Intelligent dynamic blood pressure and pulse rate monitoring method and system integrating multiple measurement methods

Abstract

The invention relates to an intelligent dynamic blood pressure and pulse rate monitoring method and system integrating multiple measurement methods, and the method comprises the steps: obtaining identity verification information inputted by a user, generating an encrypted user identifier, calling historical health data of the user, and generating a personalized monitoring scheme in combination with a preset instruction; collecting user motion state data in real time according to the body position parameters, and dynamically calibrating a cuff inflation pressure threshold value and a signal sampling frequency to obtain calibration configuration parameters; calling a corresponding sensor module to execute blood pressure and pulse rate measurement to obtain a blood pressure value collected by each blood pressure measurement method; when the standard deviation of the blood pressure values output by each blood pressure measuring method does not exceed a preset difference threshold value, outputting a weighted average value as effective blood pressure data; the effective blood pressure data and the user historical health data are merged and input into a neural network model to generate a blood pressure trend prediction curve, and the blood pressure trend prediction curve is encrypted based on the encrypted user identifier and then distributed to the user terminal equipment. According to the invention, the accuracy of measured data is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent medical device technology, and in particular to an intelligent dynamic blood pressure and pulse rate monitoring method and system that integrates multiple measurement methods. Background Technology

[0002] With the accelerating aging of society and the increasingly fast pace of life, the incidence of chronic diseases such as hypertension is rising year by year, becoming one of the major public health problems threatening human health. Traditional blood pressure monitoring methods mostly rely on professional equipment in medical institutions or single-function electronic blood pressure monitors used at home. Their application scenarios are limited, and their operation is inconvenient, making it difficult to meet the needs of long-term, continuous, and accurate monitoring.

[0003] In recent years, with the development of mobile internet, the Internet of Things, artificial intelligence and wearable devices, remote health monitoring systems have gradually become a research hotspot, especially showing broad application prospects in the fields of family health management, chronic disease management and postoperative rehabilitation.

[0004] However, most blood pressure monitoring devices currently lack the ability to effectively adapt to individual differences. They typically use fixed measurement parameters for uniform processing, failing to fully consider the impact of different users' physiological characteristics, behavioral habits, and pathological conditions on the measurement results, thus reducing the accuracy and reliability of the data. For example, during actual wear, users may be in various different body positions (such as standing, sitting, lying down, etc.), and these changes in posture can significantly affect the distribution of vascular pressure, thereby interfering with the accuracy of blood pressure readings and weakening the practical reference value of blood pressure data. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides an intelligent dynamic blood pressure and pulse rate monitoring method and system that integrates multiple measurement methods.

[0006] Firstly, this application provides an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods, employing the following technical solution: A smart dynamic blood pressure pulse rate monitoring method integrating multiple measurement methods, the smart dynamic blood pressure pulse rate monitoring method includes: Obtain the authentication information input by the user, generate an encrypted user identifier based on the authentication information, and retrieve the user's historical health data; A personalized monitoring plan is generated based on the user's historical health data and the doctor's preset instructions; wherein, the personalized monitoring plan includes a combination of body position parameters and blood pressure measurement methods; Based on the body position parameters, the user's movement status data is collected in real time, and the cuff inflation pressure threshold and signal sampling frequency are dynamically calibrated to obtain calibration configuration parameters. Based on the blood pressure measurement method combination and calibration configuration parameters, the corresponding sensor module is called to perform blood pressure and pulse rate measurement to obtain the blood pressure value collected by each blood pressure measurement method; Determine whether the standard deviation of the blood pressure value output by each blood pressure measurement method exceeds the preset difference threshold; if so, trigger a retest and mark it as abnormal; if not, output the weighted average as the valid blood pressure data. The effective blood pressure data is combined with the user's historical health data and input into a neural network model to generate a blood pressure trend prediction curve. The blood pressure trend prediction curve is encrypted based on the encrypted user identifier and distributed to user terminal devices according to preset permission rules.

[0007] By adopting the above technical solutions and integrating multiple measurement methods with intelligent dynamic adjustment mechanisms, a fundamental shift from passive, static measurement to proactive, dynamic monitoring has been achieved, improving data accuracy and user experience. The system generates personalized monitoring plans based on user authentication and historical health data, and dynamically calibrates measurement parameters in conjunction with real-time motion status, effectively reducing the impact of environmental interference on results. The collaborative work and intelligent fusion judgment of multiple blood pressure measurement methods further improve measurement reliability. The automatic retesting and marking mechanism for abnormal data enhances the system's robustness. The integration of neural network models enables blood pressure trend prediction, providing strong support for disease early warning and health management. End-to-end data encryption and access control ensure user privacy and information security, demonstrating promising clinical application prospects and widespread application value.

[0008] Secondly, this application provides an intelligent dynamic blood pressure and pulse rate monitoring system integrating multiple measurement methods, employing the following technical solution: An intelligent dynamic blood pressure and pulse rate monitoring system integrating multiple measurement methods, the monitoring system includes: The user authentication and data acquisition module is used to acquire the authentication information input by the user, generate an encrypted user identifier based on the authentication information, and retrieve the user's historical health data. A personalized generation module is used to generate a personalized monitoring plan based on the user's historical health data and the doctor's preset instructions; wherein, the personalized monitoring plan includes a combination of body position parameters and blood pressure measurement methods; The parameter calibration module is used to collect user motion state data in real time based on the body position parameters, dynamically calibrate the cuff inflation pressure threshold and signal sampling frequency, and obtain calibration configuration parameters. The multi-mode blood pressure measurement module is used to call the corresponding sensor module to perform blood pressure and pulse rate measurement according to the combination of blood pressure measurement methods and calibration configuration parameters, so as to obtain the blood pressure value collected by each blood pressure measurement method. The measurement result verification module is used to determine whether the standard deviation of the blood pressure value output by each blood pressure measurement method exceeds the preset difference threshold; if so, it triggers a retest and marks the abnormality; if not, it outputs the weighted average as the valid blood pressure data. The blood pressure trend prediction and analysis module is used to merge the effective blood pressure data with the user's historical health data and input the data into a neural network model to generate a blood pressure trend prediction curve. An encrypted distribution module is used to encrypt the blood pressure trend prediction curve based on the encrypted user identifier and distribute it to the user terminal device according to preset permission rules.

[0009] Thirdly, this application provides a computer device, which adopts the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to perform the steps of the method as described in the first aspect.

[0010] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as in any of the methods in the first aspect. Attached Figure Description

[0011] Figure 1 This is a first flowchart of an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods according to one embodiment of this application.

[0012] Figure 2 This is a second flowchart illustrating an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods, according to one embodiment of this application.

[0013] Figure 3 This is a schematic diagram of the third process of an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods according to one embodiment of this application.

[0014] Figure 4 This is a schematic diagram of the fourth process of an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods according to one embodiment of this application.

[0015] Figure 5 This is a schematic diagram of the fifth step of the intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods according to one embodiment of this application.

[0016] Figure 6 This is a schematic diagram of the sixth process of an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods according to one embodiment of this application. Detailed Implementation

[0017] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-6 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0018] This application discloses an intelligent dynamic blood pressure and pulse rate monitoring method that integrates multiple measurement methods.

[0019] Reference Figure 1 A smart dynamic blood pressure pulse rate monitoring method integrating multiple measurement methods, comprising: Step S101: Obtain the authentication information input by the user, generate an encrypted user identifier based on the authentication information, and retrieve the user's historical health data; The identity verification information includes, but is not limited to, ID card numbers, social security card NFC reading signals, or biometric data such as fingerprints and facial recognition. These diverse identity credentials not only enhance system security but also improve user convenience. Subsequently, a unique "encrypted user identifier" is generated by hashing or using other cryptographic methods on multiple identity elements. This identifier uniquely corresponds to a specific user throughout the monitoring period, ensuring that all subsequent operations revolve around the same entity and avoiding data mismatch issues caused by identity confusion.

[0020] Furthermore, after obtaining the encrypted user identifier, the system sends a request to the cloud database to retrieve the user's accumulated historical health dataset, such as historical blood pressure records, medication history, basal metabolic rate indicators, chronic disease history, and other relevant medical information. This type of data constitutes the knowledge support layer of the entire monitoring process, enabling subsequent steps to make more targeted adjustments based on a true and reliable personal background.

[0021] Step S102: Generate a personalized monitoring plan based on the user's historical health data and the doctor's preset instructions; wherein, the personalized monitoring plan includes a combination of body position parameters and blood pressure measurement methods; Specifically, a personalized monitoring plan is customized based on the physiological condition and clinical needs of different users. The personalized monitoring plan does not refer to the general regular measurement arrangement, but rather a refined scheduling strategy that takes into account factors such as the severity of the patient's condition (e.g., hypertension classification), lifestyle habits (e.g., work and rest patterns, activity intensity), and drug response characteristics.

[0022] In this context, doctors' pre-set instructions typically involve setting different measurement frequencies and timings for different levels of hypertension. For example, high-risk individuals with stage III hypertension may require high-frequency blood pressure checks, such as once an hour, to detect potential fluctuations due to their significantly increased cardiovascular risk; while for patients with mild hypertension (stage I), the frequency can be appropriately reduced to three times a day to balance the conflict between accuracy and comfort.

[0023] In addition, the plan also specifies the body position parameters, that is, to determine the posture in which the subject should be placed before the measurement to complete the sampling work. Common options include sitting, lying down and flat, because changes in body posture will affect the pressure distribution inside and outside the blood vessels and thus affect the accuracy of the measurement.

[0024] Another key component is the combination of blood pressure measurement methods. Currently, the mainstream non-invasive blood pressure measurement technologies mainly include three forms: oscillometric, electronic Korotkoff sound, and pulse wave transit time. Each type has its own applicable scenarios and technical limitations; therefore, a reasonable combination can improve overall robustness while ensuring accuracy. For example, the oscillometric method is suitable for rapid screening but is susceptible to interference, the electronic Korotkoff sound method is more suitable for the precise diagnostic stage, and the pulse wave method is often used in portable wearable devices. This combined measurement architecture can effectively cope with uncertainties in complex environments.

[0025] Step S103: Collect user movement status data in real time based on body position parameters, dynamically calibrate the cuff inflation pressure threshold and signal sampling frequency, and obtain calibration configuration parameters; This step introduces an important adaptive adjustment mechanism: the gyroscope sensor in the inertial navigation unit detects changes in the user's limb orientation in real time and adjusts the operating parameters of the core hardware components accordingly. This solution overcomes the rigidity of traditional fixed threshold settings by employing a more flexible feedback control system to optimize performance.

[0026] Specifically, the system first stores a set of standard reference angles as the ideal body position baseline. Then, it continuously collects the current actual posture angle using sensors and calculates the deviation between the two. This value reflects the degree to which the user's body deviates from the optimal position, and a larger deviation often means a higher possibility of introducing error. Therefore, the system adds a certain proportional coefficient to the original basic pressure setting and multiplies it by this deviation to obtain a new "calibration pressure threshold." This design logic essentially simulates the process of human intervention; when it detects that the user's posture is incorrect, it automatically increases the pressure to compensate for the signal attenuation effect caused by external disturbances.

[0027] At the same time, in order to match the updated changes in physical conditions, the sampling rate of audio sensors such as microphones also needs to be adjusted synchronously to better capture any subtle sound features that may appear. This adjustment, known as the "signal sampling frequency," also relies on the aforementioned body position discrimination results, together forming a complete dynamic correction framework.

[0028] Step S104: Based on the combination of blood pressure measurement methods and calibration configuration parameters, call the corresponding sensor module to perform blood pressure and pulse rate measurement to obtain the blood pressure value collected by each blood pressure measurement method; The system sequentially activates various specialized detectors according to a pre-planned measurement combination and applies calibrated configuration parameters to perform on-site sampling. The entire process follows a strict time sequence and is uniformly scheduled and commanded by a central controller, ensuring that each action is completed under optimal conditions.

[0029] For example, if the oscilloscope module is used in this measurement, the controller will drive the micro air pump to inflate the cuff according to the newly calculated pressure threshold until it reaches the predetermined upper limit, and then gradually depressurize and release the gas. During this period, the pressure sensor continuously monitors the change trajectory of the internal oscillation waveform and converts it into a digital signal to upload to the analysis engine. Meanwhile, if the electronic Korotkoff sound method is also used, the built-in microphone will be turned on at the corresponding time point to listen for arterial murmurs and the useful components will be enhanced and noise interference will be eliminated through the filtering and amplification circuit. Then, a specially developed phase extraction algorithm will be used to identify the two critical points of systolic and diastolic blood pressure. As for the part involving pulse wave conduction time, it mainly relies on a flexible photoplethysmography device attached to the wrist or other parts of the body to capture changes in blood flow velocity caused by heartbeat, and combine it with a pre-trained machine learning model to calculate the corresponding blood pressure estimate.

[0030] Step S105: Determine whether the standard deviation of the blood pressure value output by each blood pressure measurement method exceeds the preset difference threshold; if yes, proceed to step S106; if no, proceed to step S107. Step S106: Trigger a retest and mark the exception label; Step S107: Output the weighted average as valid blood pressure data; The system employs statistical variance testing to measure the consistency of results from different sources. Specifically, it first calculates the dispersion index, or standard deviation σ, among all blood pressure estimates used for comparison, and then compares it with a pre-agreed tolerance boundary. If σ is found to be below a certain accepted limit (set here as 10 mmHg), it indicates that the various measurements are closely matched, and their weighted average can be safely adopted as the final output. The weighting can be dynamically determined based on the reliability score of each method. Conversely, if the σ exceeds this tolerance range, it indicates that at least one or more data points have significantly deviated from the norm, which is highly likely to be influenced by uncontrollable external variables, leading to distortion. In this case, the current process needs to be immediately interrupted and the system switched to troubleshooting mode. On the one hand, the main control chip is notified to re-execute a complete measurement procedure; on the other hand, the original sample is labeled "suspicious" for future traceability and verification.

[0031] In addition, to prevent repeated false alarms from wasting resources, an additional verification mechanism will be added: after the second test is completed, the difference between the old and new optimal solutions will be compared again. Only if the change is less than or equal to a certain maximum allowable error range (such as 5 mmHg) will the remedial measure be recognized as successful and the latest version be identified as legal and valid data; otherwise, the entire batch of data will be completely abandoned and a warning will be sent to the management personnel to intervene and handle the situation as soon as possible.

[0032] Step S108: Merge the effective blood pressure data with the user's historical health data and input the data into the neural network model to generate a blood pressure trend prediction curve; Once a sufficient number and quality of real blood pressure observation samples are obtained, a trend analysis model for the future can be constructed. The essence of this step lies in uncovering the evolutionary patterns hidden behind massive historical records, thereby making scientifically reasonable predictions about blood pressure trends in the future.

[0033] Specifically, the most recent, filtered blood pressure data, along with a large number of previously accumulated records, are packaged together and fed into a deep learning network. Long Short-Term Memory (LSTM) networks, in particular, are well-suited for this type of application due to their excellent temporal modeling capabilities. After sufficient iterative training, this neuronal structure can autonomously summarize various implicit pattern features when faced with new inputs, such as circadian rhythm fluctuations, seasonal responses, and even emotional fluctuations. Based on these features, a smooth and continuous trend graph is drawn for medical staff to reference.

[0034] More importantly, this prediction goes beyond simply showing future numerical trends; it can be further expanded to include more practical functional modules. For example, if the systolic blood pressure remains above the normal upper limit (140 mmHg) for three consecutive days in the morning, a medication reminder mechanism will be automatically triggered, embedding the relevant information into the upcoming health report to remind the patient to take their antihypertensive medication on time to prevent the condition from worsening.

[0035] Step S109: The blood pressure trend prediction curve is encrypted based on the encrypted user identifier and distributed to the user terminal device according to the preset permission rules.

[0036] This step focuses on how to properly safeguard sensitive medical information and achieve limited sharing while protecting privacy. Because it involves everyone's private health, any disclosure could have serious consequences, so a robust security barrier must be established.

[0037] To address this, the system employs an end-to-end encrypted transmission protocol, binding all generated output files to a unique "encrypted user identifier" initially created. Only the party possessing the correct key can successfully decode and restore the original content. This ensures that even if a hacker attack occurs during transmission, the specific content details cannot be accessed. Furthermore, coupled with fine-grained access control policies, differentiated viewing permissions can be granted based on different roles (attending physician, family member, nursing staff, etc.), truly satisfying the needs of multi-party collaboration while rigorously preventing unauthorized leaks.

[0038] The above implementation integrates multiple measurement methods and an intelligent dynamic adjustment mechanism, achieving a fundamental shift from passive, static measurement to proactive, dynamic monitoring, thus improving data accuracy and user experience. The system generates personalized monitoring plans based on user authentication and historical health data, and dynamically calibrates measurement parameters using real-time motion status, effectively reducing the impact of environmental interference on results. The collaborative work and intelligent fusion of multiple blood pressure measurement methods further enhances measurement reliability. An automatic retesting and marking mechanism for abnormal data strengthens the system's robustness. The integration of a neural network model enables blood pressure trend prediction, providing strong support for disease early warning and health management. End-to-end data encryption and access control ensure user privacy and information security, demonstrating promising clinical application prospects and widespread application value.

[0039] Reference Figure 2 As one implementation of step S102, the step of generating a personalized monitoring plan based on the user's historical health data and the doctor's preset instructions includes: Step S201: Parse the user's historical health dataset stored in the server and extract the temporal fluctuation characteristics of systolic / diastolic blood pressure and abnormal event markers; In practical applications, users' long-term blood pressure records are typically stored in a structured format in cloud databases or local caches. This data includes not only the timestamps and numerical results of each measurement but also environmental variables (such as emotional state and medication status). By cleaning and normalizing this multi-dimensional time-series data, and using statistical methods to identify trends and periodic patterns in individual blood pressure changes, key indicators reflecting the stability of a patient's condition—namely, time-series fluctuation characteristics—can be extracted. For example, if a user's mean systolic blood pressure remains within the normal range for several consecutive weeks, but their standard deviation is large, it indicates weak blood pressure regulation and potential risk. Conversely, frequent occurrences of blood pressure below baseline at specific time points can be defined as "abnormal event markers," leading to labels such as "nocturnal hypotension" and "postprandial lethargy." The core objective of this step is to establish an intermediate representation model that can quantitatively express the user's current physiological state and its degree of instability, providing a basis for subsequent decision-making.

[0040] Step S202: Read the doctor's preset instruction set, which includes hypertension level labels, body position restrictions, and measurement method priority parameters; The doctor's preset instruction set is essentially a semantically structured configuration file, which may be in XML format, JSON object, or other standardized communication protocol encapsulation. It carries the clinical expert's treatment intentions and operating procedures for a specific patient.

[0041] Specifically, the hypertension grade label, as the primary classification identifier, reflects the severity of the patient's current condition. This is an assessment system clearly defined in internationally accepted hypertension prevention and treatment guidelines, with different grades corresponding to different intensities of lifestyle interventions and frequencies of medication adjustments. Postural restrictions reflect an awareness of differences in human motor function, particularly applicable to elderly patients or those recovering from surgery. For example, patients who cannot raise their arms due to shoulder surgery are clearly not suitable for measuring upper arm blood pressure using traditional cuff methods. The measurement method priority parameter further refines the preference for selecting testing equipment, allowing physicians to set priority orders based on different situations, such as auscultation over electronic sensing methods. These three elements together constitute a set of externally controllable, highly customized intervention rules to guide the next steps of the automated system.

[0042] Step S203: Generate a baseline measurement time series by matching a preset frequency rule base with hypertension grade labels; The pre-built frequency rule base refers to a pre-constructed mapping table or conditional judgment tree, which internally encodes a series of sampling frequency recommendations for various hypertension levels. This approach stems from evidence-based medicine research findings: critically ill patients require more intensive monitoring of vital signs to detect deterioration trends promptly, while mild cases can have their monitoring density appropriately relaxed to reduce the burden and improve compliance. Therefore, when the input signal is stage III hypertension, the system automatically triggers the highest frequency arrangement (once per hour) to ensure 24 / 7 coverage of all possible risk windows; for stage I patients, due to their lower overall risk and relatively stable daily fluctuations, maintaining a basic rhythm of three times a day is sufficient to meet basic monitoring needs. This timeline planning is not a simple copy of a textbook template, but a recommendation strategy optimized based on a large amount of real-world evidence, ensuring both scientific rationality and user experience comfort.

[0043] Step S204: Based on the body position constraints, select available body position parameters from the pre-configured set of body position parameters; The impact of body posture on blood pressure readings is widely recognized, especially in the elderly, where changes in posture can easily trigger transient ischemic attacks (TIAs). Therefore, modern intelligent monitoring devices often support multiple measurement positions, including sitting, lying down, and even side-lying positions, allowing for operation in a relaxed state. However, not everyone is suitable for all options. If a patient has severe cervical spondylosis, forcing them to remain supine for extended periods may worsen discomfort or even cause complications. In such cases, secondary filtering based on the "posture restrictions" field provided by the doctor is necessary.

[0044] Specifically, the system maintains a complete list of physical constraint descriptors in memory, covering multiple dimensions such as the degrees of freedom of limb movement and the stability of the support surface. Once it receives a clear instruction such as "limited left shoulder movement," it immediately eliminates movement paths that rely on the coordination of the left limbs, thus avoiding unnecessary attempts that could cause additional injury. This process is essentially a local search space compression, making the final output more closely reflect the individual's actual situation.

[0045] Step S205: Integrate measurement method priority parameters, temporal fluctuation characteristics, and abnormal event markers to construct a blood pressure measurement method combination strategy; Specifically, the strategy for constructing a combination of blood pressure measurement methods includes: if the time-series fluctuation characteristics show that the standard deviation of systolic blood pressure is >15 mmHg, then a combination of electronic Korotkoff sound method and oscillometric method is used; if the abnormal event marker includes "orthostatic hypotension", then the supine pulse wave method is forcibly enabled; the default combination strategy is a sequence of methods ordered according to the priority of the measurement methods preset by the doctor.

[0046] Traditional single-measurement methods are insufficient to fully capture the true nature behind complex pathological phenomena, especially when faced with drastic fluctuations or persistently low values. Therefore, it is necessary to introduce a multi-complementary mechanism to address this issue.

[0047] In this context, "fusion" is not simply about listing multiple candidate options for selection; more importantly, it involves making the optimal arrangement based on a thorough understanding of the advantages and disadvantages of each option. For example, if the coefficient of variation of systolic blood pressure is significantly higher over a certain period, it indicates that the body is in a relatively active state transition process. Relying solely on a shock wave detector may not be sufficient to accurately reconstruct the vascular pressure transmission characteristic curve. In this case, it is advisable to simultaneously use a stethoscope to assist in determining the timing of the first sound, thereby improving the overall reliability.

[0048] In addition, if a clear tendency toward orthostatic hypotension is known, a pulse wave analyzer that can respond quickly in a supine position should be given priority, regardless of other factors, to prevent false alarms and potential safety hazards.

[0049] Step S206: Output a personalized monitoring plan, including basic measurement time series, available body position parameters, and blood pressure measurement method combination strategies.

[0050] The personalized monitoring solution can be viewed as a complete action plan document, integrating all the conclusions derived from the preceding modules to form an operation manual directly usable by the execution unit. This document should include not only a specific task schedule (when to measure? how many times?), but also the corresponding hardware interface call order (which sensor to start first? which type of probe to activate later?), and necessary fault tolerance contingency plans (how to recover from errors? Is it permissible to skip this data collection?). Furthermore, to facilitate cross-platform portability and sharing, it can be encoded into a unified format data packet for external distribution, enabling seamless integration by downstream applications.

[0051] In the above implementation, the comprehensive modeling of users' historical physiological data and clinical intervention information realizes the technical transformation process from static data analysis to dynamic behavioral strategy formulation, opens up the information flow channel between the original health records and actual nursing practice, effectively solves the problems of blindly following the trend and one-size-fits-all in the traditional management model, and truly achieves the people-oriented service concept throughout.

[0052] Reference Figure 3 As one implementation of step S103, the steps of collecting user motion state data in real time based on body position parameters, dynamically calibrating the cuff inflation pressure threshold and signal sampling frequency, and obtaining calibration configuration parameters include: Step S301: Obtain available body position parameters from the personalized monitoring plan; Specifically, personalized monitoring programs refer to health monitoring plans designed for specific individuals in different scenarios. These programs typically include recommended measurement postures (such as sitting or lying down), environmental conditions, and historical physiological response characteristics. The "postural parameters" do not refer to general body position information, but specifically to the geometric angle reference system corresponding to the target measurement posture defined under standard medical guidelines. For example, when lying supine, the upper arm should be in a horizontal or slightly elevated angle range. These parameters are derived from ergonomic analysis and clinical experience summaries during the initial modeling phase and serve as the benchmark for all subsequent dynamic compensation operations.

[0053] Step S302: The user's motion state data, including three-dimensional spatial angles and acceleration vectors, is collected in real time through the inertial measurement unit. This step integrates sensor modules such as a three-axis gyroscope, accelerometer, and even a magnetometer, enabling the sensing of the human body's rotational and translational behavior in Euclidean space. The collected data not only includes spatial orientation information reflecting limb orientation, such as pitch, roll, and yaw angles, but also linear acceleration components along the three orthogonal XYZ directions.

[0054] It should be noted that in practical applications, since the sensor installation position may deviate from the ideal anatomical node, filtering algorithms (such as complementary filters or extended Kalman filters) are needed to eliminate noise effects and complete the mapping transformation from raw sensor readings to the actual physical posture. This high-dimensional multi-source data provides the system with comprehensive and accurate body motion feedback capabilities.

[0055] Step S303: Analyze the reference angle range of the target body position based on the available body position parameters; The analysis process is essentially a model matching and boundary definition operation, transforming abstract medical guidance into a measurable angular range. For example, in the typical case of "sitting with arms hanging naturally," theoretically, the angle between the upper arm and torso should be maintained between 80 and 100 degrees. Considering factors such as muscle relaxation and differences in joint mobility, a fluctuation of approximately ±5 degrees is allowed. By structurally encoding this type of prior knowledge, a fault-tolerant decision window can be formed, making subsequent judgments more closely reflect the complexities of real-world application scenarios.

[0056] Step S304: Calculate the deviation between the user's motion state data and the reference angle range; The system continuously compares the current attitude with the expected attitude to quantify whether the user's actual actions conform to preset specifications. Specifically, if the current pitch angle exceeds the permissible upper limit or falls below the minimum limit, it is considered to have a significant deviation; the same applies to attitude parameters in other dimensions.

[0057] Furthermore, considering that relying solely on a single-angle indicator is insufficient to fully characterize complex limb movement phenomena, it is often necessary to integrate information from multiple degrees of freedom to construct a more robust deviation evaluation function. For example, weighted Euclidean distance can be used to measure the overall deviation magnitude, or the tolerance thresholds for anisotropy can be fused into a multi-dimensional hyperrectangular region for inclusion testing. Such approaches help enhance the system's robustness and reduce the probability of false positives.

[0058] Step S305: Generate the cuff inflation pressure threshold calibration coefficient and signal sampling frequency scaling factor based on the deviation value, and output the calibration configuration parameters.

[0059] Specifically, the intelligent adjustment of the internal operating parameters of the blood pressure measuring device is driven by body posture deviations. The determination of the "cuff inflation pressure threshold calibration coefficient" essentially reflects that different body positions cause changes in the gravitational component of the arteries, thereby altering the hydrostatic pressure distribution pattern of the blood.

[0060] Therefore, the initial pressure setting must be adjusted accordingly to ensure that the prerequisite for effective blood flow closure is met. If the user is found to be in an unstable or non-standard position during measurement, appropriately increasing the initial pressure value can avoid the risk of underestimation due to incomplete closure of local blood flow pathways. On the other hand, the setting of the "signal sampling frequency scaling factor" focuses on combating the challenge of signal-to-noise ratio degradation caused by external disturbances. When severe limb tremors or large fluctuations in overall acceleration are observed, it means that the acquired heart sound and pulse signals are highly susceptible to artifact interference. Therefore, it is necessary to reduce the ADC (analog-to-digital converter) sampling rate to obtain higher channel gain and time domain resolution, and vice versa.

[0061] Ultimately, the output calibration configuration parameters are a set of specific instruction sequences for downstream hardware controllers. These are written into the register configuration area of ​​the embedded pressure measurement chip in real time to regulate multiple low-level operating parameters such as the start and stop timing of the air pump, the air release rate curve, and even the AD conversion cycle.

[0062] In the above embodiments, a highly coupled dynamic parameter adjustment system is constructed by integrating personalized user posture settings with real-time six-degree-of-freedom motion tracking. This system enables a shift from passively accepting static medical advice to actively adapting to changing situations. Utilizing continuous spatial orientation feedback provided by the IMU, it accurately identifies potential sources of deviation and triggers targeted adjustments to pressure thresholds and optimization of sampling strategies. This technical solution not only effectively mitigates the cumulative effect of systematic errors caused by improper posture but also, to some extent, suppresses transient disturbances caused by random vibrations.

[0063] As one implementation of step S104, the step of calling the corresponding sensor module to perform blood pressure and pulse rate measurement based on the blood pressure measurement method combination and calibration configuration parameters to obtain the blood pressure value collected by each blood pressure measurement method includes the following three cases: (1) When the blood pressure measurement method combination includes electronic Korotkoff sound method, the calibrated signal sampling frequency is used to control the microphone to collect arterial sound signals, and the blood pressure value is calculated by phase extraction algorithm; The electronic Korotkoff sound method originates from the classic auscultatory blood pressure measurement concept, which uses the characteristic sound signals generated during the gradual release of pressure after the cuff compresses the artery to determine the timing of systolic and diastolic blood pressure. However, traditional Korotkoff sound recognition relies on human interpretation by medical personnel, which suffers from high subjectivity and poor repeatability. Therefore, introducing the "electronic Korotkoff sound method" into modern intelligent devices means that sophisticated acoustic sensing devices and advanced signal processing methods must be used to achieve automated recognition and quantitative output.

[0064] Based on this, setting the calibrated signal sampling frequency becomes one of the key aspects of ensuring system performance. The sampling frequency refers to the number of discrete samples taken from the analog signal per unit time, directly affecting the effective resolution and response speed of subsequent digital signal processing. In this embodiment, the frequency is not fixed but dynamically adjusted after initial body position detection and physiological state assessment. For example, if the user is in a resting state (such as bed rest), due to less external interference, the system can choose a relatively low sampling rate (such as 100 Hz) to save power; conversely, in situations involving movement or frequent changes in body position, it is necessary to increase the sampling frequency (such as to 200 Hz or even higher) to more accurately capture high-frequency arterial sound signals and prevent information distortion caused by aliasing.

[0065] As a sensing unit, the microphone is responsible for converting sound pressure fluctuations in the air into voltage signals. To ensure high-quality data acquisition, the selected microphone should possess characteristics such as good directionality, a wide operating frequency band (typically covering 30 Hz to 300 Hz), and a high signal-to-noise ratio. The acquired sound signal also needs to be converted into a digital sequence that can be analyzed by a computer via an analog-to-digital converter (ADC). In this process, the choice of sampling frequency directly determines the balance between quantization accuracy and frequency domain resolution.

[0066] Subsequently, a phase extraction step is performed. By analyzing the time-frequency of the acquired audio signal, the start and end points of Korotkoff sounds are identified, and the time points of blood flow recovery and re-obstruction are deduced accordingly. These algorithms often employ mathematical tools such as Short-Time Fourier Transform (STFT) and Continuous Wavelet Transform (CWT) to analyze transient events in complex noise backgrounds. Specifically, the algorithm first denoises the original signal, then extracts its envelope curve, and subsequently tracks the changing trend of the energy concentration region. Once a sudden surge in signal intensity is detected within a certain time period and maintained for a certain duration, it can be determined as the location corresponding to systolic blood pressure; similarly, when the energy rapidly decays to below a predetermined threshold, it indicates the arrival of diastolic blood pressure. It should be noted that "phase" here does not simply refer to the angular attribute of a sine wave, but rather reflects a critical transition state in time.

[0067] Ultimately, the system combines these timestamps with the synchronously recorded pressure curves to establish a one-to-one mapping relationship between the two, thereby retrieving specific blood pressure values. The entire process achieves a complete transformation path from physical phenomena to numerical representation and then to physiological indicators, possessing the advantages of both objectivity and accuracy.

[0068] (2) When the blood pressure measurement method combination includes the oscillometric method, the air pump is controlled to inflate and deflate using the calibrated cuff inflation pressure threshold, and the oscillation wave signal is obtained through the pressure sensor to calculate the blood pressure value; Among them, the oscillometric method, a widely used classic non-invasive blood pressure measurement method, is based on the idea of ​​observing the change in pressure inside the cuff during the deflation process while monitoring the weak vibration signal generated by the elastic rebound of the arterial wall. This mechanical disturbance, known as an "oscillatory wave," essentially reflects the dynamic response of blood flow to the release of external constraints. Its amplitude exhibits a typical bell-shaped distribution, with the peak position usually corresponding to the vicinity of the mean arterial pressure, and the inflection points on both sides of the slope indicating the systolic and diastolic pressure ranges, respectively.

[0069] However, setting the initial cuff inflation pressure appropriately has always been a crucial factor affecting the stability and accuracy of oscillometric methods. Excessive pressure can lead to patient discomfort and even increased risk of tissue damage, while insufficient pressure may fail to completely seal the vascular access, resulting in missed detections and false alarms. Therefore, this application proposes a strategy to dynamically adjust the "cuff inflation pressure threshold" based on prior body position calibration results. The cuff inflation pressure threshold is a predefined upper limit value used to guide the control system when to stop the inflation action. In actual operation, this threshold is not statically fixed but is a function output determined by combining the user's real-time posture information (such as tilt angle, acceleration vector, etc.) and historical data. For example, when a user changes from a sitting to a standing position, theoretically, the central venous pressure decreases, and the pressure on the peripheral arteries also decreases accordingly. If the original inflation standard is still used at this time, the cuff may fail to effectively block blood flow, causing subsequent signal acquisition to fail. Conversely, if the inflation target value can be increased according to the new situation (e.g., from 180 mmHg to 200 mmHg), it can better match the current physiological conditions and ensure a sufficient detection window width.

[0070] At the control level, the operation of the air pump is driven by a closed-loop feedback mechanism. This mechanism integrates real-time readings from the built-in pressure sensor with a preset target pressure curve, continuously adjusting the inflation and deflation rhythm until the desired state is achieved. In this process, the pressure sensor plays a dual role: on the one hand, it provides feedback signals to support the operation of the control logic; on the other hand, it is also a true measurement probe, responsible for picking up those subtle pressure fluctuations that reflect the characteristics of vascular pulsation.

[0071] Once the formal measurement phase begins, the pressure sensor continuously monitors pressure fluctuations within the cuff cavity and converts them into continuous electrical signals for subsequent circuit conditioning and data analysis. These raw signals often contain a large number of irrelevant components (such as respiratory disturbances and muscle twitches), so they must be filtered and purified beforehand. Commonly used methods include, but are not limited to, Butterworth low-pass filtering and Kalman predictive estimation, with the aim of retaining useful information while minimizing the influence of external interference sources.

[0072] Subsequently, effective morphological parameters are extracted from the cleaned oscillatory wave signal. This step typically involves constructing a so-called "envelope" structure, which is a contour curve formed by connecting various local maxima. Studies have shown that the geometric characteristics of different parts of this envelope have a certain functional relationship with their corresponding blood pressure levels. For example, the maximum amplitude point is often close to the mean arterial pressure range, while the two inflection points roughly correspond to the systolic and diastolic pressure regions. As for the specific judgment rules, various modeling approaches can be adopted. A common practice is to find the steepest position in the rising segment of the envelope curve as a candidate point for systolic pressure, or to select a specific percentage height in the falling segment as a marker point for diastolic pressure.

[0073] (3) When the blood pressure measurement method combination includes the pulse wave method, a flexible sensor is used to detect the pulse wave signal and the blood pressure value is calculated based on the pre-trained propagation time model.

[0074] Compared to the two traditional methods that rely on cuff compression, the pulse wave method represents a completely new non-contact or semi-contact approach to blood pressure estimation. Its theoretical foundation rests on the important assumption that as blood pressure levels rise, the walls of major arteries tend to harden, leading to a faster propagation speed of the pulse wave within them—the so-called "pulse wave velocity" (PWV). Conversely, by accurately measuring the time difference required for propagation along a specific arterial path, it becomes possible to indirectly estimate the current blood pressure level.

[0075] In this embodiment, flexible electronic materials are used to fabricate the sensing components. These materials generally possess excellent ductility, breathability, and biocompatibility, minimizing the wearer's discomfort while maintaining good contact. Common manufacturing processes include printing conductive inks, depositing nano-carbon particles, and weaving fibrous electrodes, all of which can meet the requirements for long-term wear. The working mechanisms of flexible sensors mainly fall into two categories: one is piezoelectric devices based on piezoresistors, and the other is photoplethysmography (PPG) sensors that utilize differences in light reflection. The former relies on changes in impedance caused by different forces applied to the surface to sense skin vibrations induced by heartbeats; the latter emits specific wavelengths of light that penetrate subcutaneous tissue, indirectly reflecting the periodic expansion and contraction of blood volume based on changes in the degree of color absorption. Regardless of the type, they all require dedicated interface circuitry for preprocessing operations such as front-end amplification, gain control, and common-mode rejection.

[0076] Subsequently, the collected raw waveform data cannot be directly used for blood pressure conversion; further processing and extraction of key feature parameters are required. One of the most important dimensions is the so-called "pulse wave propagation time" (PTT). Simply put, this refers to the time interval between the moment the heart ejects blood and the moment the corresponding pulse is received at a fixed observation point at a distance. Theoretically, a shorter PTT indicates a higher PWV, thus suggesting higher blood pressure; and vice versa. However, the actual situation is far more complex than the ideal model because, in addition to blood pressure itself, many variables such as heart rate, blood vessel diameter, blood viscosity, age, and gender can interfere with this value. Therefore, simple univariate fitting is clearly insufficient to meet real-world challenges.

[0077] To address this, this application introduces the concept of a "pre-trained propagation time model." This is an intelligent computational framework built upon big data resources and iteratively optimized over time; its essence lies in supervised learning. Developers collect a large amount of sample data in a laboratory environment, covering various age groups, body types, lifestyles, and other factors, while simultaneously recording standard blood pressure values ​​and other physiological parameters. Then, advanced algorithms such as statistical regression, neural networks, and support vector machines are used to uncover patterns hidden behind the massive amounts of data, ultimately constructing an empirical formula or mapping rule that can be used for prediction of unknown individuals. It should be noted that to make the model more universally applicable, additional variables are often added during training, such as body mass index (BMI) and basal metabolic rate (BMR), to enhance generalization ability and robustness. In practical applications, the system only needs to input the real-time measured PTT into the trained model to quickly obtain the corresponding blood pressure estimate.

[0078] In the above implementation, by customizing various key technical parameters (such as sampling frequency, inflation threshold, and selection of sensing materials), the product's adaptability to diverse groups and complex working conditions has been significantly enhanced, and strong support has been provided for improving the reliability of the results.

[0079] Reference Figure 4 As one implementation of step S108, the step of merging effective blood pressure data with the user's historical health data and inputting it into a neural network model to generate a blood pressure trend prediction curve includes: Step S401: Merge the valid blood pressure data with the user's historical health dataset by time alignment to generate a merged dataset; The core objective of this step is to build a data framework with a unified time dimension, so that the subsequent modeling stage can fully explore the potential correlations between different time periods. Valid blood pressure data usually comes from a series of continuous blood pressure readings collected by terminals such as smart wearable devices or home self-monitoring instruments; while the user's historical health dataset may include, but is not limited to, electronic medical records, results from multiple previous measurements, and other relevant vital signs parameters (such as heart rate, sleep quality indicators).

[0080] It's important to note that since these data often come from different acquisition systems with inconsistent sampling frequencies, a mechanism is needed to ensure that all variables have corresponding observations at the same time point. To this end, a "timestamp alignment" strategy is adopted. This involves searching for matching historical data items based on the timestamp of each blood pressure sample. If some moments are missing, linear interpolation is used to fill in the gaps, thus ensuring the overall data is complete and ordered. This approach not only improves the efficiency of subsequent model training but also mitigates information loss caused by sparse sampling to some extent, making it particularly suitable for situations where frequent monitoring is lacking at night or during other low-activity periods.

[0081] Step S402: Standardize the merged dataset to obtain standardized time series data; Specifically, for the constructed multidimensional time series matrix, the overall mean and standard deviation of each column are calculated as a reference benchmark, and then the original values ​​are normalized using the classic Z-score transformation method. In this way, attribute fields, regardless of their original category, are compressed to the same distribution range, thus preventing certain dominant large-scale variables from masking other subtle but equally important patterns of change.

[0082] Furthermore, such preprocessing techniques help accelerate the convergence speed of neural networks and enhance their generalization ability, as most optimization algorithms assume that the input signal fluctuates within a small range around the zero center. It is worth noting that for practical applications in the medical field, this standardization is not only to meet engineering requirements, but also to enable data from different patient groups and even cross-regional platforms to be compared and analyzed under the same evaluation framework, further improving the compatibility and transferability of the entire system.

[0083] Step S403: Input standardized time-series data into a pre-trained neural network model; wherein the neural network model includes a time-series feature extraction layer and a regression output layer; Specifically, the core component of this step is a deep learning architecture specifically designed to handle time dependencies, consisting of two key layers: a time-series feature extraction layer and a regression output layer. The former captures the dynamic evolution patterns hidden behind complex waveforms, while the latter transforms abstract representations into concrete numerical results. By pre-accumulating a large number of labeled samples offline and learning a stable and reliable mapping function through supervised learning, it can quickly react to new user data and provide immediate feedback. This deployment approach ensures both predictive performance and practicality, making it highly suitable for high-frequency interaction scenarios such as remote monitoring and chronic disease management.

[0084] Step S404: Extract the implicit periodic features and short-term fluctuation features of standardized time series data through the time series feature extraction layer; The specific technique used in this step is typically a Bi-directional Long Short-Term Memory (Bi-LSTM) network, a special variant of recurrent neural networks that possesses excellent long-term dependency modeling capabilities and gradient propagation stability. Compared to traditional RNNs that only advance frame by frame in a single direction, Bi-LSTM allows information to flow simultaneously along both forward and reverse paths. This means it can simultaneously perceive all the contextual background before the current node and the future evolution trend.

[0085] In other words, the forward LSTM excels at capturing gradual changes in blood pressure, such as the gradual increase under the influence of circadian rhythms, while the inverse part is better at identifying sudden oscillations caused by unexpected factors, such as transient peaks after emotional excitement or strenuous exercise. The combination of the two constitutes a powerful representation space that can reflect both macroscopic cyclical properties and microscopic perturbation effects, which is crucial for characterizing the complex regulatory mechanisms inherent in the human circulatory system.

[0086] Step S405: Based on implicit periodic features and short-term fluctuation features, the regression output layer generates a blood pressure value sequence within the future time window; This part of the functionality is mainly handled by a set of densely connected fully connected subnetworks. Its task is to receive condensed spatiotemporal feature vectors from the high-level semantic encoder and decode them back into a list of blood pressure estimates with clear physical meanings. Typically, this type of decoder sets up three to four intermediate layers to progressively refine the reconstruction process. Each layer is equipped with an activation function to activate nonlinear components and prevent over-smoothing.

[0087] It's important to note that the output here is neither a simple classification label nor a fuzzy probability density distribution, but rather a precise set of numbers representing commonly used clinical parameters such as systolic blood pressure, diastolic blood pressure, and even pulse rate. In other words, given a historical observation window starting at a specific time point t, this module can infer the complete triad of blood pressure combinations at fixed intervals over the next few hours, forming a complete dynamic curve for medical staff to reference and make decisions.

[0088] Step S406: Output the blood pressure value sequence as a blood pressure trend prediction curve.

[0089] Ultimately, the result is no longer an isolated single numerical point, but a complete description of a development path including multiple time slices. This curve visually displays the blood pressure changes of the target subject over a specific future time period, which can be used to warn of impending risks and to assist in the formulation of personalized intervention measures.

[0090] The above-described implementation not only effectively addresses the data fragmentation problem prevalent in the existing health management field, but also significantly lowers the barrier to human intervention due to its high level of automation, greatly expanding the coverage and depth of smart healthcare services. At the same time, the robustness and adaptability demonstrated by this technical solution provide valuable experience and technical paradigm guidance for the development of more similar chronic disease prediction projects.

[0091] Reference Figure 5 As a further implementation of the intelligent dynamic blood pressure and pulse rate monitoring method, after outputting the weighted average as effective blood pressure data, it also includes: Step S501: Obtain the pulse rate value and abnormal label from the valid blood pressure data, and generate the pulse rate stability index by combining the diurnal rhythm characteristics from the user's historical health data. The system generates a key quantitative indicator for measuring pulse rate stability—the Pulse Rate Stability Index (PRSI)—by acquiring pulse rate values ​​and abnormal tags from the user's valid blood pressure data and combining this with the circadian rhythm characteristics contained in the user's long-term accumulated historical health data. The core of this process lies in establishing a dynamic matching relationship between the current physiological state and the individualized biological clock.

[0092] Specifically, circadian rhythms refer to a series of approximately 24-hour periodic physiological changes regulated by the suprachiasmatic nucleus of the hypothalamus. These changes include body temperature, hormone secretion, metabolic levels, and even cardiac activity intensity, all exhibiting distinct diurnal fluctuations. To accurately characterize this feature, the system employs a time-series modeling method based on cosine function fitting, known as "cosine analysis." By regressing pulse rate data measured at regular intervals over at least 30 days, an idealized 24-hour pulse rate curve representing the user's daily rhythm trend is extracted. This ideal curve not only includes amplitude information (the difference between maximum and minimum values) but also clearly defines phase information (e.g., the morning peak typically occurs around 8 AM).

[0093] Building upon this foundation, the system further incorporates three dimensions of information for the comprehensive calculation of PRSI: First, the degree of short-term pulse rate variability, calculated by taking the standard deviation from the mean of the five most recent measurements, forming the coefficient of variation (CV). This parameter directly reflects whether the pulse has fluctuated dramatically in the recent period. Second, the phase shift, which compares the actual observed peak pulse time with the optimal peak time predicted by the historical rhythm model. If there is a significant delay or advance, it is considered that the body may have an autonomic nervous system regulation disorder. Third, additional weighting factors: if the blood pressure records for a certain period are marked with clear pathological labels such as "orthostatic hypotension" or "sinus bradycardia," then the deviation items corresponding to that interval are given higher sensitivity weights in subsequent scoring processes to more accurately capture potential danger signals.

[0094] Ultimately, these diverse variables were normalized and integrated into a standardized score between 0 and 100, where a lower score indicates higher stability, while a higher score suggests an unstable pulse rate, which may indicate increased sympathetic excitation or other stress responses.

[0095] Step S502: When the pulse rate stability index exceeds the preset fluctuation threshold, the ECG signal acquisition module is triggered to synchronously acquire the user's real-time ECG waveform data. Once the pulse rate stability index exceeds the preset fluctuation threshold, the system's external response module will be activated immediately, triggering the ECG signal acquisition device to start working and simultaneously capturing the user's current ECG waveform data.

[0096] Understandably, when the pulse rate fluctuates significantly outside the normal range or deviates from an individual's habitual rhythm pattern, it often indicates a decline or even decompensation in the body's autonomic nervous system's regulatory capacity. This is one of the precursors to many serious cardiovascular events (such as malignant arrhythmias and acute coronary syndromes). Therefore, it is necessary to use electrocardiogram (ECG) signals, which have greater diagnostic value, to explore the cardiac electrophysiological behavior at this moment.

[0097] However, considering the limited resources of portable wearable devices, it's impossible to activate all sensors indiscriminately around the clock. Therefore, an event-driven sensing protocol is used for intelligent scheduling and management. Event-driven means that the corresponding hardware unit is activated to perform a task only when specific conditions are met, thus saving energy and improving efficiency. In this scenario, the "event" is determined by the aforementioned PRSI exceeding the limit. Different thresholds are set for different populations based on the classification standards of different underlying diseases. For example, a threshold of 40 points might be set for hypertensive patients, while for those with a history of chronic heart failure, whose threshold is lowered to 30 points because they are more prone to serious consequences from even slight disturbances. Once this limit is reached, the system sends an activation command to the ECG electrode integrated into the wristband or chest patch, putting it into high-speed sampling mode.

[0098] Meanwhile, to prevent other concurrent operations from consuming excessive system resources and affecting the main workflow, some secondary sensor channels will be temporarily shut down, such as pausing SpO2 oxygen saturation detection. Furthermore, a time-domain alignment mechanism is specifically designed during signal acquisition to ensure that the captured data window fully covers the two-minute intervals before and after the heart rate abnormality, thus guaranteeing the time continuity and integrity required for subsequent HRV analysis. Additionally, to address common limb movement interference issues in real-world environments, the system is equipped with an advanced noise-resistant filtering algorithm that adaptively selects the most suitable lead system (usually lead II or a modified Chest Lead) based on environmental conditions to obtain the clearest and most reliable QRS complex waveform possible.

[0099] Step S503: Extract the RR interval sequence from the real-time ECG waveform data and calculate the heart rate variability characteristic value using a variability analysis algorithm; Heart rate variability (HRV) eigenvalues, as an important tool for evaluating the function of the autonomic nervous system, essentially reflect the variation amplitude and intrinsic structural patterns of the RR interval between two adjacent heartbeats. To comprehensively reveal the physiological information hidden behind this complex time series, this approach adopts a multi-level analytical strategy, encompassing multiple perspectives including time-domain statistics, spectral decomposition, and nonlinear dynamics.

[0100] Firstly, at the time domain level, the focus is on the overall dispersion of the RR intervals and the rate of local transient changes. Two commonly used representative parameters are SDNN (Standard Deviation of NN intervals) and RMSSD (Root Mean Square of Successive Differences between adjacent NN intervals). The former describes the overall dispersion of the heartbeat over a period of time and can reflect the overall level of autonomic nervous activity; the latter focuses on the acceleration and deceleration trends of the heartbeat between adjacent moments. Because it is highly dependent on vagal innervation, it is often regarded as a good indicator for assessing parasympathetic tone.

[0101] Secondly, the frequency domain analysis requires specialized power spectrum estimation techniques, especially for non-uniformly sampled raw RR sequences. Traditional Fourier transforms are not suitable; instead, the Lomb-Scargle periodogram method, more appropriate for this specific data format, is used. This method maps the energy distribution of the entire RR sequence to different frequency components, thus identifying which frequency bands are dominant. Generally, 0.04–0.15 Hz is referred to as the low frequency (LF), where energy mainly originates from the regulatory effects of the sympathetic nervous system; while 0.15–0.4 Hz is called the high frequency (HF), primarily resulting from vagal nerve control. The power magnitude and ratio of these two frequencies (LF / HF ratio) together constitute an important indicator of autonomic nervous system balance.

[0102] In addition, it includes some more cutting-edge nonlinear methods, such as sample entropy analysis. This is a new approach to measuring the randomness and regularity of time series data. Its core idea is to examine the ability of a given substring of length to maintain a similar evolutionary trajectory in the future. In other words, the more regular and repetitive the RR sequence, the smaller its corresponding sample entropy, and vice versa. Multiple studies have shown that in disease states such as congestive heart failure, cardiac automaticity is significantly weakened, causing RR sequences to tend to become monotonic and smooth. At this time, the sample entropy value often decreases significantly. This technique provides a completely new perspective to detect early signs of disease that have not yet appeared.

[0103] Step S504: Input the heart rate variability feature value, effective blood pressure data, and pulse rate stability index into the pre-trained cardiovascular risk assessment model, and output the short-term cardiovascular risk level. In this application embodiment, a novel multimodal feature fusion architecture is proposed, which mainly includes three layers: First, the Embedding Layer, which is responsible for converting different types of semantically related input variables into a unified high-dimensional vector representation; second, the Cross-Attention Mechanism, which allows different modalities to refer to each other's contextual information, thereby strengthening their semantic associations; and finally, the Classification Output Layer, which uses the Softmax activation function to generate four-level risk rating results.

[0104] Throughout the model training process, a large amount of real-world case data was used to guide parameter optimization, ensuring that the final discrimination boundary closely approximates actual conditions. For example, when a patient exhibits multiple adverse factors simultaneously, such as elevated diastolic blood pressure, a sharp increase in the LF / HF ratio, and persistently high PRSI, the system automatically increases the risk weight assigned to this combination, thereby improving the accuracy of the early warning.

[0105] Ultimately, the risk levels are divided into four tiers: Level 1 indicates a probability of less than 3% for a major cardiovascular event within the next 72 hours, falling within the safe range; Level 2 means a probability between 3% and 10%, requiring close monitoring; Level 3 corresponds to a probability of 10% to 30%, posing a high threat and requiring immediate medical intervention; Level 4 represents an extremely high risk of sudden onset (greater than 30%), necessitating immediate activation of emergency response procedures. This tiered system facilitates rapid response measures by medical personnel and also allows end-users to easily monitor the progression of their health conditions.

[0106] Step S505: Based on the short-term cardiovascular risk level, dynamically adjust the blood pressure measurement frequency and body position parameters in the personalized monitoring plan.

[0107] This includes dynamically adjusting the existing personalized blood pressure monitoring plan, including but not limited to selecting the measurement frequency and time window, as well as formulating postural restrictions for certain special circumstances. This process demonstrates that the entire closed-loop control system goes beyond post-event diagnosis, enabling it to provide preventative intervention recommendations in the immediate future.

[0108] Specifically, for those assessed as low-risk, the existing baseline blood pressure monitoring frequency of three times daily should be maintained. However, if the risk level rises to intermediate or higher, the monitoring density needs to be appropriately increased, such as switching to hourly monitoring for six consecutive hours. For those at extremely high risk, the highest level of monitoring strategy should be implemented, with vital signs re-measured every fifteen minutes until the condition is confirmed to be relieved. Another important consideration is postural management. Many elderly people, especially those with diabetes or Parkinson's disease, are prone to so-called "orthostatic hypotension," which is a sudden drop in blood pressure when standing up from a lying position, causing dizziness or even fainting.

[0109] Therefore, when the system detects a systolic blood pressure difference exceeding 20 mmHg between the supine and standing positions, it will forcibly switch back to the supine measurement method regardless of the current risk level. Similarly, if the HRV results show excessive sympathetic nerve activity (manifested as LF / HF>4), then to prevent inducing syncope, any form of standing test will be prohibited. All these adjustment commands are directly written into the blood pressure monitor's firmware memory via Bluetooth wireless communication, achieving a truly intelligent remote control experience.

[0110] The above-described implementation not only helps users promptly detect potential cardiovascular risks but also provides clinicians with richer and more detailed vital sign profiles, thereby contributing to the goal of early screening and treatment. More importantly, thanks to event-driven sensing technology and an adaptive resource allocation mechanism, the system effectively balances the trade-off between monitoring accuracy and user experience, greatly enhancing the usability and sustainable service capabilities of portable medical devices in real life.

[0111] Reference Figure 6 As a further implementation of the intelligent dynamic blood pressure and pulse rate monitoring method, after the step of outputting the short-term cardiovascular risk level, it also includes: Step S601: Based on the short-term cardiovascular risk level and blood pressure trend prediction curve, a health intervention instruction set is generated by matching the preset intervention strategy library. The intervention strategy base, a knowledge graph, is not simply a set of rules, but a dynamic semantic network that integrates current international authoritative standards for hypertension diagnosis and treatment with extensive clinical validation. Its nodes include not only traditional classification labels (such as risk levels 1 to 4) but also various physiological time-series patterns (such as morning peak hypertension and nocturnal non-dipper curves). These patterns are behavioral representations of the time dimension abstracted from continuous blood pressure monitoring data. They reflect the patterns of blood pressure changes in different individuals within their diurnal rhythms, thereby revealing potential cardiovascular event tendencies.

[0112] In this embodiment, to achieve accurate matching, the system employs a two-layer decision engine architecture: the first layer is a master-level matching module, which determines the basic treatment path for the target population based on the cardiovascular comprehensive score output by the machine learning model; the second layer is a fine-grained sub-rule mapping module, responsible for further refining the selection logic of specific intervention methods. The latter works similarly to a conditional reasoning mechanism, identifying abnormal states such as a sharp rise in blood pressure in the morning or persistent high blood pressure at night by comparing the numerical fluctuations in the input blood pressure curve over a specific time period, and activating corresponding treatment recommendations accordingly. This two-stage screening approach, from macro to micro, effectively improves the adaptability and feasibility of intervention recommendations.

[0113] Ultimately, the resulting health intervention instruction set is a data object organized in a standardized JSON format, containing multiple fields such as a unique identifier, scope of application definition, core measure description, and auxiliary guidance suggestions. This ensures that the entire intervention process has a good structure and is easy for subsequent system parsing and processing.

[0114] Step S602: Associate the health intervention instruction set with the encrypted user identifier according to the preset permission rules; This step employs a security architecture combining Role-Based Access Control (RBAC) and Attribute-Based Encryption (ABE). The former clearly defines the operational boundaries of various stakeholders. For example, patients can view the details of their health management plan but cannot change any medical order parameters; family members can only receive key, extracted warning information; doctors have full read and write permissions to adjust intervention strategies at any time; and emergency medical services are restricted to accessing only the emergency contact information and vital signs of the highest-level (Level 4) patients. This division of permissions helps prevent the leakage of sensitive medical information while ensuring necessary business collaboration capabilities.

[0115] Meanwhile, the application of ABE technology further enhances the system's security and flexibility. This algorithm allows different information fragments within the original instruction set to be assigned different decryption attributes, ensuring that only the person holding the corresponding private key can recover the relevant portion of the content. For example, information about drug dosages might be encrypted into a separate ciphertext component, accessible only to those with the "physician qualification" attribute; while general lifestyle tips (such as low-sodium diet reminders) could be made available to a wider audience.

[0116] Furthermore, all this encrypted data is uniformly linked to a global index built based on user identity hashes, forming a clear and controllable data chain. Even if an external attack leads to database leakage, the lack of legitimate authorization credentials prevents attackers from piecing together a complete and effective intervention plan, greatly protecting individual privacy rights from infringement.

[0117] Step S603: When the user terminal device requests to retrieve the blood pressure trend prediction curve, the health intervention instruction set is simultaneously pushed to the authorized terminal.

[0118] Specifically, throughout the interaction process, each client accessing the platform needs to provide dual identity verification information, including its encrypted user ID and proof of its current login role. This is one of the prerequisites for activating subsequent services. Once identity verification is successful, the cloud server will begin retrieving the latest blood pressure evolution trajectory image file related to it from the storage space, and simultaneously search for relevant intervention instruction records that have already been bound. Subsequently, differentiated content presentation strategies are adopted according to the needs of different types of users: if the query is initiated by medical personnel, in addition to displaying graphical trends, a complete electronic operation interface will be provided for them to review and even modify online in real time; conversely, if the access request is made by an ordinary patient, only a simplified text-based action guide list will be displayed, supplemented by an intuitive and easy-to-understand calendar reminder function to help them better follow medical advice and arrange their daily life rhythm.

[0119] It should be noted that in the event of an extremely dangerous situation (such as a Level 4 level, which is considered extremely high-risk), the system will automatically trigger an independent alarm linkage program to quickly transmit the patient's location coordinates and snapshots of key physiological indicators to the designated medical institution, thus gaining valuable time for rescue.

[0120] The above implementation method achieves a one-stop service capability, from individual health status perception to targeted intervention formulation and multi-party collaborative response. Compared with the traditional static management model, this technical solution does not only remain at the passive feedback level, but truly achieves an organic combination of proactive intervention, personalized customization, and efficient delivery, greatly improving the scientific level and social benefit output efficiency of chronic disease prevention and control.

[0121] This application also discloses an intelligent dynamic blood pressure and pulse rate monitoring system that integrates multiple measurement methods.

[0122] An intelligent dynamic blood pressure and pulse rate monitoring system integrating multiple measurement methods, specifically comprising: The user authentication and data acquisition module is used to acquire the user's input authentication information, generate an encrypted user identifier based on the authentication information, and retrieve the user's historical health data. The personalization generation module is used to generate personalized monitoring plans based on the user's historical health data and the doctor's preset instructions; the personalized monitoring plan includes a combination of body position parameters and blood pressure measurement methods. The parameter calibration module is used to collect user motion state data in real time based on body position parameters, dynamically calibrate the cuff inflation pressure threshold and signal sampling frequency, and obtain calibration configuration parameters. The multi-mode blood pressure measurement module is used to call the corresponding sensor module to perform blood pressure and pulse rate measurement based on the combination of blood pressure measurement methods and calibration configuration parameters, and obtain the blood pressure value collected by each blood pressure measurement method. The measurement result verification module is used to determine whether the standard deviation of the blood pressure value output by each blood pressure measurement method exceeds the preset difference threshold; if so, it triggers a retest and marks the abnormality; if not, it outputs the weighted average as the valid blood pressure data. The blood pressure trend prediction and analysis module is used to merge effective blood pressure data with the user's historical health data and input it into a neural network model to generate a blood pressure trend prediction curve. The encrypted distribution module is used to encrypt the blood pressure trend prediction curve based on the encrypted user identifier and distribute it to the user terminal device according to the preset permission rules.

[0123] An intelligent dynamic blood pressure and pulse rate monitoring system integrating multiple measurement methods according to an embodiment of this application can implement any of the above methods, and the specific working process of each module in the system can refer to the corresponding process in the above method embodiments.

[0124] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the division of a certain module is merely a logical functional division, and in actual implementation there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0125] This application also discloses a computer device.

[0126] A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement an intelligent dynamic blood pressure pulse rate monitoring method integrating multiple measurement methods as described above.

[0127] This application also discloses a computer-readable storage medium.

[0128] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above in an intelligent dynamic blood pressure and pulse rate monitoring method integrating multiple measurement methods.

[0129] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0130] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. An integrated multi-measurement method smart dynamic blood pressure pulse rate monitoring method, characterized by, The monitoring method comprises: obtaining user input authentication information, generating an encrypted user identifier according to the authentication information, and calling user historical health data; generating a personalized monitoring scheme based on the user historical health data and doctor preset instructions; wherein the personalized monitoring scheme comprises a body position parameter and a blood pressure measurement method combination; real-time acquisition of user motion state data according to the body position parameter, dynamic calibration of the cuff inflation pressure threshold and the signal sampling frequency, and obtaining of the calibration configuration parameter; blood pressure and pulse rate measurement by calling the corresponding sensor module according to the blood pressure measurement method combination and the calibration configuration parameter, and obtaining of the blood pressure value collected by each blood pressure measurement method; judging whether the standard deviation of the blood pressure value output by each blood pressure measurement method exceeds a preset difference threshold; if yes, triggering re-measurement and marking an abnormal label; if no, outputting a weighted average value as valid blood pressure data; merging the valid blood pressure data and the user historical health data, inputting a neural network model to generate a blood pressure trend prediction curve; encrypting the blood pressure trend prediction curve based on the encrypted user identifier, and distributing the blood pressure trend prediction curve to the user terminal device according to a preset permission rule.

2. The integrated multi-measurement method smart dynamic blood pressure and pulse rate monitoring method according to claim 1, characterized in that, The step of generating a personalized monitoring scheme based on the user historical health data and doctor preset instructions comprises: analyzing a pre-stored user historical health data set in a server, extracting time sequence fluctuation characteristics and abnormal event labels of systolic pressure / diastolic pressure; reading a doctor preset instruction set, including a hypertension grade label, a body position restriction condition and a measurement method priority parameter; generating a basic measurement time sequence according to the hypertension grade label matching a pre-configured frequency rule library; based on the body position restriction condition, screening available body position parameters from a pre-configured body position parameter set; fusing the measurement method priority parameter, the time sequence fluctuation characteristics and the abnormal event label to construct a blood pressure measurement method combination strategy; outputting a personalized monitoring scheme, including the basic measurement time sequence, the available body position parameters and the blood pressure measurement method combination strategy.

3. The integrated multi-measurement method-based intelligent dynamic blood pressure and pulse rate monitoring method as claimed in claim 2, wherein, The step of real-time acquisition of user motion state data according to the body position parameter, dynamic calibration of the cuff inflation pressure threshold and the signal sampling frequency, and obtaining of the calibration configuration parameter comprises: obtaining available body position parameters in the personalized monitoring scheme; real-time acquisition of user motion state data by an inertial measurement unit, including three-dimensional space angles and acceleration vectors; analyzing a reference angle range of a target body position according to the available body position parameters; calculating a deviation value of the user motion state data and the reference angle range; generating a cuff inflation pressure threshold calibration coefficient and a signal sampling frequency scaling factor based on the deviation value, and outputting the calibration configuration parameter.

4. The integrated multi-measurement method smart dynamic blood pressure and pulse rate monitoring method according to claim 3, characterized in that, The step of blood pressure and pulse rate measurement by calling the corresponding sensor module according to the blood pressure measurement method combination and the calibration configuration parameter, and obtaining of the blood pressure value collected by each blood pressure measurement method comprises: when the blood pressure measurement method combination includes an electronic Korotkoff method, using the calibrated signal sampling frequency to control a microphone to collect an arterial sound signal, and calculating a blood pressure value by a phase extraction algorithm; When the blood pressure measurement method combination contains tonometry, the inflation pressure threshold of the calibrated cuff is used to control the inflation and deflation of the air pump, the oscillatory wave signal is obtained through the pressure sensor, and the blood pressure value is calculated; When the blood pressure measurement method combination contains pulse wave method, a flexible sensor is used to detect the pulse wave signal, and the blood pressure value is calculated based on a pre-trained propagation time model.

5. The integrated multi-measurement method-based smart ambulatory blood pressure and pulse rate monitoring method as claimed in claim 1, wherein, The step of merging the effective blood pressure data with the user historical health data and inputting the neural network model to generate the blood pressure trend prediction curve includes: Merging the effective blood pressure data with the user historical health data set in time alignment to generate a merged data set; Standardizing the merged data set to obtain standardized time series data; Inputting the standardized time series data into a pre-trained neural network model; wherein the neural network model includes a time series feature extraction layer and a regression output layer; Extracting the implicit periodic feature and short-term fluctuation feature of the standardized time series data through the time series feature extraction layer; Generating a blood pressure value sequence within a future time window based on the implicit periodic feature and short-term fluctuation feature through the regression output layer; Outputting the blood pressure value sequence as a blood pressure trend prediction curve.

6. The integrated multi-measurement method smart dynamic blood pressure and pulse rate monitoring method according to any one of claims 1 to 5, characterized in that, After outputting the weighted average value as the effective blood pressure data, further comprising: Obtaining the pulse rate value and abnormal label in the effective blood pressure data, and generating a pulse rate stability index in combination with the circadian rhythm feature in the user historical health data; When the pulse rate stability index exceeds a preset fluctuation threshold, triggering a electrocardiogram signal acquisition module to synchronously acquire real-time electrocardiogram waveform data of the user; Extracting the RR interval sequence in the real-time electrocardiogram waveform data and calculating the heart rate variability feature value through a variability analysis algorithm; Inputting the heart rate variability feature value, the effective blood pressure data, and the pulse rate stability index into a pre-trained cardiovascular risk assessment model to output a short-term cardiovascular risk level; According to the short-term cardiovascular risk level, dynamically adjusting the blood pressure measurement frequency and body position parameters in the individualized monitoring scheme.

7. The integrated multi-measurement method smart dynamic blood pressure and pulse rate monitoring method according to claim 6, wherein, After the step of outputting the short-term cardiovascular risk level, further comprising: Based on the short-term cardiovascular risk level and the blood pressure trend prediction curve, matching a preset intervention strategy library to generate a health intervention instruction set; Associating the health intervention instruction set to the encrypted user identifier according to a preset permission rule; When the user terminal device requests to call the blood pressure trend prediction curve, synchronously pushing the health intervention instruction set to the authorized terminal.

8. An integrated multi-measurement method smart dynamic blood pressure pulse rate monitoring system characterized in that, The monitoring system comprises: A user authentication and data acquisition module for acquiring user input identity verification information, generating an encrypted user identifier based on the identity verification information, and calling user historical health data; An individualized generation module for generating an individualized monitoring scheme based on the user historical health data and doctor preset instructions; wherein the individualized monitoring scheme includes body position parameters and a blood pressure measurement method combination; A parameter calibration module for acquiring user motion state data in real time according to the body position parameters, dynamically calibrating the cuff inflation pressure threshold and signal sampling frequency, and obtaining calibrated configuration parameters; A multi-mode blood pressure measurement module, configured to invoke corresponding sensor modules to perform blood pressure and pulse rate measurements according to the blood pressure measurement method combination and calibration configuration parameters, and obtain blood pressure values collected by each blood pressure measurement method; A measurement result verification module, configured to determine whether the standard deviation of blood pressure values output by each blood pressure measurement method exceeds a preset difference threshold; if yes, trigger re-measurement and mark an abnormality label; if no, output a weighted average value as valid blood pressure data; A blood pressure trend prediction analysis module, configured to combine the valid blood pressure data with the user historical health data, and input a neural network model to generate a blood pressure trend prediction curve; An encryption and distribution module, configured to encrypt the blood pressure trend prediction curve based on the encrypted user identifier, and distribute the blood pressure trend prediction curve to user terminal devices according to a preset permission rule.

9. A computer device, characterized by: A computer program product, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program product, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method of any one of claims 1 to 7.

Images