A wearable human safety monitoring device and method

Abstract

The present application relates to the technical field of wearable devices, in particular to a wearable human safety monitoring device and method, comprising: a data acquisition module for coordinating various wearable devices and collecting physiological data on the wearable devices; a data verification module for verifying and fusing various physiological data to obtain physiological state information; a data reasoning module for inputting the physiological state information into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain a hierarchical health event judgment result; and an execution module for scheduling various wearable devices to perform collaborative response operations matching the hierarchical health event judgment result according to the hierarchical health event judgment result, thereby solving the problem of insufficient monitoring accuracy, reliability and active response capability caused by the lack of effective systematic collaborative mechanism among various wearable devices.

Description

Technical Field

[0001] This invention relates to the field of wearable device technology, specifically to a wearable human safety monitoring device and method. Background Technology

[0002] With the popularization of health management awareness and the rapid development of Internet of Things technology, wearable devices have become an important tool for continuously monitoring human physiological signals and warning of health risks. They play an increasingly crucial role in the early detection of diseases, chronic disease management, and outpatient health monitoring.

[0003] Currently, human safety monitoring based on wearable devices mainly relies on single or multiple independent devices to collect users' physiological data, and then using mobile terminal applications to aggregate, analyze, and issue alerts. However, because each wearable device typically operates as an independent data source and decision-making terminal, there is a lack of a systematic mechanism for effective cross-validation, fusion analysis, and collaborative response between devices. This makes the accuracy of monitoring results susceptible to interference from single-point data errors, limits the ability to judge complex health events, and often limits response measures to information alerts, making it difficult to form an efficient, reliable, and proactive closed-loop safety monitoring system. Summary of the Invention

[0004] To address the technical problem of insufficient monitoring accuracy, reliability, and proactive response capabilities due to the lack of an effective systematic coordination mechanism among wearable devices, this application provides a wearable human safety monitoring device and method.

[0005] The wearable human safety monitoring device and method provided in this application adopts the following technical solution:

[0006] A wearable human safety monitoring device includes:

[0007] The data acquisition module is used to coordinate various wearable devices and collect physiological data from each wearable device.

[0008] The data verification module is used to verify and fuse various physiological data to obtain physiological state information;

[0009] The data reasoning module is used to input physiological state information into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment, and obtain hierarchical health event judgment results.

[0010] The execution module is used to schedule each wearable device to perform collaborative response operations that match the results of the graded health event assessment, based on the assessment results.

[0011] This application also provides a wearable human safety monitoring method, including:

[0012] Coordinate various wearable devices and collect physiological data from each device;

[0013] The physiological data are verified and fused to obtain physiological state information;

[0014] Physiological state information is input into a pre-defined hierarchical artificial intelligence model for collaborative reasoning and state judgment, resulting in a graded health event judgment result.

[0015] Based on the results of the graded health event assessment, each wearable device is scheduled to perform a coordinated response operation that matches the assessment results.

[0016] Furthermore, the steps for coordinating various wearable devices and collecting physiological data from each device include:

[0017] Register each wearable device to the same monitoring network to obtain the device attribute information of each wearable device;

[0018] Based on the attribute information of each device, the sensor type, current battery level and wearing position of the corresponding wearable device are evaluated to obtain a description of the data acquisition capability of each wearable device.

[0019] Based on the description of each data acquisition capability, corresponding data acquisition tasks are assigned to each wearable device to obtain a data acquisition instruction set;

[0020] Based on the data acquisition instruction set, the corresponding wearable device is scheduled to synchronously acquire the sensing signals from the sensors on the wearable device, and the sensing signals are preprocessed by local timestamp mapping and formatting to obtain physiological data.

[0021] Furthermore, based on the attribute information of each device, the steps for evaluating the sensor type, current battery level, and wearing position of the corresponding wearable device to obtain a description of the data acquisition capabilities of each wearable device include:

[0022] Based on the attribute information of each device, the sensor type, current battery level and wearing position of each wearable device are evaluated to obtain the sensor capability evaluation results, battery status evaluation results and location confidence evaluation results of each wearable device.

[0023] Based on the sensor capability assessment results, battery status assessment results, and location confidence assessment results, calculate the comprehensive confidence score of each wearable device under the preset physiological parameter monitoring task;

[0024] Based on the comprehensive confidence score and combined with historical data collection quality records, the role and contribution weight of each wearable device in the current monitoring network are calibrated to obtain a description of the data collection capabilities of each wearable device.

[0025] Furthermore, the steps for verifying and fusing various physiological data to obtain physiological state information include:

[0026] Noise filtering and signal integrity analysis were performed on each physiological data point to obtain the first physiological data.

[0027] The first physiological data representing the same target physiological parameter are time-aligned, outlier data are identified and removed, and the second physiological data is obtained.

[0028] Cross-validation is performed on the second physiological data that characterizes the same target physiological parameter. After obtaining the cross-validation results for each second physiological data, confidence weights are assigned to each second physiological data according to the cross-validation results to obtain the third physiological data.

[0029] Based on the confidence weights of each third physiological data point, the third physiological data points are fused and information is combined to obtain physiological state information.

[0030] Furthermore, based on the confidence weights of each third physiological data point, the steps for fusing and combining the information from each third physiological data point to obtain physiological state information include:

[0031] Based on the confidence weights of each third physiological data point, a weighted fusion calculation is performed on the third physiological data points describing the same physiological indicator to obtain the weighted fusion intermediate results of each physiological indicator.

[0032] After extracting physiological features associated with the weighted fusion intermediate results from complementary arbitrary third physiological data, physiological state information is generated by combining information from each physiological indicator and each physiological feature based on the weighted fusion intermediate results and physiological features.

[0033] Furthermore, the pre-defined hierarchical artificial intelligence model includes a primary screening sub-model and a deep analysis sub-model. The steps for inputting physiological state information into the pre-defined hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain the graded health event judgment results include:

[0034] The primary screening sub-model is used to perform parallel multi-parameter abnormal pattern matching and preliminary risk assessment on physiological state information to obtain a preliminary set of abnormal indicators and the confidence level of physiological state information.

[0035] Activate and invoke the deep analysis sub-model, and combine it with the user's historical physiological state baseline to conduct in-depth analysis of physiological indicators related to the preliminary abnormal indicator set, and obtain the health event identification results and the result level corresponding to each health event identification result;

[0036] The results of each health event identification and its level are mapped through a pre-defined risk grading strategy mapping table to generate graded health event judgment results.

[0037] Furthermore, based on the results of the graded health event assessment, the steps for scheduling wearable devices to perform coordinated response operations that match the assessment results include:

[0038] By analyzing the health event type, risk level, and corresponding response measures contained in the graded health event assessment results, response strategy parameters are obtained;

[0039] Based on the response strategy parameters and combined with the data acquisition capability description of each wearable device, corresponding task roles are assigned to each wearable device.

[0040] Based on the task role, after generating an instruction set that matches the device functions of the wearable device, the instruction set is distributed to the corresponding wearable device;

[0041] According to the instruction set, each wearable device is scheduled and controlled to perform corresponding preset task actions to complete the collaborative response operation.

[0042] Beneficial effects achieved:

[0043] This application provides a wearable human safety monitoring device, comprising: a data acquisition module for coordinating various wearable devices and collecting physiological data from each wearable device; a data verification module for verifying and fusing the physiological data to obtain physiological state information; a data reasoning module for inputting the physiological state information into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain a graded health event judgment result; and an execution module for scheduling each wearable device to perform a collaborative response operation that matches the graded health event judgment result, based on the graded health event judgment result.

[0044] In this application, the data acquisition module coordinates the collection of physiological data by various wearable devices, establishing preliminary collaboration among them and laying the foundation for a systematic collaborative mechanism. The data verification module verifies and merges the physiological data, effectively improving the accuracy and reliability of the physiological data through cross-verification and fusion. The data inference module inputs the merged physiological state information into a preset hierarchical artificial intelligence model for collaborative inference and state judgment. By utilizing the analysis of the preset hierarchical artificial intelligence model and the comprehensive evaluation of the physiological state information, more accurate graded health event judgment results are generated, further enhancing the reliability of monitoring. The execution module schedules each wearable device to perform matching collaborative response operations based on the graded health event judgment results, realizing proactive and coordinated response among multiple wearable devices. Thus, a closed-loop systematic collaborative mechanism is constructed as a whole, from collaborative data acquisition, collaborative verification, collaborative inference to collaborative response, solving the problem of insufficient monitoring accuracy, reliability, and proactive response capabilities caused by the lack of an effective systematic collaborative mechanism among wearable devices. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the wearable human safety monitoring device of this application;

[0046] Figure 2 This is a flowchart illustrating the steps of a wearable human safety monitoring method according to this application;

[0047] Figure 3 This is a flowchart illustrating the steps involved in collecting physiological data from various wearable devices in this application.

[0048] Figure 4 This is a flowchart illustrating the steps involved in verifying and fusing various physiological data to obtain physiological state information for this application.

[0049] Figure 5 This is a flowchart illustrating the steps involved in obtaining graded health event assessment results based on physiological state information in this application.

[0050] Figure 6 This is a flowchart illustrating the steps involved in scheduling various wearable devices to perform collaborative response operations based on the results of a graded health event assessment, as described in this application.

[0051] 10. Data Acquisition Module; 20. Data Verification Module; 30. Data Inference Module; 40. Execution Module. Detailed Implementation

[0052] The following combination Figures 1-6 This application will be described in further detail.

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0054] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0055] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text implies three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution that simultaneously satisfies A and B. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0056] This application discloses a wearable human safety monitoring device.

[0057] Please refer to Figure 1 The wearable human safety monitoring device proposed in this embodiment includes:

[0058] The data acquisition module 10 is used to coordinate various wearable devices to collect physiological data from each wearable device; the data verification module 20 is used to verify and fuse the physiological data to obtain physiological state information; the data reasoning module 30 is used to input the physiological state information into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain a graded health event judgment result; the execution module 40 is used to schedule each wearable device to perform a collaborative response operation that matches the graded health event judgment result based on the graded health event judgment result.

[0059] Figure 1 The specific steps and flow of the data acquisition module 10, data verification module 20, data inference module 30 and execution module 40 are shown below.

[0060] This application also discloses a wearable human safety monitoring method.

[0061] Please refer to Figure 2 The wearable human safety monitoring method proposed in this embodiment includes the following steps:

[0062] S10~S40:

[0063] Step S10: Coordinate all wearable devices and collect physiological data from each wearable device.

[0064] By coordinating various wearable devices and collecting their physiological data, the barriers of isolated operation of each device in traditional monitoring models are broken down, laying the foundation for solving the problems of single data sources and fragmented information. Through proactive coordination, multiple wearable devices worn by the user are integrated into a preliminary collaborative data collection network, ensuring the comprehensiveness and synchronicity of physiological data acquisition from the source. This provides multi-dimensional and multi-angle physiological signals for subsequent processing, enabling a more complete and three-dimensional perception of the user's physiological state. It effectively avoids monitoring blind spots that may be caused by the lack or limitation of data from a single wearable device, thus providing an indispensable data foundation for building a reliable monitoring system based on multi-source physiological data fusion.

[0065] In step S20, the physiological data are verified and fused to obtain physiological state information.

[0066] By validating and fusing various physiological data, the potential noise, errors, and inconsistencies among the original physiological data are resolved. This transforms the dispersed and heterogeneous physiological data collected through the aforementioned steps into reliable and unified physiological state information. Validation involves cross-comparison and quality assessment of homogeneous or complementary physiological data from different wearable devices, identifying and eliminating abnormal or unreliable outliers to ensure the authenticity and accuracy of the information input into the model. The fusion process, based on the validation results of each physiological data point, weighted integrates and complements the features of the validated physiological information to synthesize a more comprehensive and stable integrated physiological state information. This significantly improves the quality and consistency of the basic information upon which the system relies, effectively overcoming the potential biases or limitations of a single data source, and building a reliable and high-quality data foundation for subsequent intelligent analysis and precise decision-making.

[0067] Step S30: Physiological state information is input into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain hierarchical health event judgment results.

[0068] Verified and fused physiological state information is fed into a pre-defined hierarchical artificial intelligence model for collaborative reasoning and state judgment. This surpasses the limitations of traditional simple threshold or single-model analysis methods, addressing the problems of inaccurate identification and imprecise judgment of complex, hidden, or compound health risks. Through the hierarchical structure of the pre-defined hierarchical artificial intelligence model, physiological state information is analyzed in a step-by-step intelligent manner, from shallow to deep and from local to global. The primary screening sub-model within the pre-defined hierarchical artificial intelligence model can quickly screen for a wide range of abnormalities, while the deep analysis sub-model performs multimodal correlation and deep reasoning on suspicious indicators. This achieves a quasi-mapping and severity assessment from basic physiological parameters to comprehensive health events, generating graded health event judgment results with clear health event categories and different risk levels. This not only significantly improves the accuracy and reliability of the monitoring device in identifying real health threats and reduces false alarms and missed alarms, but more importantly, it provides clear and operable decision-making basis for subsequent steps, enabling the entire monitoring device to adopt differentiated response strategies according to different levels of risk, thus realizing the transformation from data perception to intelligent cognition.

[0069] Step S40: Based on the results of the graded health event assessment, schedule each wearable device to perform a collaborative response operation that matches the results of the graded health event assessment.

[0070] By scheduling wearable devices to perform matching collaborative response operations based on the results of graded health event assessments, this approach solves the problems of traditional monitoring solutions, which often feature single, passive responses and a lack of synergy. It transforms the graded health event assessment results from previous steps into corresponding intervention actions, and schedules wearable devices based on these assessments. This maps the abstract graded health event assessment results into an executable sequence of tasks for multiple wearable devices, driving wearable devices with different functions and locations to work collaboratively based on their own characteristics. For example, one wearable device might issue a strong alert, while another simultaneously collects data at a higher frequency or sends a distress signal. This step constructs a closed loop from risk perception and intelligent assessment to proactive intervention, enabling the entire monitoring device not only to "know" what has happened but also to "execute" what needs to be done. This significantly improves the monitoring device's proactive response capability, efficiency, and overall reliability in the face of real health threats, achieving a transformation from passive monitoring to proactive safety monitoring.

[0071] In one feasible implementation, refer to Figure 3 As shown, step S10 may specifically include steps S11 to S14:

[0072] Step S11: Register each wearable device to the same monitoring network to obtain the device attribute information of each wearable device.

[0073] Through a central node acting as a coordinator—either a smartphone, a local gateway, or a designated master wearable device—the system proactively discovers and sends pairing requests to other wearable devices worn by the user, such as smartwatches, smart bracelets, smart earphones, and smart clothing, using Bluetooth, Wi-Fi, or low-power IoT communication protocols. After user authorization, a secure peer-to-peer connection is established with each wearable device, logically integrating these independent devices into a distributed wearable device cluster under unified scheduling and management, forming a monitoring network aimed at completing collaborative monitoring tasks. Once this connection is established, the coordinator requests or queries the corresponding device attribute information from each newly registered wearable device in the monitoring network. This device attribute information is key metadata describing the wearable device's monitoring capabilities and status, typically including at least the device type, the model of the sensor it carries, the current battery level, and the device's wearing location information obtained through near-field sensing or manual user calibration.

[0074] This step enables the logical integration of multiple previously isolated and heterogeneous wearable devices into a cohesive whole that is identifiable, addressable, and manageable, providing a device capability description basis for subsequent dynamic task planning and resource scheduling based on device capabilities.

[0075] Step S12: Based on the attribute information of each device, evaluate the sensor type, current battery level and wearing position of the corresponding wearable device to obtain a description of the data acquisition capability of each wearable device.

[0076] Based on the attribute information of each device, the sensor type, current battery level and wearing position of the corresponding wearable device are evaluated. This solves the problem of how to transform static and heterogeneous device attribute information into a data acquisition capability description that can be directly used for task planning, so as to provide a basis for subsequent task assignment.

[0077] By evaluating sensor types, the types of physiological parameters that each wearable device can monitor and their technical reliability are determined. By assessing the current battery level, the duration of continuous operation and power consumption constraints of the wearable device are determined to prevent interruptions during task execution. By evaluating the wearing position, the contact quality between the sensor and the body measurement site is confirmed, as well as the representativeness and accuracy of the signals collected at that location for specific physiological parameters. These multi-dimensional evaluation information are comprehensively processed and integrated into a unified data acquisition capability description. This description not only reflects the inherent hardware capabilities of the wearable device but also includes its current operating status and the confidence level of signal acquisition.

[0078] This step establishes a perceptible device capability map for the entire collaborative monitoring network, enabling the network to accurately perceive at what time, which wearable device in the network is capable, at what reliability level, and what quality and type of physiological data it collects. This lays a crucial decision-making foundation for achieving efficient, reliable, and adaptive multi-wearable device collaborative data acquisition task scheduling.

[0079] Furthermore, step S12 may also include steps S121 to S123:

[0080] Step S121: Based on the attribute information of each device, evaluate the sensor type, current battery level and wearing position of each wearable device to obtain the sensor capability evaluation result, battery status evaluation result and location confidence evaluation result of each wearable device.

[0081] For sensor type evaluation, based on the sensor hardware model and specifications recorded in the device attribute information, it is mapped to a preset knowledge base containing the known accuracy level, measurement range and anti-interference capability of various sensors when monitoring different physiological parameters. A quantitative sensor capability evaluation result is generated for each sensor of each wearable device. The sensor capability evaluation result is usually a vector or list, indicating the types of physiological parameters that the corresponding wearable device can monitor, such as heart rate, blood oxygen, body temperature, etc., as well as the technical support level and expected accuracy level for each physiological parameter.

[0082] For the current battery level assessment, the remaining battery percentage or voltage value is read from the device attribute information, and combined with the historical average power consumption data of the corresponding wearable device model, the estimated continuous working time of the wearable device under different intensity monitoring tasks is calculated. This estimated time or a corresponding battery health index is used as the battery status assessment result.

[0083] For the assessment of wearing position, based on the wearing location identifier set by the user or automatically detected and reported by the device's built-in near-field sensor in the device attribute information, such as left wrist, right wrist, chest, ear canal, etc., a preset physiological measurement knowledge base is queried. This preset physiological measurement knowledge base defines the inherent quality of signal for monitoring various physiological parameters for different body parts. For example, the chest has a higher confidence level for ECG monitoring than the wrist. Thus, a position confidence assessment result is generated that reflects the suitability of the position for the monitoring tasks that the wearable device can support. This is usually expressed as a confidence score or level.

[0084] Through the above steps, the device attribute information is transformed into three independent, quantifiable, and clearly physically meaningful evaluation results. This provides a precise and unified input basis for the next step of comprehensively calculating these multi-dimensional evaluation results into a description that can fully and objectively reflect the data collection capabilities of wearable devices for specific tasks at the current moment.

[0085] Step S122: Calculate the comprehensive confidence score of each wearable device under the preset physiological parameter monitoring task based on the sensor capability assessment results, battery status assessment results, and location confidence assessment results.

[0086] The monitoring device predefines a set of quantitative evaluation rules or an algorithm model for each preset physiological parameter monitoring task, such as heart rate monitoring, blood oxygen saturation monitoring, electrocardiogram monitoring, and fall detection.

[0087] When calculating the overall confidence score of a wearable device for a specific physiological parameter monitoring task, the process first involves retrieving and matching the corresponding sensor entries from the device's sensor capability assessment results, based on the signal type of the current preset physiological parameter monitoring task, such as ECG, photoplethysmography (PPG), etc. The quantified support level (e.g., support, preferred support) and accuracy parameters (e.g., sampling rate, resolution, measurement error range) of each sensor entry are directly extracted. These values ​​are then input into a standardized scoring function for this type of sensor. This standardized scoring function maps these values ​​to a dimensionless basic technical capability score. A pre-defined mathematical function or rule model maps the specific technical specifications of the sensor to a uniform and comparable dimensionless fraction. When calculating the overall score for the combination of position confidence scores, a pre-defined physiological measurement knowledge base is accessed. This knowledge base defines the inherent signal quality weights for different body wearing positions for various physiological parameter monitoring tasks. For example, when the task is ECG monitoring, the pre-defined physiological measurement knowledge base returns a very high weight coefficient for the chest position and a lower weight coefficient for the wrist position. The position confidence assessment result of the corresponding wearable device is then multiplied by the position weight coefficient for the corresponding pre-defined physiological parameter monitoring task to obtain the corresponding score. The location confidence score of the wearable device under the corresponding preset physiological parameter monitoring task; to convert the battery status assessment result into a task sustainability score, based on the preset power consumption level of the corresponding physiological parameter monitoring task (e.g., continuous high-frequency acquisition is high power consumption, intermittent sampling is low power consumption) and the historical average power consumption data of the corresponding wearable device model, the estimated remaining time that the wearable device can maintain the operation of the corresponding physiological parameter monitoring task under the current battery level is calculated. Then, this estimated remaining time is mapped to a task sustainability score through a preset standardized mapping function bound to the task characteristics. This standardized mapping function defines the relevant parameters of the corresponding physiological parameter monitoring task. The two key time thresholds are: one is the minimum sustainable threshold required by the task, which is usually set based on the typical execution cycle or the minimum effective monitoring window of the task; the other is the ideal sustainable threshold of the task, which is usually set much higher than the minimum requirement to ensure sufficient monitoring margin. The mapping function takes the estimated remaining time as input, and its calculation logic is as follows: if the estimated remaining time is lower than or equal to the minimum time threshold, the minimum sustainable threshold is directly output as the task sustainability score, indicating that the power is almost insufficient to guarantee the completion of the task; if the estimated remaining time reaches or exceeds the ideal time threshold, the ideal sustainable threshold is directly output as the task sustainability score, indicating that the power is sufficient.

[0088] Finally, a pre-defined weighted fusion algorithm, such as weighted summation or a multilayer perceptron-based scoring model, is used to comprehensively calculate the basic technical capability score, location confidence score, and task sustainability score. Specifically, firstly, a normalized initial weight is defined for each of the basic technical capability score, location confidence score, and task sustainability score. These initial weights are set based on prior knowledge of a large number of monitoring scenarios. For example, for ECG monitoring tasks with extremely high accuracy requirements, the initial weight of the sensor's basic technical capability score will be set higher, while for continuous monitoring tasks requiring long-term battery life, the initial weight of the task sustainability score will be increased accordingly. Next, based on the characteristics of the current pre-defined physiological parameter monitoring task... The initial weights are dynamically fine-tuned using a task adaptation factor driven by an embedded task knowledge base. This knowledge base defines the sensitivity of different physiological parameter monitoring tasks to each dimension. For example, blood oxygen saturation monitoring is extremely sensitive to the spectral capabilities of the sensor, while body movement or posture monitoring is relatively insensitive to positional information. The fine-tuned weights are multiplied by the corresponding normalized dimension scores and then summed to obtain a comprehensive confidence score. This transforms the discrete and heterogeneous capability and status assessment of wearable devices in the three dimensions of sensors, battery power, and location into a unified quantitative score for each specific physiological parameter monitoring task, thus providing a data-driven decision-making basis for assigning data collection tasks in subsequent steps.

[0089] Step S123: Based on the comprehensive confidence score and combined with historical data collection quality records, the role and contribution weight of each wearable device in the current monitoring network are calibrated to obtain a description of the data collection capabilities of each wearable device.

[0090] First, a historical data acquisition quality record is maintained. The historical data acquisition instruction record is a structured database that records the actual quality evaluation indicators of the output data of each wearable device when performing various physiological parameter monitoring tasks in the past, including but not limited to the signal-to-noise ratio of the data, the degree of consistency with standard device data or other device data, the adoption rate in cross-validation, and the frequency of signal interruption.

[0091] Using the comprehensive confidence score as an immediate snapshot of the wearable device's theoretical capabilities, this data is fused and analyzed with historical data collection quality records reflecting the wearable device's long-term actual performance. For example, the actual data quality index obtained after each successful execution of a physiological parameter monitoring task is taken as newly observed data, and the average quality level calculated by the wearable device based on long-term historical data collection quality records is taken as the prior probability distribution describing the inherent reliability of the wearable device. The comprehensive confidence score calculated in the current step is regarded as a new input of an observation that can reflect the immediate capabilities. The core calculation of the Bayesian update model is to combine the prior probability distribution with the observation and perform probability inference through Bayes' theorem. The result is an updated posterior probability distribution, and the mean of this posterior probability distribution is the calibrated capability score. Based on this post-calibration capability score, roles are assigned to all wearable devices in the monitoring network within the context of the current physiological parameter monitoring task. Here, a role is a classification label that defines the main function of the wearable device in the collaborative task. For example, the wearable device with the highest and most reliable post-calibration capability score is designated as the main monitoring device for critical decision-making, while other wearable devices are designated as auxiliary monitoring devices for redundancy verification or supplementary monitoring. At the same time, based on the post-calibration capability score, a quantified contribution weight is assigned to each wearable device. This weight is a value between 0 and 1, which directly determines the proportion of data adopted by the wearable device in subsequent data fusion or its task priority in collaborative scheduling.

[0092] Ultimately, the equipment attributes, the calibrated capability scores for each physiological parameter monitoring task, the assigned roles, and the contribution weights are encapsulated into a data acquisition capability description. This makes task scheduling no longer based on a one-time, superficial capability score, but on a deep understanding of the long-term comprehensive performance of the equipment, which greatly improves the robustness, prediction accuracy, and resource utilization efficiency of task allocation in the entire collaborative monitoring network.

[0093] Step S13: Based on the description of each data acquisition capability, assign corresponding data acquisition tasks to each wearable device to obtain a data acquisition instruction set.

[0094] The assignment process begins with a clear list of task requirements, which defines the various data acquisition tasks that need to be executed synchronously. Each data acquisition task specifies the type of physiological parameter to be monitored, the required signal quality, the sampling frequency, and the duration of the task.

[0095] The decision is made through a task allocation optimization algorithm, such as a resource scheduling algorithm based on a greedy strategy or linear programming. The core objective of this task allocation optimization algorithm is to maximize the overall confidence or reliability of the entire monitoring network while satisfying all task requirements. Its decision logic is as follows: for each data acquisition task, select one or more of the most suitable wearable devices from the monitoring network to undertake it. The main basis for selection is the calibrated capability score and role in the device's data acquisition capability description. Usually, the data acquisition task is assigned to the wearable device with the highest score and the main monitoring role as the primary data source. At the same time, the second highest score and the auxiliary role of the wearable device may be assigned as redundant or supplementary data sources. Furthermore, the specific task parameters assigned to each wearable device are precisely set according to the contribution weight of each wearable device. For example, a higher sampling frequency is assigned to wearable devices with high weight. Ultimately, an executable data acquisition instruction is generated for each selected wearable device. This instruction includes all operational details such as target physiological parameters, sampling configuration, start time, and duration. The combined instructions from all wearable devices constitute a data acquisition instruction set that coordinates the synchronous operation of the entire monitoring network. This transforms abstract device capability assessments into executable task scheduling schemes, ensuring that each physiological parameter monitoring task can be completed in the most efficient way by the most suitable and reliable combination of wearable devices in the current monitoring network. This optimizes the allocation of data acquisition resources and guarantees the overall quality and reliability of subsequent multi-source data fusion.

[0096] Step S14: According to the data acquisition instruction set, the corresponding wearable device is scheduled to synchronously acquire the sensing signals on the sensors on the wearable device, and the sensing signals are preprocessed by local timestamp correspondence and formatting to obtain physiological data.

[0097] The received data acquisition instruction set is parsed into a series of independent device instructions. Then, through the established monitoring network communication link, the instructions, containing specific start times, sampling parameters, and task details, are accurately distributed to each wearable device specified in the instruction set. Upon receiving the instruction, each wearable device will activate its designated sensor at the time indicated by the instruction (usually based on a synchronization clock signal broadcast by the coordinator) to begin acquiring sensing signals. These sensing signals are direct outputs from the sensors and reflect changes in physical quantities, such as electrical signals or digital readings (e.g., changes in light intensity received by a photodiode, potential differences between electrodes, or voltage values ​​from an accelerometer). Next, local timestamp matching is performed simultaneously with or after sensing signal acquisition. This is done by adding a timestamp from the local high-precision clock of the wearable device to each acquired sensing signal at the moment of its generation. Subsequently, using the monitoring network established in step S11, a time synchronization protocol (such as a simple offset compensation algorithm based on network delay estimation) is used to uniformly calibrate the local timestamps of all wearable devices to the same reference time axis, thereby ensuring strict time alignment of sensing signals from different wearable devices. The formatting preprocessing involves performing a series of transformations on the sensor signal according to a preset unified data standard. This includes, but is not limited to, converting analog signals into digital quantities, applying sensor-specific calibration coefficients to convert readings into physically meaningful values ​​(such as converting voltage values ​​into acceleration units g), and encapsulating the data into a standard data structure containing timestamps, device IDs, physiological parameter types, and values. The structured information with a unified time reference and standard format obtained after the above processing is physiological data.

[0098] This technology transforms sensor signals collected from multiple heterogeneous and independent wearable devices into physiological data that is precisely aligned in time, standardized in format, and clear and explicit in semantics, thereby fundamentally solving the most critical problems of spatiotemporal consistency and data isomorphism in multi-source data fusion.

[0099] In one feasible implementation, refer to Figure 4 As shown, step S20 may specifically include steps S21 to S24:

[0100] Step S21: Noise filtering and signal integrity analysis are performed on each physiological data to obtain each first physiological data.

[0101] Noise filtering targets physiological data from different wearable devices. Based on the corresponding sensor type and signal characteristics, it applies pre-defined digital signal processing algorithms. For example, for motion-sensitive signals like photoplethysmography (PPG) pulse waves, adaptive filters or null noise cancellation algorithms based on accelerometer signals are used to suppress motion artifacts. For ECG signals, wavelet transforms or notch filters may be used to eliminate power line interference and baseline drift. The core operation is to mathematically separate and attenuate interference components in the physiological signal that are unrelated to the target physiological information. Signal integrity analysis evaluates the noise-filtered physiological signal. Its core operation is to determine whether the physiological signal is complete and usable through a series of pre-defined rules and threshold detections. For example, it checks whether the signal amplitude is within a reasonable range, whether there are long-term flat lines or saturation distortions due to poor sensor contact, and whether there are discontinuous gaps in the physiological signal due to data packet loss caused by wireless transmission interruptions.

[0102] Through the above two steps, the physiological signals that have been effectively cleaned by noise filtering and passed signal integrity analysis are retained and output, and defined as the first physiological data. In this way, unreliable noise components are systematically identified and removed from the physiological data collected from multiple sources, while high-quality and usable effective signals are screened out. This provides input for subsequent fine alignment, cross-validation and fusion, and significantly improves the data quality foundation and anti-interference capability of the entire system from the starting point of the data processing flow.

[0103] Step S22: Time-align the first physiological data representing the same target physiological parameter, identify and remove outlier data, and obtain the second physiological data.

[0104] The time alignment operation in this step involves multiple first physiological data points from different wearable devices that represent the same physiological parameter. This ensures precise correspondence of the first physiological data at the signal feature level. Here, feature points, such as the R-wave peak of an ECG signal or the peak point of a pulse wave, are used to calculate and compensate for possible small phase differences and cumulative clock drift between different first physiological data sequences. Specifically, firstly, pairwise similarity analysis is performed on all the first physiological data to be aligned. For example, cross-correlation analysis is used to calculate the cross-correlation coefficient of each first physiological data point at different time offsets, thereby constructing a global model describing the relative phase relationship between all the first physiological data points. Alternatively, dynamic time warping is used to perform overall nonlinear path planning on all the first physiological data points to find a sequence that aligns the overall morphological features of all the first physiological data points at the lowest cost. Based on this global model, the phase offset of each first physiological data point relative to a common time axis is solved by least squares fitting. Furthermore, the change trend of this relative offset within a continuous time window is analyzed to identify and quantify the possible clock drift rate that accumulates linearly over time. Then, the phase offset is regarded as the initial time difference of the corresponding first physiological data sequence relative to the common time axis at the start time, while the clock drift rate is regarded as the rate deviation that changes linearly over time between its local clock and the common time axis. A corresponding linear compensation function is generated for each first physiological data point. The input of the linear compensation function is the local timestamp of the first physiological data point, and its mathematical form is: calibrated timestamp = (local timestamp - initial reference time point) / (1 - clock drift rate) + initial reference time point + phase offset.

[0105] The core function of the above linear compensation function is to map the local timestamp onto the calibrated common time axis. The term "(local timestamp - initial reference time point) / (1 - clock drift rate)" is used to compensate for the time axis scaling caused by clock frequency differences, while "phase offset" is used to compensate for the fixed initial phase difference.

[0106] After generating the linear compensation function, it is applied to the local timestamp of each data point of the corresponding first physiological data to calculate its new timestamp on the common time axis. If the compensation process involves time axis scaling (i.e., the clock drift rate is not zero), a resampling algorithm is usually required to generate a calibrated data sequence that is uniformly distributed on the common time axis. This completes the compensation process from parameters to final application, thereby eliminating the small phase difference and accumulated clock drift between the first physiological data and achieving high-precision time synchronization of all first physiological data. This ensures that the first physiological data from different wearable devices achieves accurate time correspondence at the sampling point level at each physiological event point (such as each heartbeat).

[0107] The process of identifying and removing outlier data involves analyzing physiological values ​​of the target physiological parameters from different wearable devices at the same point in time, based on time alignment. Individual data points that significantly deviate from the consistent trend or reasonable physiological range formed by the majority of current physiological values ​​are identified as outlier data, marked as invalid, and then removed.

[0108] After the above time alignment and outlier removal, the data that is precisely synchronized in time and consistent in value is defined as the second physiological data. This achieves a high degree of consistency of multi-source data at the signal level, which not only eliminates the slight time difference between wearable devices, but also automatically filters out erroneous values ​​caused by the instantaneous error of a single wearable device.

[0109] Step S23: Cross-validate the second physiological data that characterize the same target physiological parameter. After obtaining the cross-validation results corresponding to each second physiological data, assign confidence weights to each second physiological data according to the cross-validation results to obtain the third physiological data.

[0110] The specific operation of cross-validation is to perform consistency analysis on multiple second physiological data representing the same target physiological parameter from different wearable devices within a set sliding time window. By calculating the correlation coefficient, dynamic time warping distance, or directly calculating the standard deviation or standard variance between the second physiological data within the sliding time window, the consistency or dispersion between the second physiological data is quantified, thereby generating a cross-validation result for each second physiological data. The cross-validation result is usually a consistency score that indicates the degree of agreement between the corresponding second physiological data and the overall trend of other second physiological data.

[0111] Next, the results of the cross-validation are normalized, for example by using the softmax function or a simple proportional allocation. The second physiological data with a high consistency score is given a higher confidence weight, while the second physiological data with a low consistency score, that is, the second physiological data with a large difference from other consistency scores, is given a lower confidence weight. Finally, each second physiological data is given a quantified confidence weight to form the third physiological data.

[0112] Step S24: Based on the confidence weight of each third physiological data, the third physiological data are fused and information is combined to obtain physiological state information.

[0113] The fusion and information composite operation based on the confidence weights of each third physiological data point aims to integrate multiple third physiological data points with varying reliability into a comprehensive physiological state information, thereby providing input for the final classification of health events. Specifically, the confidence weight fusion operation assigns greater weight to highly reliable third physiological data to reduce the impact of low-reliability data, thus optimizing homogeneous physiological indicators and obtaining a stable and accurate weighted fusion intermediate result. The subsequent information composite operation aims to integrate the optimized weighted fusion intermediate result with correlation features extracted from complementary third physiological data. This transforms validated and weighted physiological data into an internally consistent, multi-dimensional, and deeply informative physiological state information, improving the completeness, accuracy, and interpretability of the data relied upon by the subsequent pre-defined hierarchical artificial intelligence model for analysis and decision-making.

[0114] Furthermore, step S24 may also include steps S241 to S242:

[0115] Step S241: Based on the confidence weight of each third physiological data, perform weighted fusion calculation on the third physiological data describing the same physiological indicator to obtain the weighted fusion intermediate result of each physiological indicator.

[0116] It should be noted that physiological indicators refer to physiological parameters such as heart rate, blood oxygen saturation, or respiratory rate.

[0117] We collect all third-party physiological data describing the same physiological indicator from different wearable devices. At each identical, precisely aligned time point, we extract the physiological values ​​corresponding to the third-party physiological data from each wearable device at that time point and their associated confidence weights. Then, we calculate the weighted average using a weighted average algorithm: multiply the physiological values ​​from all wearable devices at that time point by their respective confidence weights, sum the resulting weighted values, and then divide by the sum of all weights to obtain the weighted fusion intermediate result of the physiological indicator at that time point. We repeat this operation for each time point in the time series to obtain the weighted fusion intermediate result of each physiological indicator.

[0118] This step reduces the influence of third-party physiological data with poor consistency and questionable reliability in cross-validation on the final result, while increasing the contribution of highly consistent and reliable data, thus generating a weighted fusion intermediate result that is more stable, accurate and reliable than any single wearable device data.

[0119] Step S242: After extracting physiological features associated with the weighted fusion intermediate results from the complementary arbitrary third physiological data, information is combined on each physiological indicator and each physiological feature based on the weighted fusion intermediate results and physiological features to generate physiological state information.

[0120] It should be noted that the physiological features in this embodiment refer to complementary third physiological data that are different from the physiological indicators represented by the weighted fusion intermediate results but have physiological correlations. For example, in accelerometer data and skin conductance data from the same time period, derivative parameters that can describe the physiological state are calculated by a specific algorithm, such as the activity intensity level feature extracted from acceleration data or the vascular elasticity index feature extracted from pulse wave morphology.

[0121] Within a time window perfectly aligned with the intermediate results of the weighted fusion, multi-dimensional feature extraction is performed on the complementary third physiological data. Specifically, in terms of time-domain statistics, the mean of the complementary third physiological data sequence within this time window is calculated to reflect its average amplitude level; the variance is calculated to quantify its dispersion around the mean to characterize the fluctuation intensity; the root mean square is calculated to obtain its effective amplitude to measure the overall signal energy; and the zero-crossing rate, i.e., the frequency at which data points cross zero or the mean, is calculated to characterize the speed of its changes. These indicators collectively describe the amplitude distribution and fluctuation characteristics of the data sequence from a statistical perspective. Next, in terms of frequency domain transformation, a Fast Fourier Transform is applied to the data sequence to convert it from a time-domain representation to a frequency-domain representation. The frequency component with the largest amplitude is then identified from the resulting spectrum as the dominant frequency to characterize its dominant rhythm. The spectral entropy is calculated to measure the disorder of the energy distribution of the spectrum to reflect the regularity of the rhythm. Furthermore, numerical... The ratio of energy in a specific physiologically relevant frequency band (e.g., 0.1-0.3 Hz corresponding to human gait) to the total energy of the entire frequency band is calculated by integral calculation. This quantifies the intensity proportion of the specific rhythmic component, thereby comprehensively characterizing the periodicity and rhythmic information contained in the data sequence. Finally, in terms of morphological analysis, for data sequences with clear waveforms (e.g., pulse waves), morphological features are extracted by detecting key points on the waveform. For example, the rise time is obtained by locating the starting point and peak point of the rising edge of the waveform and calculating their time difference; the fall time is obtained by locating the time difference from the peak point to the subsequent baseline or trough point; and the waveform area is obtained by numerically integrating the data points within a complete waveform cycle. These parameters, which are directly derived from the waveform geometry, are morphological features closely related to specific physiological mechanisms (e.g., vascular elasticity, cardiac pumping efficiency). Among these, physiological features include the amplitude distribution and fluctuation characteristics of the data sequence, the periodicity and rhythmic information of the data sequence, and morphological features.

[0122] After generating a series of physiological features, the Pearson correlation coefficient between each feature sequence and its corresponding weighted fusion intermediate result sequence is calculated simultaneously within the same time window. This Pearson correlation coefficient is obtained by dividing the covariance of the two data sequences by the product of their respective standard deviations. Its absolute value measures the strength of the linear correlation. Simultaneously, the p-value of the Pearson correlation coefficient is calculated to assess statistical significance. Finally, based on preset dual thresholds (e.g., the absolute value of the Pearson correlation coefficient must be greater than 0.5 and the p-value less than 0.05), all physiological features are screened. Only those physiological feature sequences that have a statistically significant and sufficiently strong correlation with the fused core physiological indicators are retained as effective input for subsequent information compositing, thereby ensuring that the composited information has clear physiological relevance and interpretative value.

[0123] One or more relevant physiological feature sequences (such as activity intensity feature sequences and vascular elasticity index feature sequences) selected above are strictly aligned in time with the weighted fusion intermediate result sequences of one or more physiological indicators (such as heart rate and respiratory rate fusion sequences) to ensure that these sequences share the exact same sampling time points or time windows, thus forming a multidimensional time series matrix where rows represent time points and columns represent different features and indicators. This multidimensional time series matrix is ​​then fed into a pre-defined composite model. Structurally, this pre-defined composite model can be a lightweight neural network (e.g., a multilayer perceptron with several fully connected layers) or a multi-input regression model (e.g., support vector regression or a multi-task learning linear model). Before deployment, the composite model is trained using historical data. The training data consists of similar, aligned multidimensional sequences (containing intermediate results of various physiological indicators and physiological features extracted from complementary data) as input, and multidimensional reference vectors that comprehensively reflect physiological states and are annotated by experts or measured by gold standard equipment as supervision labels. During training, the pre-set composite model automatically learns the complex nonlinear mapping relationship between the input multidimensional time series matrix and the output multidimensional reference vector through optimization algorithms (such as gradient descent). In practical applications, the trained pre-set composite model performs forward computation on the real-time input aligned multidimensional time series matrix, and its output is an integrated, multidimensional vector or structured data object, i.e., physiological state information.

[0124] The p-value is a probability value calculated through statistical hypothesis testing. It is used to quantify the likelihood that the hypothesis that "the observed correlation is purely due to random chance" is true. The smaller the p-value (usually less than a preset threshold such as 0.05), the less likely the calculated Pearson correlation coefficient is to be caused by random factors, and thus the correlation is considered to be statistically significant.

[0125] In one feasible implementation, refer to Figure 5As shown, step S30 may specifically include steps S31 to S33:

[0126] Step S31: Parallel multi-parameter abnormal pattern matching and preliminary risk assessment of physiological state information are performed using the primary screening sub-model to obtain a preliminary set of abnormal indicators and the confidence level of physiological state information.

[0127] It should be noted that the primary screening sub-model is a trained, lightweight artificial intelligence model with low computational complexity, such as an ensemble learning model based on random forest or a simple feedforward neural network, designed to achieve high processing speed and low energy consumption.

[0128] The primary screening sub-model receives physiological state information as input, which includes fused values ​​of multiple physiological indicators and physiological features extracted from complementary data. This sub-model simultaneously analyzes all input values ​​and physiological features in real time, rapidly comparing them with pre-stored normal patterns and various typical abnormal patterns learned from large datasets. For example, it matches the current combination of elevated heart rate, rapid breathing, and changes in skin conductance with normal exercise patterns and early fever abnormal patterns, respectively, and outputs the matching results. Based on these matching results, the primary screening sub-model performs a preliminary risk assessment, calculating the probability that the current input state belongs to various abnormal patterns and marking the main physiological indicators associated with abnormal patterns whose matching probabilities exceed a preset threshold, forming a preliminary set of abnormal indicators.

[0129] Simultaneously, the primary screening sub-model assesses the quality of the input physiological state information by analyzing the data completeness and internal consistency of each component dimension. This includes checking for missing core physiological indicators, whether physiological features extracted from complementary data are within reasonable ranges, and whether there are physiological contradictions in the short-term trends of different physiological indicators. At the same time, the primary screening sub-model evaluates its own certainty in pattern matching by analyzing the confidence index of its internal decision-making process. For example, in a random forest model, this is represented by the consistency of votes from all decision trees for the classification result; in a neural network, it is represented by the maximum value and entropy of the softmax probability distribution in the output layer (the more concentrated the distribution and the lower the entropy, the higher the certainty). Next, the primary screening sub-model combines the evaluation score representing the quality of the input physiological state information with the evaluation score representing the certainty of the primary screening sub-model's own decisions using a pre-defined fusion function (e.g., a weighted average or a small regression sub-network). The weight parameters of this fusion function are learned during model training to ensure that the final output comprehensive confidence value accurately reflects the overall reliability of the screening results.

[0130] Step S32: Activate and call the deep analysis sub-model, and combine it with the user's historical physiological state baseline to perform in-depth analysis on the physiological indicators related to the preliminary abnormal indicator set, so as to obtain the health event identification results and the result level corresponding to each health event identification result.

[0131] It should be noted that the deep analysis sub-model is an artificial intelligence model with a more complex structure, more parameters, and deeper analytical dimensions than the primary screening sub-model. For example, it is a deep learning model that includes a long short-term memory network layer or an attention mechanism. Its design goal is to perform refined identification and attribution of potential anomalies. It is normally in a dormant state to save power consumption and is activated only when the set of preliminary anomaly indicators is not empty and the confidence level of its accompanying physiological state information exceeds a preset threshold.

[0132] The user's historical physiological baseline refers to a personalized reference model derived from statistical modeling based on long-term, continuous monitoring of the user's historical data. It reflects the normal fluctuation range of various physiological indicators and characteristics and circadian rhythms under normal conditions.

[0133] In the specific deep analysis, the deep analysis sub-model first receives a preliminary set of abnormal indicators as input. It then extracts other physiological indicators and features highly correlated with the abnormal indicators from the complete physiological state information, forming a focused analysis dataset. The deep analysis sub-model compares this focused current data with the individual's historical normal range extracted from the user's historical physiological state baseline for the same time period, analyzing the degree, duration, and trend of deviation from the individual baseline. Based on this, the deep analysis sub-model utilizes its internally learned complex patterns, such as ECG waveform feature sequences of specific arrhythmias, and specific combinations of heart rate variability and skin conductance in the pre-hypoglycemic stage, to perform comprehensive reasoning on the focused analysis dataset. This not only determines whether the abnormality is real but also categorizes it into specific health event identification results, such as paroxysmal atrial fibrillation or orthostatic hypotension.

[0134] Meanwhile, the deep analysis sub-model assesses the severity of deviations from an individual's baseline based on the magnitude, duration, and trend of change, assigning a quantitative result level to each identification result, such as Level 1 being mild and requiring observation, and Level 3 being severe and requiring urgent response.

[0135] Step S33: The identification results and result levels of each health event are mapped through a preset risk grading strategy mapping table to generate graded health event judgment results.

[0136] It should be noted that the preset risk grading strategy mapping table is a set of decision rules predefined and stored in the monitoring device's system. This preset risk grading strategy mapping table uses health event types, such as atrial fibrillation and hypoglycemia, and outcome levels as key indices for joint input. Each event type-level combination corresponds to a specific, predefined response strategy. This response strategy at least includes explicit action suggestions, notification priorities, and resource scheduling instructions. In specific operation, the system receives the health event identification result and its corresponding outcome level as input, then searches and matches it in the preset risk grading strategy mapping table to locate the entry that completely corresponds to the input combination. It then extracts all the response strategy fields defined in that entry, encapsulates and integrates the extracted strategy fields with the input health event identification result and outcome level to form a graded health event judgment result. This establishes an automated and standardized bridge from intelligent medical analysis conclusions to clear and executable instructions, ensuring that the system can generate consistent, accurate, and appropriate graded response decisions for different types and severity of health events. This transforms complex analysis outputs into a clear and reliable action blueprint that drives subsequent collaborative response operations.

[0137] In one feasible implementation, refer to Figure 6 As shown, step S40 may specifically include steps S41 to S44:

[0138] Step S41: By parsing the health event type, risk level, and corresponding response measures contained in the graded health event judgment results, response strategy parameters are obtained.

[0139] Receive the graded health event judgment result from step S33. The result is a structured data object that explicitly includes the health event type (i.e., the specific medical event identified and classified by the model, such as "atrial fibrillation" or "hypoglycemia trend"), risk level (i.e., the quantitative severity of the event, such as "Level 2: Moderate risk"), and corresponding response measures (i.e., action recommendations determined according to the preset strategy mapping table and bound to the type and level, such as "start continuous electrocardiogram monitoring and issue a local alert").

[0140] The parsing process is executed by a rule engine within the system responsible for policy transformation. This rule engine identifies and transforms each field of the input object based on a predefined parsing rule table. For example, it maps "health event type: atrial fibrillation" to the policy parameter target_event, "risk level: level 2" to the policy parameter urgency_level, and breaks down "corresponding response measures: initiate continuous ECG monitoring and issue local alerts" into a series of specific policy parameters, such as data_collection_mode:continuous_ecg and alert_level:local_warning. Finally, the extracted and transformed policy parameters representing specific action dimensions constitute the response policy parameters, ensuring the accuracy and consistency of the entire system's response actions.

[0141] Step S42: Based on the response strategy parameters and combined with the data acquisition capability description of each wearable device, assign corresponding task roles to each wearable device.

[0142] It should be noted that a task role refers to a series of abstract positions with specific functions that are predefined based on the overall needs of the coordinated response action, such as "main alarm", "high-precision verification data collector", "auxiliary information prompter" and "external communication repeater".

[0143] In specific assignment, the first step is to determine which roles are required for the response and the required number of each role based on the response strategy parameters. Then, the data acquisition capability description of each wearable device is queried. This description includes the device's real-time capability score, available sensor types, remaining battery power, and current network role. The matching degree is calculated between each wearable device in the candidate device pool and the role to be assigned. The core of this process is to compare the capability requirements of the role in the response strategy parameters (e.g., the "high-precision verification data collector" role requires the device to have high-precision ECG monitoring capabilities and sufficient battery power) with the corresponding quantitative indicators in the wearable device's data acquisition capability description. Priority is given to selecting the wearable device with the highest current matching degree for each role. Simultaneously, it ensures that a wearable device assumes only one primary role in a single response to concentrate resources. If multiple wearable devices meet the requirements for the same role, the best device is selected based on its contribution weight in the capability description. This achieves dynamic intelligent matching between the response strategy parameters and wearable device resources, enabling each collaborative response action to automatically form an optimal task execution team based on the latest monitoring network, with the most suitable wearable devices fulfilling their respective roles. This ensures optimal resource utilization and reliable execution of the response action at the system level.

[0144] Step S43: Based on the task role, generate an instruction set that matches the device function of the wearable device, and then distribute the instruction set to the corresponding wearable device.

[0145] First, a predefined role-instruction template mapping library is accessed. This library stores corresponding instruction templates for each possible task role. When an instruction is generated for a wearable device, the corresponding instruction template is retrieved from the role-instruction template mapping library based on its task role. Then, combined with the specific sensor sampling rate, maximum alarm volume, screen type, etc. supported by the device model recorded in the wearable device's data acquisition capability description, the parameters in the instruction template are instantiated and populated. For example, an instruction containing the specific parameters "vibrate continuously at maximum intensity and display a red warning icon and text on the screen" is generated for a smartwatch acting as the "main alarm". For a smart chest patch acting as the "high-precision data collector", an instruction containing the specific parameters "start the ECG monitoring module and work continuously at a sampling rate of 500Hz for 300 seconds" is generated. This generates a set of instructions that is fully adapted to the hardware capabilities of the wearable device.

[0146] After the instruction set is generated, it is distributed point-to-point or via broadcast to the corresponding wearable devices in the form of encrypted data packets through the established local wireless network (such as Bluetooth or StarFlash) using a low-power communication protocol. This lays a direct operational foundation for the coordinated physical response of multiple wearable devices under unified scheduling, enabling them to work in unison and perform their respective functions.

[0147] Step S44: According to the instruction set, schedule and control each wearable device to execute the corresponding preset task action to complete the collaborative response operation.

[0148] After confirming that all wearable devices have successfully received the instruction set, the system first sends synchronous execution commands to each wearable device or triggers each wearable device to start its task according to a preset order, based on the execution timing logic or dependencies embedded in the instruction set, through a unified coordination protocol. Upon receiving the execution command, each wearable device's local control unit immediately parses the specific action parameters (such as alarm type, intensity, duration, or data acquisition mode and sampling rate) assigned to it in the instruction set, and drives the corresponding hardware modules (such as vibration motors, displays, speakers, or sensors) to execute preset task actions. During this process, the coordination center monitors the execution status feedback of each wearable device in real time and dynamically adjusts the scheduling instructions as needed to ensure the synchronization and complementarity of actions. For example, it ensures that alarm prompts and high-precision data acquisition are launched simultaneously, or that a backup wearable device is immediately activated to take over its role when a wearable device fails. In this way, the detailed response strategies and instruction plans formulated in the early stage are transformed into a coordinated physical response action by multiple wearable devices that are synchronized in time, complementary in function, and coherent in logic through precise timing scheduling and reliable device control. This ensures that the response to health events is timely, orderly, and efficient as a whole, and ultimately completes the entire closed loop from intelligent perception, analysis and decision-making to collaborative execution.

[0149] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A wearable human safety monitoring device, characterized in that, include: The data acquisition module is used to coordinate the various wearable devices and collect physiological data from each wearable device. The data verification module is used to verify and fuse the physiological data to obtain physiological state information; The data reasoning module is used to input the physiological state information into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment, and to obtain the graded health event judgment result; The execution module is used to schedule each wearable device to perform a collaborative response operation that matches the graded health event judgment result, based on the graded health event judgment result. The preset hierarchical artificial intelligence model includes a primary screening sub-model and a deep analysis sub-model. The step of inputting the physiological state information into the preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain the graded health event judgment result includes: The physiological state information is subjected to parallel multi-parameter abnormal pattern matching and preliminary risk assessment through the primary screening sub-model to obtain a preliminary abnormal indicator set and the confidence level of the physiological state information, wherein the confidence level is used to reflect the reliability of the preliminary abnormal indicator set. The deep analysis sub-model is activated and invoked, and combined with the user's historical physiological state baseline, the physiological indicators related to the preliminary abnormal indicator set are analyzed in depth to obtain the health event identification results and the result level corresponding to each of the health event identification results. The identification results of each health event and the result level are mapped through a preset risk grading strategy mapping table to generate the graded health event judgment result; The step of scheduling each wearable device to perform a collaborative response operation matching the graded health event judgment result includes: By analyzing the health event type, risk level, and corresponding response measures contained in the graded health event judgment results, response strategy parameters are obtained; Based on the response strategy parameters and combined with the data acquisition capability description of each wearable device, a corresponding task role is assigned to each wearable device. Based on the task role, after generating an instruction set that matches the device function of the wearable device, the instruction set is distributed to the corresponding wearable device; According to the instruction set, each wearable device is scheduled and controlled to perform corresponding preset task actions to complete the collaborative response operation.

2. A wearable human body safety monitoring method, characterized in that, include: Coordinate various wearable devices to collect physiological data from each wearable device; The physiological data are verified and fused to obtain physiological state information; The physiological state information is fed into a preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain a graded health event judgment result. Based on the graded health event assessment results, each wearable device is scheduled to perform a collaborative response operation that matches the graded health event assessment results; The preset hierarchical artificial intelligence model includes a primary screening sub-model and a deep analysis sub-model. The step of inputting the physiological state information into the preset hierarchical artificial intelligence model for collaborative reasoning and state judgment to obtain the graded health event judgment result includes: The physiological state information is subjected to parallel multi-parameter abnormal pattern matching and preliminary risk assessment through the primary screening sub-model to obtain a preliminary abnormal indicator set and the confidence level of the physiological state information, wherein the confidence level is used to reflect the reliability of the preliminary abnormal indicator set. The deep analysis sub-model is activated and invoked, and combined with the user's historical physiological state baseline, the physiological indicators related to the preliminary abnormal indicator set are analyzed in depth to obtain the health event identification results and the result level corresponding to each of the health event identification results. The identification results of each health event and the result level are mapped through a preset risk grading strategy mapping table to generate the graded health event judgment result; The step of scheduling each wearable device to perform a collaborative response operation matching the graded health event judgment result includes: By analyzing the health event type, risk level, and corresponding response measures contained in the graded health event judgment results, response strategy parameters are obtained; Based on the response strategy parameters and combined with the data acquisition capability description of each wearable device, a corresponding task role is assigned to each wearable device. Based on the task role, after generating an instruction set that matches the device function of the wearable device, the instruction set is distributed to the corresponding wearable device; According to the instruction set, each wearable device is scheduled and controlled to perform corresponding preset task actions to complete the collaborative response operation.

3. The wearable human safety monitoring method as described in claim 2, characterized in that, The step of coordinating various wearable devices and collecting physiological data from each wearable device includes: Register each of the wearable devices to the same monitoring network to obtain the device attribute information of each wearable device; Based on the attribute information of each device, the sensor type, current battery level and wearing position of the corresponding wearable device are evaluated to obtain a description of the data acquisition capability of each wearable device. Based on the data acquisition capability descriptions of each device, corresponding data acquisition tasks are assigned to each wearable device to obtain a data acquisition instruction set. According to the data acquisition instruction set, the corresponding wearable device is scheduled to synchronously acquire the sensing signals on the sensors on the wearable device, and the sensing signals are preprocessed by local timestamp correspondence and formatting to obtain the physiological data.

4. The wearable human safety monitoring method as described in claim 3, characterized in that, The step of evaluating the sensor type, current battery level, and wearing position of the corresponding wearable device based on the attribute information of each device to obtain a description of the data acquisition capability of each wearable device includes: Based on the attribute information of each device, the sensor type, current battery level, and wearing position of each wearable device are evaluated to obtain the sensor capability evaluation result, battery status evaluation result, and location confidence evaluation result of each wearable device. Based on the sensor capability assessment results, the battery status assessment results, and the location confidence assessment results, calculate the comprehensive confidence score of each wearable device under the preset physiological parameter monitoring task; Based on the comprehensive confidence score and combined with historical data acquisition quality records, the role and contribution weight of each wearable device in the current monitoring network are calibrated to obtain the data acquisition capability description of each wearable device.

5. The wearable human safety monitoring method as described in claim 2, characterized in that, The step of verifying and fusing the physiological data to obtain physiological state information includes: Noise filtering and signal integrity analysis of each physiological data point are performed to obtain each first physiological data point. The first physiological data representing the same target physiological parameter are time-aligned, outlier data are identified and removed, and the second physiological data is obtained. Cross-validation is performed on the second physiological data that characterizes the same target physiological parameter. After obtaining the cross-validation results corresponding to each second physiological data, a confidence weight is assigned to each second physiological data according to the cross-validation results to obtain the third physiological data. Based on the confidence weights of each of the third physiological data, the third physiological data are fused and information is combined to obtain the physiological state information.

6. The wearable human safety monitoring method as described in claim 5, characterized in that, The step of fusing and combining information from the third physiological data based on the confidence weights of each third physiological data to obtain the physiological state information includes: Based on the confidence weight of each of the third physiological data, the third physiological data describing the same physiological indicator are weighted and fused to obtain the weighted fusion intermediate result of each physiological indicator; After extracting physiological features associated with the weighted fusion intermediate result from complementary arbitrary third physiological data, the physiological indicators and physiological features are combined according to the weighted fusion intermediate result and the physiological features to generate the physiological state information.

Images