighting system of historical reliability and current quality to determine initial blood pressure boundary values. Based on these initial blood pressure boundary values, an elastic boundary template set with a relative pressure index is established, including: Based on the baseline pressure of all effective heartbeats, the effective heartbeats are sorted to obtain a pressure sorting sequence.

[0035] For each valid heartbeat in the pressure sorting sequence, area features are extracted, and the change in area features between adjacent valid heartbeats is calculated.

[0036] Among them, the area characteristics include: overall positive area, area to the left of the main peak, area to the right of the main peak, left-right area ratio, right-side area proportion, area after notch, area of ​​descending branch, and at least two of the area centroids.

[0037] In this embodiment, the area feature is obtained from a continuous positive region containing the main peak above the area analysis baseline; the area analysis baseline is a local zero line, a line connecting adjacent troughs, a local chord, a trend line, an envelope midline, or a robust fitting baseline; when multiple positive regions exist, the continuous region with the largest positive integral area is selected. Using the main peak, the area centroid, the midpoint of time, or a preset proportion as boundaries, in addition to calculating the left-side area, right-side area, and right-side area ratio, the left-right area ratio, the area after the notch, the area of ​​the descending branch, and the area centroid are also calculated. Specifically, for systolic blood pressure monitoring, the right-side area, the right-side area ratio, and the area after the notch are preferentially used; for diastolic blood pressure monitoring, the overall positive area, the degree of lower envelope opening, the area of ​​the descending branch, and the area centroid are preferentially used.

[0038] ; ; ; ; ; ; ; in, Indicates the area on the left side; , and They represent the first The starting point, dividing point, and ending point of the positive region of the heartbeat; Indicates the baseline for area analysis; This indicates that only positive values ​​should be retained. Indicates the area on the right side; The total positive area; Indicates the area ratio on the right side; This indicates a positive number that prevents the denominator from being zero; Indicates the ratio of the left and right areas; Indicates the area after the notch; This indicates the first effective notch after the main peak, the starting point of the secondary wave, or the preset proportional position. Indicates the centroid of the area.

[0039] For each valid heartbeat in the pressure sorting sequence, wavelet time-frequency features are extracted, and the changes in wavelet time-frequency features between adjacent valid heartbeats are calculated.

[0040] Among them, wavelet time-frequency features include at least two of the following: target region energy within a fixed-scale set, intensity of time-separated components after the main peak, energy of curvature enhancement in the valley neighborhood, length of time-frequency ridge, area of ​​connected regions, and the appearance or disappearance of target components.

[0041] In this embodiment, for wavelet time-frequency analysis, the Mexican hat wavelet, Ricker wavelet, or second-order Gaussian derivative wavelet is preferred, and the same scale set or frequency boundary is used for all valid heartbeats in the same initial scan.

[0042] Specifically, wavelet time-frequency analysis uses the same mother wavelet, scale upper and lower limits, frequency boundaries, and scale step size in the same initial scan and subsequent dynamic monitoring, and sets the region after the main peak in the normalized single-beat time-frequency plane. Right-side delay region Candidate areas for descending branches or collapses and complete and effective analysis area The aforementioned regions are defined by peaks, troughs, notches, normalized time ratios, time-frequency ridges, or individualized templates. This embodiment tracks the energy, temporal location, scale location, time-frequency ridge length, connected region area, and transition from traceable to untraceable states of the target components along the cuff pressure sorting sequence; for diastolic pressure-related changes, trough depth, decrease amplitude, and... The curvature within the wavelet enhances its energy. For the first... One effective heartbeat Perform continuous wavelet transform: ; Setting the region after the main peak in the normalized single-beat time-frequency plane Right-side delay region Candidate areas for descending branches or collapses and complete and effective analysis area For any preset time-frequency region The region energy is defined as: ; The residual energy ratio after the main peak and the intensity of the delayed component on the right side are as follows: ; ; in, For the first The heartbeat in the time frequency region Internal energy; For the first The heartbeat is in the area behind the main peak. Internal energy; For the first Each heartbeat is within the complete and valid analysis area. Internal energy; For the first The heartbeat was in the right delayed area. Internal energy; For scale Time translation Continuous wavelet coefficients at; Mother wavelet; for The complex conjugate; The ratio of residual energy after the main peak; The intensity of the delayed component on the right.

[0043] This embodiment adopts a unified weighting system for historical reliability and current quality, and defines the set of valid items as follows: ; ; in, This indicates the group of objects to be weighted, including the area feature group, wavelet time-frequency feature group, elastic matching mode group, waveform branch group, or discrimination channel group; For the set of valid items; Historical reliability indicates the degree to which this item has the ability to distinguish between different calibrated pressure states and remains stable within the same pressure state. For the first The current mass of the term in each heartbeat indicates the term's mass in the th heartbeat. The real-time availability and reliability of individual heartbeats; Indicates the candidates within this group; The median separation between the low cuff pressure side, the boundary state, and the high cuff pressure side; The median absolute deviation or interquartile range within the same pressure state; before the initial template is established, and Iterative estimation can be performed using equipment calibration data or candidate boundary neighborhoods from the initial pressure scan; after the template group is established, its historical reliability is updated using samples from the low cuff pressure side, boundary state, and high cuff pressure side. When this happens, all weights of the feature group are reset to 0 and the group is deemed invalid.

[0044] when At that time, for the set of valid items Perform normalization, and express the normalized weights as follows: ; in, For the first Normalized parent weights of the items; For summation index, it does not represent a new feature. If If the weights of the group are all reset to zero, the group will be deemed invalid. The weights of area, wavelet time-frequency, the three elastic matching methods, normalized waveform branches or non-normalized waveform branches, and the three discrimination channels of elastic-area-wavelet are all calculated by the normalized weight formula of the unified weight system of historical reliability and current quality; the second weight system will not be established or explained in the following text.

[0045] The changes in area characteristics between adjacent valid heartbeats are robustly standardized, and the weights of area characteristics are determined according to their historical reliability and current quality. An initial area score is then calculated.

[0046] In this embodiment, to compensate for differences in separation capability and current availability due to different area features, let... For the first A set of non-empty effective area features of each heartbeat. Let the area feature number be the initial score; the area initial score and area feature weights are expressed as follows: ; in, Initial score for area; For the first The area feature weights of the area features are calculated by the normalized weight formula; a larger weight is assigned to area features that have high separation between adjacent pressure states and low dispersion within the same state. This indicates robust standardization using the median and median absolute deviation, interquartile range, or equivalent scale. For the first Item area characteristics; Representing area characteristics The directional change between adjacent effective heartbeats or relative to local steady state, i.e., the change in area characteristics.

[0047] In this embodiment, systolic blood pressure candidate scores preferentially utilize right-side area reduction, right-side area percentage or left-right area ratio transitions, and post-notch area reduction; diastolic blood pressure candidate scores preferentially utilize overall positive area opening, lower envelope expansion, post-notch area appearance, and area centroid shift. When area change directions conflict with each other, the area channel quality is reduced, and output boundaries are not forced based on a single area feature.

[0048] The location where the initial area score jumps or peaks is determined along the pressure sorting sequence, and this location is used as a candidate location for area.

[0049] The changes in wavelet time-frequency features between adjacent valid heartbeats are robustly standardized. The wavelet feature weights are determined according to the historical reliability and current quality of the wavelet time-frequency features, and the initial wavelet score is calculated.

[0050] set up For the first The set of wavelet time-frequency features currently available in a valid heartbeat, in order to This indicates the wavelet time-frequency feature number. The time-frequency features of the wavelet are denoted as Its directional change relative to the adjacent effective heartbeat or local steady-state reference in the pressure sorting sequence is denoted as The initial score for wavelet is defined as: ; in, Initial score for wavelet; For the first The wavelet time-frequency feature weights are calculated based on the normalized weight formula. The time-frequency boundary state feature vectors output by wavelet time-frequency analysis can be selected from factors such as regional energy, energy ratio, intensity of time-separated components after the main peak, energy of curvature enhancement in the valley neighborhood, ridge length, ridge continuity, area of ​​connected regions, or the appearance / disappearance state of the target component.

[0051] The position where the initial wavelet score reaches a stable transition or a peak is determined along the pressure sorting sequence, and this position is used as a candidate position for the wavelet.

[0052] Based on the historical reliability and current quality of the initial area score and the initial wavelet score, the initial area score and the initial wavelet score are weighted and summed to obtain the initial area-wavelet joint score. ; in, For the first Initial area wavelet joint score of stress-ordered heartbeats; and These are the fusion weights for the area initial scoring branch and the wavelet initial scoring branch, respectively. When neither branch has a valid term, only the other valid branch forms the candidate; when both branches are invalid, no initial boundary is output.

[0053] by This indicates the blood pressure boundary type to be defined in this study, among which... These represent the states corresponding to the systolic pressure SBP boundary and the diastolic pressure DBP boundary, respectively. Determined along the pressure sorting sequence. The stable transition position or peak candidate position is determined, and the first peak position is determined by combining the corresponding area change direction, wavelet time-frequency change direction and physiological sequence constraints. Initial blood pressure boundary; the representative cuff pressure at this position is denoted as and will As a subsequent relative pressure index The only benchmark.

[0054] The location where the initial area wavelet joint score transitions or peaks is determined along the pressure sorting sequence, and the baseline pressure of the effective heartbeat corresponding to this location is used as the initial blood pressure boundary value.

[0055] In this embodiment, at least one blood pressure boundary is jointly determined based on the area morphology of the effective heartbeats after pressure sorting and the wavelet time-frequency characteristics. The systolic blood pressure boundary is jointly defined by the reduction of the right-side area, the transition of the right-side area ratio or the left-right area ratio, and the weakening of the wavelet component after the main peak in time separation; the diastolic blood pressure boundary is jointly defined by the opening of the overall area, the expansion of the lower envelope, and the energy change of the curvature enhancement in the trough neighborhood.

[0056] It should be noted that the initial blood pressure boundary value is obtained when the area candidate position and the wavelet candidate position overlap within the preset pressure neighborhood and satisfy the physiological order constraint. If the area and wavelet candidates are inconsistent, the effective set corresponding to any weight denominator is empty, the effective heartbeats are insufficient, or the physiological order constraint is not satisfied, it is necessary to expand the initial candidate neighborhood, extend the pressure scan range and supplement the effective heartbeats, return to feature extraction, or output a remeasurement prompt. The initial blood pressure boundary value and its corresponding initial locking cuff pressure, boundary confidence, and effective heartbeats near the boundary can be used to establish the template group.

[0057] A template consistency check is performed on all valid heartbeats. After the template consistency check is passed, effective single-beat oscillation waves in the boundary state, the low pressure side of the reference, the high pressure side of the reference, and the preset pressure state are selected from the oscillation waves of all valid heartbeats, based on the initial blood pressure boundary value. The preset pressure state is within the template coverage area and the pressure index calibration has been completed.

[0058] Before establishing the initial template set, a template consistency check must be performed, including a template consistency quality check and a bidirectional scan consistency check. In this embodiment, the template consistency quality check is as follows: normalized effective heartbeats under the same or adjacent pressure states are sequentially used as evaluation templates, and full-cycle and characteristic band DTW matching is performed with the remaining effective heartbeats; only when the set of templates to be evaluated... satisfy At that time, the average value of the normalized distance of the path length is calculated; otherwise, it is determined that the template sample is insufficient. Only after the template consistency check is passed and the template group is established can the subsequent short-term dynamic monitoring be carried out.

[0059] ; in, This is the set of valid heartbeats to be used to build the template group; and These are the serial numbers of two different heartbeats; and for and The corresponding normalized waveform; For heartbeat and The path length is normalized to the dynamic time warped distance; To obtain the arithmetic mean; This is the average distance for template consistency. A pressure index template group is only allowed to be established if this average distance does not exceed the template consistency threshold calibrated by the equipment; if it exceeds the threshold, it indicates abnormal body movement, contact status, or waveform instability, and dynamic monitoring is not initiated. Check whether the average distance for each pressure state is lower than the template consistency threshold, and whether the optimal indexing order of low-pressure side—boundary state—high-pressure side is stable. If the dispersion within a certain state is too large, abnormal beats are removed, the acquisition time for that pressure state is extended, or a rescan is performed.

[0060] A template is created for each selected effective single-beat oscillation wave to store the actual cuff pressure, oscillation wave waveform, area characteristics, wavelet time-frequency characteristics, and quality characteristics corresponding to that effective single-beat oscillation wave.

[0061] Calculate the relative pressure index of each template with respect to the initial blood pressure boundary value, arrange all templates in ascending order of relative pressure index, and establish an elastic boundary template group.

[0062] In this embodiment, with Indicates the first The initial blood pressure boundary value of the blood pressure class boundary, in Effective single-beat oscillations were selected for the low cuff pressure side, boundary state, and high cuff pressure side, respectively, to establish: ; ;

[0063] in, For the first Elastic boundary template set for blood pressure-like boundaries; For the first One template; Number of templates; To establish the first Cuff pressure when using a template; Indexed by relative pressure; each template is arranged according to... Arranged from smallest to largest, the template group must include at least one... The low-pressure side template of the cuff, a neighborhood located within a preset boundary state (i.e. A boundary state template (close to zero) and a High-pressure side template with cuffs; To normalize the complete single-shot template; This is a non-normalized complete single-shot template; Templates for boundary-sensitive feature bands; It is the area feature vector; This represents the wavelet time-frequency eigenvector; This refers to the template quality vector or quality flag.

[0064] The template heartbeat can be located at the target blood pressure boundary itself, near the boundary, on the low cuff pressure side, on the high cuff pressure side, or at other preset pressure states that have been calibrated with pressure indexing within the template coverage area, without being required to be located at the target blood pressure boundary itself. Regardless of the preset pressure state from which the template originates, each template saves the actual cuff pressure at the time of creation. Compared to the already calibrated first Relative pressure index of target blood pressure boundary The waveform, area features, wavelet time-frequency features, and quality features are also included; waveforms that do not save the actual cuff pressure and relative pressure index are not directly used to infer the absolute blood pressure boundary.

[0065] Each template stores at least the normalized waveform, non-normalized waveform, boundary-sensitive feature bands, area feature vector, and wavelet time-frequency feature vector. The template group includes a baseline template group that remains unchanged during a short-term monitoring period and a working template group that is updated only when the update gating conditions are met. The historical reliability before the initial template establishment is estimated by the equipment calibration data or the candidate boundary neighborhood of the initial pressure scan. After the template group is established, it is updated by samples from the low cuff pressure side, boundary state, and high cuff pressure side.

[0066] The elastic boundary template group may include 3 to 11 pressure states, with pressure intervals between adjacent states ranging from 1 mmHg to 5 mmHg and not requiring equal intervals; each template must retain a clear relative pressure index. The template group shall at least cover the low cuff pressure side, the boundary state, and the high cuff pressure side, and may include other preset pressure states within the coverage area that have completed relative pressure index calibration.

[0067] After the initial template group is established, the system can release some of the cuff pressure and re-reach the target relative pressure according to the selected monitoring mode. . This indicates the cuff pressure that is expected to be maintained during dynamic monitoring relative to the current cuff pressure. The target bias for blood pressure boundary conditions can be taken near the boundary state or at a preset safe side position that meets safety requirements. Its value is determined by the monitoring mode, template coverage, and device safety strategy, and remains unchanged during the same short-term dynamic monitoring and trend conversion process. If the low-cuff pressure side or high-cuff pressure side of the template group is incomplete, or the joint difference sequence between adjacent templates is unstable, closed-loop dynamic monitoring will not be entered; supplementary data collection or re-execution of the initial pressure scan will be performed first.

[0068] For systolic blood pressure monitoring, the preset safety side position can be set on the low cuff pressure side of the target systolic blood pressure boundary, so that distal blood flow passes through part or periodic heartbeats to reduce the duration of near-complete blockage. Closed-loop tracking is only used when the safety side position is within the template coverage area that has been calibrated with relative pressure index, and boundary back-inference, quality gating and safety conditions are all effective.

[0069] If the low-pressure side and high-pressure side of the template group are incomplete, or the joint differences between adjacent templates cannot present a stable sequence, the equipment will not enter dynamic monitoring, but will prompt for remeasurement or expansion of the initial scan.

[0070] Select the dynamic monitoring mode, use the initial blood pressure boundary value as the current reference blood pressure boundary, and perform the following operations: The dynamic monitoring mode includes: short-term dynamic monitoring of systolic blood pressure and short-term dynamic monitoring of diastolic blood pressure; the working pressure of the dynamic monitoring is the boundary state corresponding to the target blood pressure boundary or the preset safe side relative pressure within the template coverage area, wherein the preset safe side relative pressure of the systolic blood pressure mode is located on the low cuff pressure side of the target systolic blood pressure boundary to allow distal blood flow to pass through partially or periodically.

[0071] The duration of a single short-term dynamic monitoring session shall not exceed 5 minutes; the maximum continuous occlusion time and pressure release interval shall be set for systolic blood pressure monitoring. Pressure regulation shall be stopped and safe venting shall be performed when safety conditions such as severe physical activity, hypoperfusion, decreased blood oxygenation, pain, cuff overpressure, valve control malfunction, or sensor drift are triggered; if continuous low quality or failure of all templates to reliably match occurs, pressure regulation and working template updates shall be frozen, and area-wavelet joint recalibration shall be performed.

[0072] like Figure 4 As shown, systolic blood pressure boundaries are identified during the initial scan, and pressure index elastic templates are established for the low-pressure side (blood flow through state), the occlusion boundary state, and the high-pressure side (compression enhancement state). After monitoring begins, the cuff pressure is maintained at the target relative pressure at the systolic blood pressure boundary state or a preset safe position. If the current heartbeat best matches the low-pressure side template, and the area channel also selects the low-pressure side index from its template group, the right-side area, post-notch area, or blood flow reappears through relevant wavelet components, increasing the cuff pressure. If the elastic channel best matches the high-pressure side template, and at least one of the area or wavelet channels is in the same direction, the right-side area and blood flow continuously weaken through relevant wavelet components, then the system reduces the cuff pressure. Any pressure adjustment requires qualified template matching, confirmation of continuous effective heartbeats in the same direction, and at least two effective channels having consistent directions. When a safety abnormality occurs, pressure adjustment is stopped and automatic deflation occurs. If the non-safety quality remains unqualified or the template fails, freeze-clean pressure adjustment and template update are performed, area-wavelet joint recalibration is executed, and the template group is rebuilt or expanded before returning to monitoring.

[0073] When blood pressure rises, the current cuff pressure decreases relative to the new systolic pressure boundary, and the previously nearly disappeared right-side area, post-notch area, or post-peak time-separated wavelet component may reappear, making the current heartbeat closer to the low-pressure side blood flow pattern template. (Followed by the following section...) The blood pressure boundary inverse calculation formula is used to gradually increase the cuff pressure after confirming continuous effective heartbeats. After each pressure adjustment, the pressure is allowed to stabilize and then rematched until the optimal template returns to the target boundary state or the preset safe side relative pressure state.

[0074] When blood pressure decreases, the current cuff pressure rises relative to the new systolic pressure boundary, the current heartbeat is closer to the systolic template, and the right-side area and blood flow continue to decrease through relevant wavelet components. The system reduces the cuff pressure and relocks the boundary using amplitude-limited control. If the cumulative systolic pressure maintenance time reaches the upper limit during monitoring, distal perfusion is abnormal, or the user experiences pain, the cuff is immediately deflated regardless of the algorithm's outcome.

[0075] Short-term dynamic monitoring of diastolic blood pressure aims to maintain the state near the transition between the open and collapsed diastolic morphology. The template set includes at least the open state on the low-pressure side, the diastolic pressure collapse boundary state, and the collapse-enhanced state on the high-pressure side. The low-pressure side template may exhibit overall open area, lower envelope expansion, or high similarity between non-collapsed templates; the boundary state and the high-pressure side template may exhibit enhanced curvature energy in the trough neighborhood, local wavelet energy transitions in the descending branch, or increased similarity between collapsed templates.

[0076] If the elastic channel of the current heartbeat matches the low-pressure side template, and at least one of the area or wavelet channels selects the same-side pressure index, increase the cuff pressure to return to the diastolic pressure boundary; when at least two effective channels match the high-pressure side state, decrease the cuff pressure. Since the local morphology near the diastolic pressure may be affected by respiration and heart rate, window-level consistency determination of multiple effective heartbeats is preferred. When the optimal template falls on the low-pressure side and the reversed boundary is higher than the currently locked value, increase the cuff pressure after confirmation of consecutive heartbeats; when the optimal template falls on the high-pressure side and the reversed boundary decreases, decrease the cuff pressure. For patients with significant respiratory changes, first take the median of the reversed boundaries of multiple heartbeats within a sliding window, and then decide whether to adjust the pressure.

[0077] like Figure 5 As shown, templates for the low-pressure side open state, the diastolic pressure boundary state, and the high-pressure side collapsed enhanced state are established near the diastolic pressure boundary. The low-pressure side template retains the overall area open, lower envelope expansion, and non-collapsed wavelet features; the boundary state and the high-pressure side template retain the energy enhancement of the valley neighborhood curvature, the local energy transition of the descending branch, and the collapsed state morphology.

[0078] Systolic and diastolic blood pressure are typically monitored at different times near their respective pressures. This embodiment does not require a fixed cuff pressure to be near both boundaries simultaneously.

[0079] Based on the current reference blood pressure boundary, determine the current target relative pressure, adjust the cuff pressure applied during the current control cycle to the target relative pressure, wait for the preset stabilization time, acquire the original cuff pressure signal and preprocess it to obtain the effective heartbeat and the pressure value of the current effective heartbeat.

[0080] The target relative pressure is: the boundary state of the current target blood pressure boundary or the preset safe side relative pressure within the template coverage area.

[0081] In this embodiment, during dynamic monitoring, quality assessment can be performed by calculating a comprehensive quality score for each cardiac oscillation. The comprehensive quality score can be composed of local signal-to-noise ratio, cuff pressure stability, cardiac segmentation reliability, elastic regularity path anomaly, intratemplate correlation similarity, area and wavelet direction consistency, proportion of abnormal cardiac beats, body motion risk, and distal perfusion status. The definition of the first... The overall quality score of the current heartbeat is: ; in, For the first The overall quality score of each heartbeat; The quality score is obtained from the local signal-to-noise ratio; The quality score is obtained from the cuff pressure stability. The quality score is derived from the completeness of heartbeat segmentation and the reliability of key points. The elastic quality score is obtained by considering elastic matching similarity, dynamic time warping path anomaly, and registration rationality. The area quality score is derived from the area baseline, area region integrity, and area feature availability. The wavelet time-frequency quality score is derived from the effectiveness of the target time-frequency region, the continuity of the ridge line, and the degree of local artifacts. The body motion risk is derived from acceleration, abrupt changes in pressure baseline, high-frequency energy of oscillating waves, or combinations thereof, and therefore For the dynamic mass term; For remote perfusion quality scoring; all quality items are normalized to ; , All are non-negative quality weights, and their sum is 1. Area and wavelet direction consistency can be incorporated separately. , Or it can be used as an additional gating condition. Low-quality heartbeats are not used for template updates, voltage regulation triggers, or trend outputs.

[0082] Feature extraction is performed on the current effective heartbeat to obtain the joint features of the current effective heartbeat.

[0083] The joint features include: the normalized waveform, the non-normalized waveform, the boundary-sensitive feature band, the area feature, and the wavelet time-frequency feature of the current oscillation wave.

[0084] In this embodiment, the cuff pressure is brought into a constant or quasi-constant pressure maintenance state near the blood pressure boundary, and the normalized waveform, non-normalized waveform, boundary-sensitive feature band, area feature, and wavelet time-frequency feature of the current oscillation wave are extracted stroke by stroke.

[0085] The elastic boundary template group has been established and the target relative pressure has been selected. Afterward, the system brings the cuff pressure to the target working position and maintains it stable. Before the initial dynamic pressure adjustment, valid heartbeats meeting the following conditions are continuously collected: the cuff pressure has entered the preset stable range and there is no pressure adjustment transient; the heartbeat segmentation is complete; the overall baseline quality is qualified; and there is no severe body movement, air leakage, overpressure, pain, or distal perfusion abnormalities. Heartbeats meeting the above conditions constitute the initial stable heartbeat set. .

[0086] by This indicates the minimum initial stable heart rate preset by the device. Determined by equipment calibration, and must at least meet the requirements for calculating robust statistics such as the median and median absolute deviation. Only when Only when Estimate the dynamic time-warped distance scale, amplitude difference scale, template matching quality threshold, overall quality threshold, and fundamental dead zone; if If a stable heartbeat is still not collected, continue collecting it. If the heartbeat is still insufficient, use the backup threshold that has been calibrated by the device or re-execute the initialization.

[0087] like Figure 6 As shown, based on the joint features of the current effective heartbeat, the current effective heartbeat is jointly matched with the elastic boundary template group to obtain the optimal matching template for the current effective heartbeat, including: Traverse each template in the elastic boundary template group and calculate the amplitude difference between the current effective heartbeat and the currently traversed template.

[0088] In this embodiment, the amplitude difference between the current effective heartbeat and any one of the peak-to-peak value, root mean square value, energy, or overall positive area of ​​the current traversed template is calculated.

[0089] The normalized and non-normalized waveforms of the current valid heartbeats are directly matched, linearly resampled, and dynamically time-warped matched with the current traversal template, respectively. The similarity of the normalized waveforms under the three matching methods, the similarity of the non-normalized waveforms under the three matching methods, and the cumulative distance of the dynamic time warping are obtained.

[0090] During monitoring, this embodiment retains both normalized and non-normalized waveforms for each current valid heartbeat and extracts the same feature bands as the template. For each candidate template, direct correlation, linear resampling correlation, and DTW registration correlation are calculated sequentially. The DTW path is limited to a certain proportion of the search band within the single-beat cycle to avoid mismatching the main peak to the descending branch or mismatching the notch to adjacent bands.

[0091] In this embodiment, let the first A template in the waveform branch The reference sequence inside is , No. The corresponding sequence of each currently valid heartbeat is The waveform branch This includes normalized and non-normalized waveform branches. Direct matching first ensures the two sequences have the same length by truncation, padding, or selection of common valid intervals, and then calculates the direct correlation similarity. Linear resampling matching linearly stretches or compresses the current sequence to the same length as the reference sequence before calculating the correlation similarity.

[0092] ; ; in, and These are the similarities for direct matching and linear resampling matching, respectively. Represents Pearson correlation, robust correlation, cosine similarity, or equivalent normalized similarity; This indicates linear resampling. Direct matching is more sensitive to local time misalignment. Linear resampling can correct for overall cardiac cycle expansion and contraction, but it cannot completely correct for nonlinear misalignment of local peaks, troughs, and notches.

[0093] For lengths respectively and Reference sequence With the current sequence Establish a local distance matrix, where, and All are reference sequences Sampling points in; and All are current sequences The sampling points in the data. The local distance of a complete single cycle can be expressed using the squared Euclidean distance, while the local distance of boundary-sensitive characteristic bands can be expressed using the absolute difference or the Euclidean distance after robust scaling.

[0094] For the reference sequence sampling point number and the current sequence sampling point number The cumulative distance with dynamic time warping is obtained recursively: ; in, Sampling points and sampling points Local distance between them; The cumulative distance; the boundary conditions are: , ; Sampling points and sampling points Local distance between them; Sampling points and sampling points Local distance between them; Sampling points and sampling points Local distance between them; This indicates a boundary condition where the cumulative distance is 0; , All of these are boundary conditions that are not allowed to be matched.

[0095] The regularized path satisfies start-end constraints, continuity, and monotonicity, and is subject to Sakoe-Chiba bands, Itakura parallelograms, or maximum time scaling constraints; the cumulative distance is divided by the number of grid points in the optimal regularized path and then normalized by distance mapping.

[0096] Map the current sequence to the reference template time axis according to the optimal regularization path, and calculate the similarity of dynamic time regularization matching.

[0097] ; in, Similarity for dynamic time-warped matching; This indicates nonlinear registration along the optimal regularized path. When the path is excessively bent, the path length is abnormally increased, the main peak is matched to the descending branch, the notch is matched to a non-adjacent band, or the complete single cycle significantly conflicts with the characteristic band results, the regularized path is judged as abnormal and the quality of the elastic channel is reduced or the heartbeat is rejected.

[0098] The similarity of the normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the normalized waveform. The similarity of the non-normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the non-normalized waveform.

[0099] For any waveform branch ,by These represent any one of the following matching methods: direct matching, linear resampling matching, and DTW matching. Define the first... The current heartbeat in the waveform branch The set of effective elastic matching methods within is , denoted as: ; in, For the first The historical reliability of the elastic matching method is obtained according to the historical reliability rule in the normalized weight formula; For the first The current quality of the matching method in a given heartbeat can be obtained from the proportion of common valid intervals, linear resampling residuals, DTW path anomaly degree, and the confidence level of related similarity.

[0100] when At that time, the weight of the three-level matching The fusion similarity within the waveform branch is obtained by directly calculating the normalized weight formula.

[0101] ; in, For waveform branches Fusion similarity; For waveform branches In the Similarity under various flexible matching methods; For similarity Normalized parent weights.

[0102] like If the waveform branch is invalid, it will not be included in the subsequent elastic template difference calculation.

[0103] The fusion similarity of the normalized waveform and the fusion similarity of the non-normalized waveform are weighted and fused to obtain the elastic similarity of the dual waveform fusion.

[0104] Let the normalized waveform branch be denoted as Non-normalized waveform branches are denoted as Define the set of currently valid waveform branches as .

[0105] ; in, For waveform branches Historical reliability; For this branch road in the The current quality of each heartbeat; both are still determined according to the normalized weight formula to determine the branch weights. Define the elastic similarity of dual-waveform fusion as: ; in, For dual-waveform fusion elastic similarity; superscript This represents the fusion result of the elastic discrimination channel.

[0106] like Then the elastic discrimination channel invalid.

[0107] After mapping the normalized dynamic time warping distance to the normalized and non-normalized waveforms respectively, a weighted fusion is performed to obtain the dual-waveform fused dynamic time warping distance.

[0108] In this embodiment, the regularized path satisfies start-end constraints, continuity constraints, and monotonicity constraints. To reduce the impact of sequence length and enable its integration with similarity differences, the channels... Dynamic time-normalized distance based on optimal normalized path length First, map channel by channel, then fuse. This can be represented as: ; ; ; in, Represents the reference sequence The Middle Each sampling point and the current sequence The Middle The cumulative distance of each sampling point; For the first The heartbeat and the first A template in the channel The number of optimal regular path grid points within the range; To normalize to The dynamic time warp distance; The distance scale is determined by the median distance, interquartile range, or other robust statistics within the initial stable window. This embodiment also sets the Sakoe-Chiba band, Itakura parallelogram, or a preset maximum time scaling ratio to limit excessive regularization that is not physiologically meaningful. The distance for dual-channel fusion dynamic time warping.

[0109] After normalizing the dynamic time warping distance, elastic similarity, and amplitude difference of the dual waveform fusion, a weighted fusion is performed to obtain the elastic template difference between the current effective heartbeat and the current traversed template.

[0110] In this embodiment, let and They represent the current effective heartbeat and the number of beats, respectively. The amplitude of a template is determined by one of the following: peak-to-peak value, root mean square value, energy, or overall positive area. The relevant similarity range is... Convert it to the difference quantity before merging it. Differences in scope.

[0111] ; ; in, For logarithmic magnitude differences; This represents the normalized amplitude difference; This is the initial stable window amplitude difference scale.

[0112] ; in, Differences in flexible templates; and These are the non-negative weights for dynamic time warping differences and correlation similarity differences, respectively. The amplitude difference is used to ensure that the non-normalized channel retains absolute amplitude information. The complete single-beat DTW distance, boundary-sensitive feature band DTW similarity, and third-level similarity can be used together, or a portion of them can be selected according to the computing power of the embedded processor. Regardless of the combination used, the logic of "taking the maximum similarity or minimum difference in template library matching" is maintained. The decrease in similarity or increase in difference is only used as a change indication in the monitoring of adjacent cardiac aberrations.

[0113] In this embodiment, direct correlation similarity, correlation similarity after linear resampling, and correlation similarity after dynamic time warping are calculated separately within the complete single-cycle and boundary-sensitive feature bands. The dynamic time warping distance is first normalized according to the warped path length, and then mapped to the stable window distance scale. The matching results of normalized and non-normalized waveforms are first fused within each candidate template. The non-normalized channel simultaneously introduces differences in peak-to-peak value, root mean square value, energy, or overall positive area amplitude, and then the joint differences between different candidate templates are compared. Complete single-cycle matching primarily determines whether the overall waveform and periodic structure are consistent with the pressure index template. Feature band matching primarily compares the latter part of the main peak, the descending branch, the notch, the trough neighborhood, or other boundary-sensitive regions. The two types of results are fused after forming non-negative weights based on their respective historical separation degree and current segmentation quality. When the current complete and valid heartbeat is used as the working template for an adjacent cycle, its result is only used for rapid mutation indication; the pressure direction is still determined by the pressure index baseline template group.

[0114] Calculate the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template.

[0115] The calculation of the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template includes: Extract the area features of the current valid heartbeat and several area features of the current traversal template respectively; the area features include: overall positive area, left side area, right side area, left-right area ratio, right side area percentage, area after notch, area of ​​descending branch, degree of lower envelope openness and area centroid.

[0116] The differences between the current valid heartbeat and the area features of the current traversed template are calculated and weighted to obtain the area difference between the current valid heartbeat and the current traversed template.

[0117] In this embodiment, area difference is used to measure the change in integral distribution between the current heartbeat and a certain pressure index template. To avoid directly adding features with different dimensions such as total area, area ratio, and area centroid, the feature is first dimensionless based on the median absolute deviation or interquartile range within the initial stable window, and then the area feature weights are calculated according to historical reliability and current quality. The area features include total positive area, left side area, right side area, left-right area ratio, right side area proportion, area after notch, descending limb area, lower envelope openness, and area centroid. For dynamic monitoring of systolic blood pressure, right side area, right side area proportion, left-right area ratio, and area after notch are preferentially retained; for dynamic monitoring of diastolic blood pressure, total positive area, lower envelope openness, descending limb area, and area centroid are preferentially retained. Area features with higher historical reliability, greater separation between adjacent pressure states, and smaller dispersion within the same pressure state receive higher weights.

[0118] If the current effective area feature set Calculate the area difference between the current valid heartbeat and the area of ​​the current traversed template.

[0119] ; ; in, and The first The current effective heartbeats and the first The first template Item area characteristics; For the first Robust scales for area features, such as the median absolute deviation or interquartile range of the corresponding feature; For the first Dimensionless area differences of area characteristics; For the first Weights of the area feature; This represents the area channel difference. The sum of the weights of the effective area features is 1; features that fail to segment, have an unreliable baseline, or have a current quality of zero are not included in the effective set; the area channel is invalid when the effective set is empty.

[0120] Extract the wavelet time-frequency features of the current valid heartbeat and the current traversed template respectively; the wavelet time-frequency features include: within the fixed scale set , , or At least two of the following: regional energy and energy ratio, intensity of time-separated components after the main peak, energy of curvature enhancement in the trough neighborhood, ridge length, ridge continuity, area of ​​connected regions, appearance or disappearance of target components, and significance of double notches.

[0121] The differences between the current valid heartbeat and the wavelet time-frequency features of the current traversed template are calculated and weighted and fused to obtain the wavelet time-frequency difference between the current valid heartbeat and the current traversed template.

[0122] In this embodiment, wavelet time-frequency differences are used to measure changes in local peaks and valleys, notches, time-delay components, regional energy, ridges, and connected region structure. Each wavelet time-frequency feature is first dimensionless based on its median absolute deviation or interquartile range, and then weighted according to historical reliability and current quality. The wavelet time-frequency features are derived from a fixed-scale set. , , or The analysis comprises at least two of the following: regional energy and energy ratio, intensity of time-separated components after the main peak, energy enhanced by curvature in the trough neighborhood, ridge length, ridge continuity, area of ​​connected regions, appearance or disappearance of the target component, and salience of the double notch. During cross-wave comparisons, the mother wavelet, upper and lower scale limits, frequency boundaries, scale step size, and region definition are kept consistent, and the target component is tracked sequentially along the pressure index. The Mexican hat wavelet serves as the main analysis branch; when the target component cannot be stably separated from the main oscillatory component, Morlet wavelet, complex Morlet wavelet, stationary wavelet transform, multi-resolution decomposition, or equivalent time-frequency method can be invoked within the candidate template neighborhood.

[0123] If the current effective wavelet time-frequency feature set Calculate the wavelet time-frequency difference between the current valid heartbeat and the current traversed template.

[0124] ; ; in, and These are the current effective heartbeats and the number of... The first template Wavelet time-frequency characteristics; For the first Robust scaling of wavelet time-frequency features, such as the median absolute deviation or interquartile range of the corresponding features; The time-frequency difference is a dimensionless wavelet. Wavelet feature weights; This represents the time-frequency channel difference of wavelet features. The sum of the weights of all valid wavelet features is 1; features that exceed the target region boundary, have interrupted and uninterpretable ridge lines, have excessively strong local artifacts, or have a current quality of zero are not included in the valid set.

[0125] If the current heartbeat is a high-quality heartbeat, the weight of the elastic channel can be set to the highest among the three channels; if the regular path has abnormal bends, the path length increases significantly, or the normalization and non-normalization results conflict, then enter the observation state or recalibrate, and do not directly adjust the voltage by replacing the elastic channel with area or wavelet single channel.

[0126] In this embodiment, let Let each of the following represent any discriminant channel among the elastic template, area, and wavelet time-frequency channels, for each currently valid heartbeat: With each pressure index template The differences in flexible templates are already located in Mapping area differences and wavelet time-frequency differences to... Each effective channel then independently selects its optimal pressure index and generates a signed direction.

[0127] ; ; in, Differences in elastic template mapping; This represents dimensionless channel differences; For channel mapping differences; For the stable window distance scale of the corresponding channels, including the stable window distance scale of the area channels. and the stable window distance scale of the wavelet time-frequency channel .

[0128] Define the set of currently valid discrimination channels. for: ; in, For the first The historical reliability of each discrimination channel is derived from the cross-state separation and same-state dispersion of samples from the low cuff pressure side, boundary state, and high cuff pressure side in the equipment calibration data or initial pressure scan. For the first The current heartbeat The current quality of each discriminant channel is derived from the signal quality, feature integrity, and corresponding channel quality evaluation of the current heartbeat; both have been used to calculate the fusion weight of the three discriminant channels according to the normalized weight formula. .like If the current heartbeat is determined to have no effective discrimination channel, it will not proceed to the subsequent joint difference and blood pressure regulation determination.

[0129] For any discrimination channel The template with the smallest difference for the discrimination channel is selected from all candidate pressure index templates to obtain the optimal template index for the discrimination channel.

[0130] ; in, For the first The discrimination channel is for the first The optimal template index is independently selected for each current heartbeat. This channel only uses the established template relative pressure index to determine direction, without recalculating new template pressure parameters. Its channel direction is defined as: ; in, For the first The signed boundary direction is given by each discrimination channel; It is a symbolic function; For the first The current cuff pressure corresponding to the current heartbeat; The first template group has been obtained during the initial template group creation phase. Blood pressure boundary, template Corresponding relative pressure index; The currently locked and used benchmark Blood pressure boundary. This indicates that the reverse direction of this channel is an increase in the blood pressure boundary. This indicates a lowered blood pressure threshold. This indicates that the boundary traced back from this channel matches the currently locked boundary. This occurs upon first entering dynamic monitoring. The initial locked blood pressure boundary is set; after each blood pressure adjustment and relocking, the new locked blood pressure boundary is updated.

[0131] After obtaining the directions of each valid discrimination channel, the current evaluation is being conducted on the [number]th [channel]. Each candidate pressure index template determines the direction of the candidate template.

[0132] ; in, Indicates if the current number Individual heartbeats are provided by candidate templates. The boundary offset direction corresponding to the representation; The first template group already obtained in the aforementioned template group creation phase Each template has a relative pressure index. This formula only calls... , and No additional pressure parameters are added.

[0133] The candidate template direction is compared with the directions of each valid discrimination channel, and the cross-level jump of the candidate template index and the DTW regularization path anomaly are considered at the same time. The full-text relevance penalty is uniformly defined as the penalty for direction conflict and candidate anomaly. , represented as: ; in, [·] is a conditional indicator function; it takes the value 1 if the condition is true, and 0 otherwise. The number of currently valid channels; This is the overall optimal template index obtained from the previous valid heartbeat; For the current heartbeat and the first The normalized path anomaly score already obtained when performing DTW matching on a template is directly called from the DTW path quality evaluation result, and the second set of path anomaly parameters is not calculated. , and The non-negative internal weights for the three components—direction conflict, index jump, and path anomaly—are determined by device calibration or preset parameters and remain fixed during a single monitoring period; their sum is 1. For the first dynamic heartbeat, since there is no... The index jump term is set to 0, and the internal weights are renormalized only among the remaining valid penalty components.

[0134] We obtain the joint difference between the current valid heartbeat and the current traversed template by weighted summation of the elastic template difference, area difference, and wavelet time-frequency difference.

[0135] exist At that time, the current i-th fusion weight and mapping difference are used to calculate the current i-th . The current heartbeat and the first Joint difference among candidate templates for: ; in, The weights for the three-discrimination channel fusion are directly derived from the normalized weight formula. and We obtain, and for the same current heartbeat, all candidate templates. Remain unchanged; The overall intensity coefficient of the penalty term is determined by equipment calibration or preset control parameters and is used to adjust the overall impact of the penalty term relative to the differences in the three discrimination channels.

[0136] Compare the joint differences of all candidate templates to determine the first... The overall optimal template index for each current heartbeat and its corresponding minimum joint difference: ; in, For the first The overall optimal template index for the current heartbeat is the only template index called when retroactively inferring the current blood pressure boundary. This represents the minimum joint difference corresponding to the overall optimal template. Subsequent template matching quality thresholds, singlet confirmation, and trend filtering directly utilize the values ​​obtained here. No recalculation At this point, this section only obtains the overall optimal template and the minimum joint difference; the final boundary offset direction, which has not yet been obtained, is not invoked here in advance.

[0137] Based on the relative pressure index of the optimal matching template, the relative pressure state corresponding to the current effective heartbeat is determined, and based on the pressure value of the current effective heartbeat, the relative pressure state, and the relative pressure index of the optimal matching template, the current blood pressure boundary and offset direction are inferred.

[0138] After obtaining the overall optimal template index Then, the relative pressure index, which was determined during the initial template group creation phase, is directly called. , reverse the first The current heartbeat corresponds to the first Blood pressure boundary.

[0139] ; in, The first one obtained by reverse calculation is the current one. Blood pressure boundary; For the overall optimal template In the The relative pressure index in the blood pressure boundary template group. This relative pressure index was obtained during the initial template group creation phase and is only called here without being resolved.

[0140] With the currently locked number Blood pressure boundary To establish a benchmark, the relative pressure state corresponding to the current effective heartbeat is determined, and the signed boundary offset of the current heartbeat is defined. for: ; in, This indicates that the current inversely calculated blood pressure boundary is higher than the currently locked boundary. This indicates that the current reverse-calculated blood pressure boundary is lower than the currently locked boundary. This indicates that the two are consistent. Subsequent comparisons of the dead zone and dynamically adjusted threshold use the offset magnitude. The final voltage regulation direction is used .

[0141] Only in the elastic discrimination channel Belongs to the current set of valid discrimination channels And at least two valid discrimination channels provide With the final offset direction Only when consistent are the current heartbeats allowed to enter the subsequent dead zone, continuous confirmation, and blood pressure regulation determination. If the elastic discrimination channel is rejected due to DTW path anomalies or other quality gating, the current heartbeat enters the observation, template search range expansion, or recalibration state, and blood pressure regulation is not directly triggered by only the remaining area discrimination channel or wavelet time-frequency discrimination channel. When a channel exhibits feature loss, increased noise, unstable direction, or abnormal candidate index order, its quality decreases and it is automatically downweighted. Low quality does not directly increase the boundary offset, and low quality is not considered as evidence of blood pressure changes. If the quality of all channels is insufficient, the output signal is invalid and blood pressure regulation is stopped.

[0142] The quality of the optimal matching template for each currently valid heartbeat within the current control cycle is verified.

[0143] If the current monitoring period is continuous If a number of currently valid heartbeats that pass quality verification meet the continuous same-direction confirmation condition, then according to the continuous The system uses the current blood pressure boundary and offset direction of the current effective heartbeat to generate the blood pressure regulation command for the next control cycle; otherwise, it performs monitoring anomaly handling.

[0144] Obtain an initial set of stable heartbeats, and determine the template matching quality threshold based on the median and median absolute deviation of the minimum joint differences of each heartbeat in the initial set of stable heartbeats.

[0145] This embodiment distinguishes between template matching quality, overall signal quality, and boundary change amplitude. Template matching quality is used to determine whether the current waveform can be interpreted by existing template groups, while boundary change amplitude is used to determine whether a blood pressure change requiring adjustment has occurred. These two factors are not interchangeable. Let the initial set of stable heartbeats be non-empty and have a sufficient sample size to calculate the median absolute deviation. When the number of samples is insufficient, stable heartbeats should be re-collected or the device should be calibrated to a threshold. The template matching quality threshold... for: ; in, This represents the median; Indicates the absolute deviation of the median; To match the quality threshold coefficient. When This indicates that the current heartbeat cannot be reliably interpreted by the existing elastic boundary template group, and it is not allowed to trigger voltage regulation.

[0146] The quality threshold is determined based on the median and median absolute deviation of the overall quality score of each heartbeat in the initial stable heartbeat set.

[0147] The quality threshold is determined by subtracting the product of the quality threshold coefficient and the median absolute deviation from the median overall quality of the initial stable heartbeat set. When the optimal template is continuously located at the edge of the pressure index template group, the boundary offset exceeds the template coverage range, the minimum joint difference continuously exceeds the template matching quality threshold, or there is continuous conflict in the three-channel direction, the system expands the pressure index search range or re-executes the area and wavelet joint blood pressure boundary calibration, instead of forcibly absorbing the anomalies by adjusting the pressure or updating the template.

[0148] If all templates fail to achieve a reliable match, the overall quality remains below the quality threshold, the optimal template is continuously located at the edge of the template group, there are conflicts in the directions of each effective channel, or the back-calculation results exceed the template coverage range, then the quality verification is considered to have failed, and monitoring anomaly handling is performed, including: pausing pressure adjustment and freezing the working templates, not interpreting low quality as a change in blood pressure; re-performing the area and wavelet combined blood pressure boundary calibration, redetermining the target blood pressure boundary and reconstructing or expanding the elastic boundary template group, and returning to the control cycle loop after passing the consistency and quality checks.

[0149] when And the current effective cardiac comprehensive quality score Not lower than the quality threshold When this happens, perform the following operations: The basic dead zone is determined based on the median and median absolute deviation of the absolute values ​​of the boundary offsets of each heartbeat in the initial stable heartbeat set.

[0150] ; ; in, This is the signed boundary offset; Basic dead zone; It is the basic dead zone coefficient.

[0151] With continuous Each current effective heartbeat is used as a control cycle.

[0152] In the Within each control cycle, a current valid heartbeat that meets the following conditions is defined as a qualified heartbeat: ; ; Flexible discrimination channel The cuff pressure is within the preset stable range; no safety abnormality has been triggered. [The following appears to be a separate, unrelated sentence:] The cuff pressure is within the preset stable range; no safety abnormality has been triggered. The set of qualified heartbeats within a control cycle is denoted as .like If the voltage regulation command is not generated in this control cycle, continue to observe, supplement data acquisition, or recalibrate.

[0153] The dynamic dead zone threshold is determined by statistically analyzing the area characteristic variation, wavelet time-frequency characteristic variation, cuff pressure variation, cardiac interval variation, and body motion risk within the current control period, and combining this with the baseline dead zone.

[0154] exist The dynamic dead zone threshold is defined by incorporating internal statistical area characteristics, wavelet time-frequency characteristics, cuff pressure and cardiac interval variability, combined with body motion risk. for: ; in, , , and They are respectively Robust variability in inner area characteristics, wavelet time-frequency characteristics, cuff pressure, and RR interval or heartbeat interval can be measured using median absolute deviation or interquartile range and the same algorithm can be used in the same monitoring process. For the first The risk of bodily movement within a control cycle; , , , and The non-negative threshold coefficient preset or calibrated for the device.

[0155] For the first The set of qualified heartbeats for each control cycle Define the robust aggregate value of overall quality. for: ; in, It represents one of the median, truncated mean, or quality-weighted mean, and uses the same aggregation method in the same monitoring process.

[0156] The dynamic adjustment threshold is determined based on the dynamic dead zone threshold, the minimum interval between the dead zone and the adjustment threshold, and the overall quality of qualified heartbeats within the current control window; the dynamic adjustment threshold is defined as: ; in, For the first Dynamic adjustment threshold for each control cycle; The minimum interval between the preset or calibrated dead zone of the device and the adjustment threshold; This is the non-negative correction coefficient for the dynamic adjustment threshold based on the overall quality. Therefore, it always satisfies... .

[0157] A hysteresis relationship is established between the adjustment threshold and the return dead zone threshold. Only after re-entering the dynamic dead zone and maintaining a preset stable time after a pressure adjustment is a new locking boundary allowed to be confirmed. If two consecutive pressure adjustments are in opposite directions with small offsets, the system extends the observation window, reduces the control gain, or maintains the current pressure to avoid repeated actions near the threshold. When the blood pressure boundary offset exceeds the template coverage area, expanded search or recalibration is triggered first. Automatic deflation is mainly triggered by safety conditions such as pressure, time, distal perfusion, pain, leakage, and control abnormalities, and does not directly equate large changes in actual blood pressure themselves with deflation conditions.

[0158] Based on the template matching quality threshold, quality threshold, and dynamic adjustment threshold, the single beat confirmation indicator is calculated according to the minimum joint difference of the current valid heartbeat, the comprehensive quality score, the boundary offset, the validity of the elastic template, and the directional consistency of the elastic template difference, area difference, and wavelet time-frequency difference.

[0159] set up Indicates the first The control cycle to which the current effective heartbeat belongs.

[0160] ; in, For the first The single-beat confirmation indicator of a currently valid heartbeat. This is achieved when template matching quality, overall quality, offset amplitude, valid elastic discrimination channel, and at least two valid discrimination channels are aligned in direction. .

[0161] Based on the single-shot confirmation indicator, the consecutive same-direction confirmation rate is calculated. Only when the consecutive same-direction confirmation rate is not less than the consecutive confirmation threshold is the consecutive... The current blood pressure boundary and offset direction of the current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle.

[0162] ; in, This refers to the consecutive same-direction confirmation rate; The direction of pressure increase or decrease is yet to be confirmed; For the first The single beat confirmation indicator of a currently valid heartbeat; For the first The signed boundary offset of each currently valid heartbeat. Only when When the change is confirmed, it is determined to be valid and the action is taken. The direction in which the maximum value is obtained generates the cuff pressure adjustment command; Continuous confirmation threshold; when and When the same maximum value is obtained, the system enters observation mode and does not generate voltage regulation commands.

[0163] In this embodiment, four states are set: (1) stable state , and (2) Maintain the current cuff pressure; (3) Observe the template matching quality and overall quality to ensure they are qualified, but Or, if the channel direction has not yet formed a continuous consistency, record the offset, freeze the working template and continue to collect data; (3) Confirm the change status: , And if all quality conditions are met, start a limit voltage regulation; (4) Signal invalid, extended search or safe exit status: , If the elastic discrimination channel is invalid, the template index is continuously located at the coverage edge, or the three discrimination channels are in continuous conflict, do not start pressure regulation and enter observation, expand the search range, or recalibrate; when body movement, air leakage, overpressure, pain, low perfusion, maximum continuous occlusion time, pressure regulation oscillation, or monitoring timeout reach the safety conditions, stop tracking and safely vent the gas.

[0164] For the Multiple back-inferred blood pressure boundaries are gated by template matching quality, overall quality, and directional consistency within a single control cycle. By performing robust aggregation, we obtain the periodic back-inference boundary. for: ; Robust aggregation can use the median, truncated mean, or quality-weighted mean, and the same method should be used in the same monitoring process.

[0165] When the minimum joint difference is continuously greater than the template matching quality threshold, the comprehensive quality score is continuously lower than the quality threshold, the optimal template is continuously located at the edge of the template group, the directional consistency does not meet the requirements, or the back-inference result exceeds the template coverage range, the blood pressure regulation is paused and the working template is frozen. The area and wavelet joint blood pressure boundary calibration is re-executed to rebuild or expand the elastic boundary template group.

[0166] Target cuff pressure is updated using the following formula: ; in, and The first The and the first The target cuff pressure for each control cycle; The target relative pressure that was determined when entering dynamic monitoring and remains unchanged during this short-term monitoring period; The pressure control gain preset or calibrated for the equipment; The maximum single pressure adjustment amount set for the equipment safety strategy.

[0167] Define the limiting function as follows: ; in, The pressure control increment is to be limited. and These represent the lower and upper limits, respectively; in this formula, , .

[0168] set up This refers to the cuff pressure stability error limit, which is a pressure control parameter preset or calibrated by the equipment. When the cuff pressure meets the preset stability time... If the current target pressure is considered stable, or if an equivalent stability criterion is met, then the current target pressure is considered stable. The trend output and template update gating described later both call the same... No further "pressure error limits" will be redefined.

[0169] This embodiment also sets the minimum effective adjustment amount per cycle, the minimum operating time of the air pump or proportional valve, the stabilization waiting time after pressure regulation, the pressure control error limit, and the continuous reverse pressure regulation suppression condition.

[0170] Adjust the pressure according to the target pressure of the next control cycle, wait until the pressure stability condition is met, re-execute joint matching until the boundary offset returns to the dynamic dead zone, and calculate the blood pressure trend value.

[0171] In the same paragraph In short-term dynamic monitoring of blood pressure boundaries, the same target relative pressure is used for both initial locking and relocking in subsequent control cycles. .by Indicates that after initialization is complete... Stable locking cuff pressure under certain conditions, in order to Indicates the first After each control cycle of voltage adjustment, the voltage is readjusted at the same time. The cuff pressure at which a stable locked state is achieved under certain conditions can be used to express the short-term blood pressure trend as follows: ; in, For the first The control cycle relative to the initial boundary is the [number]th [cycle]. Blood pressure trend value; This is the initial boundary; and All are relative pressures at the same target The actual cuff pressure at which stable locking is achieved is used, rather than the new blood pressure boundary estimation parameter.

[0172] In addition to calculating the trend by changing the pressure of the locking cuff, it is also possible to do so in the first... Within a trend output window, the reliable single-beat back-inferred blood pressure boundary is calculated. Perform robust statistics. An example using the median is: ; in, Output the trend window number; For the first The set of currently valid heartbeats within a trend output window; For window level Blood pressure trend value; For the first Each heartbeat belongs to a control cycle; In relation to the aforementioned pressure stability conditions This is the error limit for pressure stability of the same cuff. Transient pressure changes, abnormal beats, and unstable heartbeats are not included in the trend statistics.

[0173] The output includes at least one of the following: short-term blood pressure trend, locked cuff pressure, boundary confidence, template matching quality, number of effective heartbeats, proportion of abnormal heartbeats, pressure control status, and safety status. For low quality, the corresponding weights are reduced, the template is frozen, and pressure is adjusted, with continued observation or recalibration. Tracking is stopped and automatic deflation is only initiated when overpressure, exceeding the maximum continuous occlusion time, severe body movement, cuff leakage, abnormal pressure control, pain feedback, or safety conditions indicating distal perfusion risk occur.

[0174] In this embodiment, the flexible boundary template set includes a baseline template set that remains unchanged during this short-term dynamic monitoring, and a working template set that is slowly updated only when the update gating condition is met. The baseline template is used to preserve the reference for changes relative to the initial state; the working template is used to adapt to slight contact or gradual morphological changes under stable conditions. The current heartbeat is mapped to the template time axis along the optimal DTW path at the update front, avoiding direct averaging of waveforms of different time lengths.

[0175] If the optimal template is continuously located at the edge of the template group, or the minimum joint difference continuously exceeds the matching threshold, this embodiment considers the boundary to be outside the template coverage area, the cuff contact state to have changed, or the physiological state to have changed significantly. In this case, the working template is not continued to be iterated, but the pressure search range is expanded and the area and wavelet combined blood pressure boundary calibration is re-executed.

[0176] The area-wavelet joint recalibration can be performed using a full pressure scan or a local pressure scan covering the neighborhood of the current candidate boundary. The system re-extracts area features and wavelet time-frequency features, performs weighted fusion based on historical reliability and current quality, redetermines the target blood pressure boundary, and reconstructs or expands the elastic boundary template group; short-term quasi-continuous blood pressure trend monitoring is only restored after template sequence consistency, matching quality, overall signal quality, and safety status have all passed the checks.

[0177] In the After the state determination is completed in the first control cycle, if no voltage adjustment is needed and the boundary remains stable, or if one voltage adjustment has been completed and the circuit has been relocked, then upon entering the second control cycle... Perform a work template update gating check before each control cycle. Only gating flags are present. Work templates are only updated occasionally; The working template should remain unchanged. Updating the working template is prohibited during observation, confirmation of changes, voltage regulation transients, continuous low-quality conditions, and safety anomalies. The baseline template set should remain unchanged during this short-term dynamic monitoring period.

[0178] ; in, For the first At the end of each control cycle, the work template updates the gate control flag; The aforementioned robust aggregate value for overall cardiac quality is considered acceptable. For the first A qualified heartbeat during each control cycle The robust aggregation value; The acceptable cardiac boundary offset for this control cycle The robust aggregation value; To ensure robust actual cuff pressure during the relocking phase; The current target cuff pressure; The same pressure stability error limit as mentioned above; To control the risk of bodily movement during the current control period; The maximum permissible risk of physical movement is preset for the equipment; This indicates that the pressure, time, perfusion, pain, leakage, and control status are all within acceptable ranges.

[0179] When updates are allowed, the updatable numerical content is defined as a template numerical vector formed by concatenating normalized waveforms, non-normalized waveforms, boundary-sensitive feature bands, area vectors, and wavelet vectors. The working template is updated class by class of numerical components only when the overall quality is qualified, the minimum joint difference does not exceed the template matching quality threshold, the boundary offset returns to the dynamic dead zone, the cuff pressure is stable, the body movement risk is low, and the safety status is normal. ; in, and These are the working template numerical vectors before and after the update, respectively; This is a representative numerical vector of several recent effective heartbeats after elastic registration; Update the step size for the template and ; Maximum template update step size; relative pressure index, blood pressure boundary type, and quality or safety metadata are not included in linear averaging; template updates are frozen during observation, confirmed changes, blood pressure regulation, or low-quality states. Updates are performed item by item according to the same numerical component; relative pressure index. Blood pressure borderline types Quality indicators and security metadata are not included in the linear average.

[0180] For atrial fibrillation, premature beats, or arrhythmias, stratification is performed according to RR interval or cardiac cycle, and elastic template matching is performed between heartbeats in the same stratum; robust aggregation with sliding window, deweighting of abnormal beats, and continuous same-direction confirmation are adopted. A single irregular heartbeat is not directly interpreted as a change in blood pressure boundary, nor does it trigger blood pressure regulation or template updates on its own; if sufficient qualified heartbeats cannot be obtained in any stratum, the output quality is insufficient and the case is entered into observation or area-wavelet joint recalibration.

[0181] For premature beats and their compensatory pauses, the premature beat and the immediately following recovery beat may not participate in pressure regulation triggering. For body movement, the body movement window is identified by combining acceleration, pressure baseline abrupt changes, oscillatory wave high-frequency energy, and template matching failure. During body movement, the template is not updated, the cuff pressure is not adjusted, and it is maintained, released, or remeasured according to the level of risk.

[0182] Because systolic blood pressure boundary monitoring may approach a state of blood flow occlusion, this invention is limited to short-term quasi-continuous monitoring and is not used for prolonged continuous occlusion. A single monitoring session is 10 seconds to 5 minutes, preferably 30 seconds to 2 minutes. The system is configured with maximum cuff pressure, maximum single pressure adjustment, maximum continuous occlusion time, pressure release interval, distal perfusion monitoring, and user pain feedback. When a safety condition is triggered, safety logic takes precedence over all algorithm outputs, and the system immediately stops pressure adjustment and performs safe deflation.

[0183] In this embodiment, the sampling frequency of pressure and oscillation waves can be from 50Hz to 2000Hz, preferably from 200Hz to 1000Hz; the example range of the continuous wavelet scale set is 10 to 200, and the specific scale or equivalent frequency boundary is determined according to the sampling rate and beat duration, and remains consistent in the same monitoring process. The elastic template group can contain 3 to 11 relative pressure states; the pressure interval between adjacent states can be 1 mmHg to 5 mmHg; each pressure state can be represented by 2 to 8 valid heartbeats. The historical reliability of direct matching, linear resampling, and DTW similarity can be obtained by device calibration or the initial stabilization window, and the current quality is updated beat by beat; the number of consecutive same-direction offset confirmed heartbeats can be 2 to 6, and the consecutive confirmation threshold can be 0.5 to 1; the single pressure adjustment amount can be 0.5 mmHg to 10 mmHg, and is constrained by the maximum pressure, the maximum continuous occlusion time, and the stabilization time after pressure adjustment; a single short-term dynamic monitoring can be 10 seconds to 5 minutes, preferably 30 seconds to 2 minutes. The above values ​​are within the scope of the embodiments and do not constitute the sole limitation on the scope of protection.

[0184] This invention can be implemented using a microcontroller, digital signal processor, general-purpose processor, field-programmable gate array, or a combination thereof. Template matching can be performed at the device end, or assisted by a local terminal or edge computing unit, provided that the requirements for safe real-time control are met. Pressure control and safe venting must be performed by a local controller capable of independent execution in the event of communication interruption.

[0185] In arm-type electronic blood pressure monitors, the cuff, miniature air pump, proportional valve or deflation valve, pressure sensor, and local processor form a complete closed loop; in wrist-type devices, the control gain, pressure interval between adjacent templates, and safety upper limit are recalibrated based on the local pressurization chamber volume and airway response; in bedside or anesthesia monitoring equipment, ECG R waves, photoplethysmography pulse waves, acceleration, and distal perfusion signals can be used to improve heart rate segmentation and safety gating; in ambulatory blood pressure monitoring or home devices, the short-term dynamic monitoring phase is initiated as needed after a routine cuff measurement and is fully deflated after the preset monitoring duration is reached.

[0186] Example 2: This embodiment presents a short-time dynamic monitoring system based on area and wavelet time-frequency boundary locking. The system includes: A cuff or localized pressure unit is used to apply adjustable pressure to the subject.

[0187] The cuff pressure generation and adjustment unit is connected to the cuff or local pressurization unit through an air path and is used to control the cuff or local pressurization unit to generate or release pressure according to the drive signal.

[0188] Pressure sensors are used to collect real-time pressure signals within the cuff or localized pressure unit.

[0189] The signal acquisition and synchronization module is used to transmit real-time pressure signals to the processor.

[0190] The processor is used to perform oscillatory wave preprocessing, heartbeat segmentation, initial blood pressure boundary calibration, elastic boundary template group establishment, elastic template matching, area difference calculation, wavelet time-frequency difference calculation, joint difference and boundary back-inference, threshold and continuity confirmation judgment, short-term trend calculation, and template updating and recalibration.

[0191] The recalibration is a joint recalibration of the blood pressure boundary using area and wavelet, which includes re-extracting area features and wavelet time-frequency features, redetermining the target blood pressure boundary, and reconstructing or expanding the elastic boundary template group.

[0192] The pressure controller is used to convert the pressure adjustment command into a drive signal and send it to the cuff pressure generation and adjustment unit.

[0193] The quality control and safety protection module is used to monitor signal quality and perform exception handling when an anomaly is detected.

[0194] The quality control and safety protection module is used to enable, freeze, or stop template updates, trend outputs, and pressure regulation based on the current signal quality, body movement, pressure status, remote perfusion, and safety conditions.

[0195] The memory is used to store real-time pressure signals, effective heartbeats and pressure values, joint features, elastic boundary template groups, blood pressure boundary values, offset directions, and pressure regulation commands. The display or communication output module is used to display the effective heart rate and pressure value, elastic boundary template group, blood pressure boundary value, and offset direction in real time.

[0196] In this embodiment, the local pressure unit includes a cuff and other pressure structures capable of applying controllable external pressure to the local tissue where the target artery is located. The cuff or local pressure unit is connected to the cuff pressure generation and adjustment unit and a pressure sensor, respectively. The output of the pressure sensor is connected to a signal acquisition and synchronization module, which is connected to a processor. The processor is connected to a memory, a quality control and safety protection module, and a display or communication output module, and outputs pressure adjustment commands to the pressure controller. The pressure controller is connected to the cuff pressure generation and adjustment unit and is used to perform pressure scanning, constant or quasi-constant pressure maintenance, limited pressure increase, limited pressure decrease, and safe deflation.

[0197] In this embodiment, the cuff pressure generation and adjustment unit includes a micro air pump, an air storage chamber, a proportional valve, a vent valve, a one-way valve, a throttle orifice, a pressure servo mechanism, or a combination thereof; the pressure servo mechanism is used to perform pressure scanning, constant or quasi-constant pressure maintenance, limited pressure increase, limited pressure decrease, and safe venting, and waits for the cuff pressure to enter a preset stable range after each pressure adjustment.

[0198] In this embodiment, the pressure controller is set with a single pressure adjustment range of 0.5 to 10 mmHg, a single minimum effective adjustment amount, a stabilization waiting time after pressure adjustment, a pressure hysteresis interval, and a continuous reverse pressure adjustment suppression condition; it only performs a limited pressure adjustment once when the continuous confirmation rate reaches the continuous confirmation threshold, freezes the template update and trend output during the pressure adjustment transient, and recalculates the minimum joint difference and boundary offset after the pressure stabilizes.

[0199] In this embodiment, the system is integrated into an arm-type electronic blood pressure monitor, a wrist-type electronic blood pressure monitor, an ambulatory blood pressure monitoring device, a bedside electrocardiogram monitor, an anesthesia monitor, an emergency monitoring device, a home blood pressure monitoring device, or a wearable short-term blood pressure trend monitoring device.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the present invention.

Claims

1. A short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking, characterized in that, This method includes the following steps: The raw cuff pressure signal of a single pressure scan is acquired and preprocessed to extract the oscillation wave signal. Based on the oscillation wave signal, several effective heartbeats and the baseline pressure corresponding to each effective heartbeat are obtained. The area features and wavelet time-frequency features of each effective heartbeat are extracted and weighted and fused to determine the initial blood pressure boundary value. Based on the initial blood pressure boundary value, an elastic boundary template group with relative pressure index is established. Select the dynamic monitoring mode, and use the initial blood pressure boundary value as the current reference blood pressure boundary to enter the control cycle loop. Perform the following operations in each control cycle: Based on the current reference blood pressure boundary, determine the current target relative pressure, adjust the cuff pressure applied during the current control cycle to the target relative pressure, wait for the preset stabilization time, acquire the original cuff pressure signal and preprocess it to obtain the effective heartbeat and the pressure value of the current effective heartbeat; Feature extraction is performed on the current effective heartbeat to obtain the joint features of the current effective heartbeat, and the current effective heartbeat is jointly matched with the elastic boundary template group to obtain the optimal matching template of the current effective heartbeat; Based on the relative pressure index of the optimal matching template, the relative pressure state corresponding to the current effective heartbeat is determined, and based on the pressure value of the current effective heartbeat, the relative pressure state and the relative pressure index of the optimal matching template, the current blood pressure boundary and offset direction are inferred. Perform quality verification on the optimal matching template for each currently valid heartbeat within the current control period; if continuous heartbeats within the current monitoring period If a number of currently valid heartbeats that pass quality verification meet the continuous same-direction confirmation condition, then according to the continuous The current blood pressure boundary and offset direction of the current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle; Conversely, if the anomaly is detected, then abnormal handling will be performed. Adjust the blood pressure according to the adjustment command, calculate and output the blood pressure trend value, use the blood pressure boundary locked after adjustment as the current reference blood pressure boundary, and repeat the above operation until the preset single monitoring time is reached.

2. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 1, characterized in that, The process of acquiring and preprocessing the raw pressure signal from a single pressure scan to obtain several effective heartbeats and the baseline pressure for each effective heartbeat includes: Acquire the raw cuff pressure signal from a single boost or depressurization scan; The original cuff pressure signal is baseline separated and filtered to obtain the pressure baseline and oscillation wave component; The oscillating wave component is divided into several heartbeats, and the baseline pressure of each heartbeat is determined based on the pressure baseline. Each heartbeat is evaluated for quality, and those that fail the evaluation are removed to obtain a number of valid heartbeats. The normalized and non-normalized waveforms of each valid heartbeat are saved.

3. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 2, characterized in that, The process involves extracting the area features and wavelet time-frequency features of each effective heartbeat, performing weighted fusion to determine the initial blood pressure boundary value, and establishing an elastic boundary template group with a relative pressure index based on the initial blood pressure boundary value, including: Based on the baseline pressure of all effective heartbeats, the effective heartbeats are sorted to obtain a pressure sorting sequence; For each valid heartbeat in the pressure sorting sequence, the area feature is extracted, and the change in area feature between adjacent valid heartbeats is calculated; For each valid heartbeat in the stress sorting sequence, wavelet time-frequency features are extracted, and the changes in wavelet time-frequency features between adjacent valid heartbeats are calculated. The changes in area characteristics between adjacent valid heartbeats are robustly standardized, and the weights of area characteristics are determined according to their historical reliability and current quality. An initial area score is then calculated. Robustly standardize the changes in wavelet time-frequency features between adjacent valid heartbeats, determine the weights of wavelet time-frequency features based on their historical reliability and current quality, and calculate the initial wavelet score; The initial area and wavelet joint scores are weighted and summed to obtain the initial area and wavelet joint scores. The location where the initial area wavelet joint score transitions or peaks is determined along the pressure sorting sequence, and the baseline pressure of the effective heartbeat corresponding to the location is used as the initial blood pressure boundary value. A template consistency check is performed on all valid heartbeats. After the template consistency check is passed, effective single-beat oscillation waves in the boundary state, the low pressure side of the benchmark, the high pressure side of the benchmark, and the preset pressure state are selected from the oscillation waves of all valid heartbeats, based on the initial blood pressure boundary value. A template is created for each selected effective single-stroke oscillation wave to store the actual cuff pressure, oscillation wave waveform, area characteristics, wavelet time-frequency characteristics, and quality characteristics corresponding to that effective single-stroke oscillation wave; Calculate the relative pressure index of each template with respect to the initial blood pressure boundary value, arrange all templates in ascending order of relative pressure index, and establish an elastic boundary template group.

4. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 3, characterized in that, The dynamic monitoring mode includes: short-term dynamic monitoring of systolic blood pressure and short-term dynamic monitoring of diastolic blood pressure; the target relative pressure for dynamic monitoring is: the boundary state of the current target blood pressure boundary or the preset safe side relative pressure within the template coverage area.

5. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 4, characterized in that, The step of jointly matching the current effective heartbeat with the elastic boundary template group to obtain the optimal matching template for the current effective heartbeat includes: Traverse each template in the elastic boundary template group and calculate the amplitude difference between the current effective heartbeat and the currently traversed template; The normalized and non-normalized waveforms of the current valid heartbeats are directly matched, linearly resampled, and dynamically time-warped matched with the current traversal template, respectively. The similarity of the normalized waveforms under the three matching methods, the similarity of the non-normalized waveforms under the three matching methods, and the cumulative distance of the dynamic time warping are obtained. The similarity of the normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the normalized waveform. The similarity of the non-normalized waveform under the three matching methods is weighted and fused to obtain the fused similarity of the non-normalized waveform. The fusion similarity of the normalized waveform and the fusion similarity of the non-normalized waveform are weighted and fused to obtain the elastic similarity of the dual waveform fusion. The cumulative distance of dynamic time warping matching is normalized according to the optimal warping path length to obtain the normalized dynamic time warping distance. After mapping the normalized dynamic time warping distance to the normalized waveform and the non-normalized waveform respectively, a weighted fusion is performed to obtain the dual-waveform fused dynamic time warping distance. After normalizing the dynamic time warping distance, elastic similarity, and amplitude difference of the dual waveform fusion, a weighted fusion is performed to obtain the elastic template difference between the current effective heartbeat and the current traversed template. Calculate the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template, respectively; We perform a weighted summation of the elastic template difference, area difference, and wavelet time-frequency difference to obtain the joint difference between the current valid heartbeat and the current traversed template. The template with the smallest joint difference is selected as the optimal matching template for the current valid heartbeat.

6. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 5, characterized in that, The calculation of the area difference and wavelet time-frequency difference between the current valid heartbeat and the current traversed template includes: Extract the area features of the current valid heartbeat and several area features of the current traversed template respectively; The differences between the current valid heartbeat and the area features of the current traversed template are calculated and weighted and fused to obtain the area difference between the current valid heartbeat and the current traversed template. Extract the wavelet time-frequency features of the current valid heartbeat and the current traversed template respectively; The differences between the current valid heartbeat and the wavelet time-frequency features of the current traversed template are calculated and weighted and fused to obtain the wavelet time-frequency difference between the current valid heartbeat and the current traversed template.

7. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 6, characterized in that, The weighted summation of elastic template differences, area differences, and wavelet time-frequency differences yields the joint difference between the current valid heartbeat and the current traversed template, including: The elastic template difference, area difference, and wavelet time-frequency difference are respectively mapped to the same dimension to obtain the elastic template mapping difference, area mapping difference, and wavelet time-frequency mapping difference; The weights of elastic template differences, area differences, and wavelet time-frequency differences are determined based on historical reliability and current quality. The elastic template mapping difference, area mapping difference, and wavelet time-frequency mapping difference are weighted and summed according to their respective weights, and a direction conflict penalty term is introduced to obtain the joint difference between the current effective heartbeat and the current traversed template. The method for calculating the direction conflict penalty term is as follows: determine the candidate template direction of the current traversal template; determine the independent directions of elastic template difference, area difference, and wavelet time-frequency difference respectively; calculate the proportion of each independent direction that is inconsistent with the candidate template direction to obtain the direction inconsistency ratio; calculate the degree of cross-level jump of the template index of the current traversal template relative to the optimal template index of the previous effective heartbeat to obtain the index jump penalty term; calculate the dynamic time warping path anomaly degree of the current traversal template to obtain the path anomaly penalty term; and sum the direction inconsistency ratio, index jump penalty term, and path anomaly penalty term by weight to obtain the direction conflict penalty term of the current traversal template.

8. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 7, characterized in that, If the current monitoring period is continuous If a number of currently valid heartbeats that pass quality verification meet the continuous same-direction confirmation condition, then according to the continuous The current blood pressure boundary and offset direction of each current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle, including: Obtain an initial stable heartbeat set, and determine the template matching quality threshold based on the median and median absolute deviation of the minimum joint differences of each heartbeat in the initial stable heartbeat set; The quality threshold is determined based on the median and median absolute deviation of the overall quality score of each heartbeat in the initial stable heartbeat set; If the minimum joint difference of the current effective heartbeat is greater than the template matching quality threshold, the current effective heartbeat does not trigger blood pressure regulation; if the minimum joint difference of the current effective heartbeat is not greater than the template matching quality threshold, and the overall quality score of the current effective heartbeat is not lower than the quality threshold, then the following operations are performed: The basic dead zone is determined based on the median and median absolute deviation of the absolute values ​​of the boundary offsets of each heartbeat in the initial stable heartbeat set. With continuous One current effective heartbeat is used as a control cycle; A valid heartbeat is defined as a qualified heartbeat if it meets the following conditions: the minimum joint difference of the current valid heartbeat is greater than the template matching quality threshold; the comprehensive quality score of the current valid heartbeat is not lower than the quality threshold; there is a valid discrimination channel; the cuff pressure is within the preset stable range; no safety anomaly is triggered; if there is no qualified heartbeat in the current control cycle, pressure adjustment is not triggered. The area feature variation, wavelet time-frequency feature variation, cuff pressure variation, cardiac interval variation, and body motion risk within the current control period are statistically analyzed, and the dynamic dead zone threshold is determined in combination with the baseline dead zone. The dynamic adjustment threshold is determined based on the dynamic dead zone threshold, the minimum interval between the dead zone and the adjustment threshold, and the overall quality of qualified heartbeats within the current control window. Based on the template matching quality threshold, quality threshold, and dynamic adjustment threshold, the single beat confirmation indicator is calculated according to the minimum joint difference of the current valid heartbeat, the comprehensive quality score, the boundary offset, the validity of the elastic template, and the directional consistency of the elastic template difference, area difference, and wavelet time-frequency difference. Based on the single-shot confirmation indicator, the consecutive same-direction confirmation rate is calculated. Only when the consecutive same-direction confirmation rate is not less than the consecutive confirmation threshold is the consecutive confirmation rate determined. The current blood pressure boundary and offset direction of the current effective heartbeat are used to generate the blood pressure regulation command for the next control cycle.

9. The short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking according to claim 8, characterized in that, The step of adjusting blood pressure according to the pressure adjustment command and calculating the blood pressure trend value includes: Calculate continuous The target pressure for the next control period is obtained by summing the robust aggregate value of the current blood pressure boundary of the current effective heartbeat with the target relative pressure of the current monitored blood pressure boundary, and then subtracting the target pressure of the current control period. Adjust the pressure according to the target pressure of the next control cycle, wait until the pressure stability condition is met, re-execute joint matching until the boundary offset returns to the dynamic dead zone, and calculate the blood pressure trend value.

10. A short-time dynamic monitoring system based on area and wavelet time-frequency boundary locking, characterized in that, The system includes: Cuffs or localized pressure units are used to apply adjustable pressure to the subject; The cuff pressure generating and regulating unit is connected to the cuff or local pressurization unit through an air path, and is used to control the cuff or local pressurization unit to generate or release pressure according to the drive signal; Pressure sensors are used to collect real-time pressure signals within the cuff or localized pressurization unit; The signal acquisition and synchronization module is used to transmit real-time pressure signals to the processor; The processor is configured to execute the short-time dynamic monitoring method based on area and wavelet time-frequency boundary locking as described in any one of claims 1-9, and output voltage regulation commands; The pressure controller is used to convert the pressure adjustment command into a drive signal and send it to the cuff pressure generation and adjustment unit; The quality control and safety protection module is used to monitor signal quality and perform exception handling when an anomaly is detected. The memory is used to store real-time pressure signals, effective heartbeats and pressure values, joint features, elastic boundary template groups, blood pressure boundary values, offset directions, and pressure regulation commands. The display or communication output module is used to display the effective heart rate and pressure value, elastic boundary template group, blood pressure boundary value, and offset direction in real time.

Images