Cross-device health data fusion method

Abstract

The invention relates to the technical field of health data processing, in particular to a cross-device health data fusion method, which comprises the following steps of monitoring a device connection state, identifying data missing and resending, calibrating multi-device time and performing data difference, calculating a device stability score to adjust confidence fusion data, extracting periodic health parameters and analyzing a change trend. And detecting the factor activation state and combining the sensitivity to obtain a risk score, and outputting a health processing result. According to the method, through combination of equipment connection state identification and data caching progress calculation, the precision of breakpoint data recovery is enhanced, a timestamp alignment and interpolation correction mechanism is adopted, continuous compensation of multi-source data is realized, the data weight is adjusted by using a stability score, the credibility of fused data is improved, and the reliability of the fused data is improved. In combination with behavior characteristic trend extraction and risk factor sensitivity determination, the accuracy and individualized distinguishing ability of chronic disease risk identification are optimized, and the individualized health management and disease prevention ability is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of health data processing, and particularly to a cross-device health data fusion method. Background Art

[0002] The technical field of health data processing includes methods and technologies for dynamically collecting, standardizing and systematically managing multi-dimensional health information generated by individuals during daily life and medical management. The core content of this technical field lies in constructing a comprehensive health data system through real-time monitoring and processing of physiological parameters, behavioral characteristics and environmental data, improving the ability of health risk identification and management. The overall technical system covers cross-device data collection standards, heterogeneous data format conversion, time series synchronization, extraction of key health indicators and data security management mechanisms, supports continuous tracking of health status and chronic disease management, and is widely used in disease early warning, health intervention and remote medical services, playing an important role in chronic disease prevention and control and personalized health management.

[0003] Among them, the cross-device health data fusion method refers to multi-source heterogeneous health data obtained from intelligent wearable devices, mobile health terminals and auxiliary monitoring devices. Depending on the device identification protocol to confirm the data source, standardize physiological and behavioral parameters such as heart rate, blood pressure, blood sugar, blood oxygen and exercise behavior output by different devices through unified data format conversion rules, adopt a time alignment mechanism to achieve the temporal consistency of cross-device data, perform multi-dimensional association on chronic disease-related risk factors and stroke warning indicators in combination with a preset data integration process, and store the fused health data in a centralized management platform to achieve stroke risk monitoring, medication management, rehabilitation guidance and implementation and management of various health intervention measures.

[0004] Traditional cross-device health data fusion technologies lack an efficient data missing compensation mechanism and time offset calibration measures when processing multi-source data, resulting in problems of inconsistent time series and data breakpoints during the data integration process, limiting the depth and breadth of data analysis, leading to delays or misjudgments in health risk early warning, lacking the ability to dynamically adjust for differences in individual device signal quality during data fusion, limiting data accuracy and the personalization of health management. When facing diverse and large-scale user groups, it causes health intervention measures to not accurately meet the actual needs of users, affecting the health management effect and user satisfaction. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art and propose a cross-device health data fusion method.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A cross-device health data fusion method, including the following steps: S1: Monitor the device connection status parameters. By detecting the device reconnection request, call the data capture progress parameter and cache pointer position parameter in the device data cache, calculate the data missing interval amount in combination with the current data sequence number, and send a data reissuance request to generate a breakpoint resumption compensation value; S2: Call the breakpoint resumption compensation value, obtain the acquisition data sequences of multiple devices. By comparing the local timestamp and the reference timestamp of each device, calculate the time offset and perform calibration, identify the missing values in the data sequence and perform data interpolation to generate an interpolation compensation result; S3: Call the interpolation compensation result, extract the signal interruption times, signal abnormal fluctuation times parameter, and sampling interval offset parameter of each device, calculate the stability score of the device, adjust the data confidence weight corresponding to each device and fuse the same type of data to obtain a fused data set; S4: Call the fused data set, extract the walking amount parameter, sitting duration parameter, sleep duration parameter, and heart rate fluctuation parameter, calculate the change rate characteristics of each parameter in multiple cycles, and analyze the time series change trend of the rate characteristics to generate a health data feature set.

[0007] As a further solution of the present invention, the breakpoint resumption compensation value is specifically the data missing interval amount, reissuance request instruction, and data capture progress positioning information. The interpolation compensation result includes the time offset calibration value, data sequence missing segment identifier, and interpolated generated data points. The fused data set is specifically the weighted fusion parameter value, device confidence mapping table, and unified time series data structure. The health data feature set includes the cycle change rate matrix, trend change direction feature, and continuous fluctuation range value.

[0008] As a further solution of the present invention, the steps for obtaining the breakpoint resumption compensation value are specifically as follows: S111: Monitor the device connection status parameters, real-time identify the connection status of the device, including connection and disconnection, real-time detect the device reconnection request, obtain the reconnection request identifier and the current device connection status code, and generate a device reconnection status identifier value; S112: Call the device reconnection status identifier value, extract the data capture progress parameter and cache pointer position parameter in the device data cache, calculate the offset between the data capture progress parameter and the cache pointer position parameter, and generate a cache offset value; S113: Based on the cache offset value, call the current data sequence number and the last valid data sequence number, calculate the difference between the current data sequence number and the last valid data sequence number, obtain the data missing interval amount, and send a data reissuance request to generate a breakpoint resumption compensation value.

[0009] As a further solution of the present invention, the steps for obtaining the interpolation compensation result are specifically as follows: S211: Invoke the breakpoint resuming compensation value, obtain the acquisition data sequences of multiple devices, extract the device numbers and timestamps corresponding to each device, and generate a set value of device acquisition sequences; S212: Based on the set value of device acquisition sequences, calculate the difference between the local timestamp and the reference timestamp of each device, obtain the time offset, and correct the acquisition data sequence of each device to generate a time calibration sequence value; S213: Invoke the time calibration sequence value, identify the missing values in the continuous data time axis, extract the adjacent data points of the missing values, calculate the values of the interpolation points, perform data interpolation on the acquisition data sequence, and generate an interpolation compensation result.

[0010] As a further solution of the present invention, the steps for obtaining the fusion data set are specifically as follows: S311: Invoke the interpolation compensation result, extract the signal interruption times parameter, signal abnormal fluctuation times parameter, and sampling interval offset parameter of each device, and generate a set of basic device stability parameters; S312: According to the set of basic device stability parameters, extract the signal interruption times and abnormal fluctuation times of each device in each data segment, obtain the sampling interval offset and signal duration parameters, and invoke the device data segment number parameter to calculate the device stability score value; S313: Invoke the device stability score value, adjust the corresponding data confidence weights according to the score values of each device, perform data fusion on the same type of data, and obtain the fusion data set.

[0011] As a further solution of the present invention, the steps for obtaining the health data feature set are specifically as follows: S411: Invoke the fusion data set, extract the walking amount parameter, sedentary duration parameter, sleep duration parameter, and heart rate fluctuation parameter within multiple time periods, and generate a set of periodic health parameters; S412: Based on the set of periodic health parameters, extract the numerical change conditions and periodic time points of each type of health parameter in continuous periods, construct a periodic number sequence and the total number of periods, and calculate the average change rate of multiple health parameters in the periodic sequence to obtain a periodic rate feature sequence value; S413: Invoke the periodic rate feature sequence value, arrange the rate feature values of each health parameter in chronological order, extract its time trend change pattern and judge the trend range, and obtain the health data feature set.

[0012] As a further solution of the present invention, the method further includes: S5: Invoke the health data feature set to obtain the periodic change rate feature values and time series fluctuation feature values of multiple parameters. By comparing with the preset chronic disease risk factor feature thresholds, identify the activation status of risk factors, and combine the sensitivity parameters of multiple risk factors to calculate the chronic disease risk score of the user and obtain the health data processing result; The health data processing result specifically refers to the chronic disease risk score value, the type of activated risk factors, and the combination of sensitivity impact parameters.

[0013] As a further solution of the present invention, the steps for obtaining the health data processing result are specifically as follows: S511: Invoke the health data feature set, extract the periodic change rate feature values and time series fluctuation feature values of multiple health parameters, and compare with the preset chronic disease risk factor feature thresholds to identify and mark the activation status of each risk factor and obtain the activation status comparison result; S512: According to the activation status comparison result, by identifying the activation status of the risk factors of each health parameter, extract the sensitivity parameters of each activated risk factor, evaluate the activation intensity of each parameter, and generate activation intensity data; S513: Invoke the activation intensity data, calculate the chronic disease risk score of the user by calculating the risk score of each activated factor, and output the health data processing result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by combining device connection status recognition and data cache progress calculation, the accuracy of breakpoint data recovery is enhanced. By adopting a timestamp alignment and interpolation correction mechanism, the continuity compensation of multi-source data is realized. By using a stability score to adjust data weights, the credibility of the fused data is improved. By combining behavior feature trend extraction and risk factor sensitivity determination, the accuracy of chronic disease risk identification and the ability of individualized differentiation are optimized, and the ability of personalized health management and disease prevention is enhanced. Description of the Drawings

[0015] Figure 1 It is a schematic diagram of the main steps of the present invention; Figure 2 It is a flowchart for obtaining the breakpoint continuation compensation quantity value of the present invention; Figure 3 It is a flowchart for obtaining the interpolation compensation result of the present invention; Figure 4 It is a flowchart for obtaining the fused data set of the present invention; Figure 5 It is a flowchart for obtaining the health data feature set of the present invention; Figure 6 It is a flowchart for obtaining the health data processing result of the present invention. Detailed implementation manners

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0017] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation on the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0018] Please refer to Figure 1 , the present invention provides a technical solution: a cross-device health data fusion method, including the following steps: S1: Monitor the device connection status parameters, detect the device reconnection request, call the data capture progress parameter and the cache pointer position parameter in the device data cache, calculate the data missing interval amount in combination with the current data serial number, and send a data reissuance request to generate a breakpoint resumption compensation value; S2: Call the breakpoint resumption compensation value, obtain the acquisition data sequences of multiple devices, calculate the time offset and perform calibration by comparing the local timestamp and the reference timestamp of each device, identify the missing values in the data sequence and perform data interpolation to generate an interpolation compensation result; S3: Call the interpolation compensation result, extract the signal interruption times, signal abnormal fluctuation times parameter, and sampling interval offset parameter of each device, calculate the stability score of the device, adjust the corresponding data confidence weight of each device and fuse the same type of data to obtain a fused data set; S4: Call the fused data set, extract the walking amount parameter, sedentary duration parameter, sleep duration parameter, and heart rate fluctuation parameter, calculate the change rate characteristics of each parameter in multiple cycles, and analyze the time series change trend of the rate characteristics to generate a health data feature set; S5: Call the health data feature set, obtain the periodic change rate characteristic values and time series fluctuation characteristic values of multiple parameters, compare with the preset chronic disease risk factor characteristic thresholds, identify the activation status of the risk factors, and combine the sensitivity parameters of multiple risk factors to calculate the chronic disease risk score of the user to obtain the health data processing result.

[0019] The breakpoint resumption compensation value specifically includes the data missing interval amount, the retransmission request instruction, and the data capture progress positioning information. The interpolation compensation result includes the time offset calibration value, the data sequence missing segment identifier, and the interpolated generated data points. The fused data set specifically includes the weighted fusion parameter value, the device confidence mapping table, and the unified time series data structure. The health data feature set includes the periodic change rate matrix, the trend change direction feature, and the continuous fluctuation range value. The health data processing result specifically refers to the chronic disease risk score value, the activated risk factor type, and the sensitivity influence parameter combination.

[0020] Please refer to Figure 2 , and the steps for obtaining the breakpoint resumption compensation value are specifically as follows: S111: Monitor the device connection status parameter, identify the connection status of the device in real time, including connection and disconnection, detect the device reconnection request in real time, obtain the reconnection request identifier and the current device connection status code, and generate the device reconnection status identifier value; Number each monitoring device identifier connected to the network. For example, devices A, B, and C are numbered 1, 2, and 3 respectively. When the system background detects a new device access or disconnection event, call the connection status detection module to read the real-time identification bit information of the device status register, obtain the current device connection status code. The status code of 0 indicates not connected, the status code of 1 indicates connecting, the status code of 2 indicates stably connected, and the status code of 3 indicates disconnecting. If the device reports a heartbeat packet again within 30 seconds, the system identifies it as a "device reconnection request". At this time, record the device unique identifier (such as MAC address or device number), reconnection timestamp, and reconnection status identifier (for example, changing from 0 to 1) in the heartbeat packet to the system log. Obtain the disconnection duration by performing a time difference operation on the reconnection timestamp and the disconnection timestamp. In the example, device A disconnected at 09:15:03 on April 24, 2025, and resumed connection at 09:15:17. Its disconnection duration is 14 seconds, which is greater than the system-set minimum fault tolerance interval of 5 seconds. It is determined as a valid disconnection event, and the reconnection status identifier value is generated as "R1-14", where "R1" represents the reconnection event sequence number 1, and "14" represents the disconnection duration in seconds.

[0021] S112: Call the device reconnection status identifier value, extract the data capture progress parameter and the cache pointer position parameter in the device data cache, calculate the offset between the data capture progress parameter and the cache pointer position parameter, and generate the cache offset value; Read the data capture progress pointer position and the current cache write pointer position stored in the device buffer area in a circular array structure. The device collects 1 record per second, and the cache space can retain up to 1800 records (i.e., 30 minutes of data). For example, in a certain reconnection event, read the cache write pointer position as , and the data capture progress pointer is , then the calculated offset is ,in, is the cache offset, Write the pointer position for the current cache, The data capture progress pointer position indicates that the device has cached 26 uncaptured data records during the disconnection period. The system uses the offset value The index cache area is read one by one from the 998th item to the 1024th item. Each piece of data is saved in a fixed structure, including timestamp, heart rate value, number of steps, and status flag. During the reading process, the system marks the currently processed cache segment and updates the crawl progress pointer, and finally generates a cache offset value of 26.

[0022] S113: Based on the cache offset value, the current data sequence number and the last valid data sequence number are called, the difference between the current data sequence number and the last valid data sequence number is calculated, the data missing interval is obtained, and a data retransmission request is sent to generate a breakpoint resume compensation value; After the device is disconnected and restored, the system compares its current data sequence number The last valid data sequence number , are 14276 and 14247 respectively, so the amount of missing data is ,in, is the data missing interval, is the data sequence number reported by the current device. is the valid data sequence number captured last time. If the cache offset is 26, it means that there are still 3 data not stored in the cache, that is, 3 data are lost during the actual disconnection. The system will send a retransmission request to the device according to the cache offset, with the request segment sequence number interval [14248,14273] and the data type mask. After receiving the request, the device retrieves the local cache and retransmits the data in the interval. Finally, the receiving end merges the original data sequence with the retransmitted data to form a complete data stream, generating a breakpoint retransmission compensation value of 29, and the compensation rate is , where the numerator 26 is the number of records successfully reissued, and the denominator 29 is the total number of theoretical missing records, indicating that the system successfully recovered most of the data. Table 1 lists the device cache status and compensation result samples.

[0023] Table 1 Device cache offset and resend compensation table As shown in Table 1, the compensation request content and the actual success rate are calculated by the offset and the sequence number difference.

[0024] See also Figure 3 , the specific steps for obtaining the interpolation compensation results are: S211: Call the breakpoint resumption compensation value, obtain the acquisition data sequences of multiple devices, extract the device numbers and timestamps corresponding to each device, and generate the device acquisition sequence set value; Call the breakpoint resumption compensation value, and the system sequentially retrieves the previously recorded breakpoint resumption compensation value array , and locates the device number list corresponding to each compensation interval through the internal indexing mechanism. The device numbers are D1, D2, and D3 respectively. The system verifies the legality of each device through the device identity authentication module, and obtains its original acquisition data within the compensation period in the form of time segments. The acquisition data field structure is {timestamp, heart rate, steps, signal flag}. For each piece of data, the system extracts its timestamp value and pairs it with the device number to which the data belongs, constructs a set of multi-group time series data with the device number as the index, and the time series data is automatically sorted in ascending order of the timestamp to form a set value. For example, device D1 obtains 30 records from 09:15:00 to 09:15:30, and the timestamp interval is 1 second for each record. Then the set form is , and finally the system completes the construction of the set structure and generates the device acquisition sequence set value.

[0025] S212: Based on the device acquisition sequence set value, calculate the difference between the local timestamp and the reference timestamp of each device, obtain the time offset, and correct the acquisition data sequence of each device to generate the time calibration sequence value; Based on the device acquisition sequence set value, the system extracts the local timestamp of each data item in the set and performs a difference operation with the reference timestamp uniformly set by the server . The calculation method is , where is the time offset value, is the device-reported timestamp, is the reference timestamp corresponding to the current receiving moment of the server. If the reporting time of device D1 is 09:15:03 and the server receiving moment is recorded as 09:15:00, the offset is 3 seconds. The system performs the same offset correction on all the time series data of device D1, that is, subtracts 3 seconds from all timestamps to correct the offset, so as to complete the standardization operation of the acquisition time series. The system records the corrected time series structure as , completes the data time standardization processing of all devices, and generates the time calibration sequence value.

[0026] S213: Call the time calibration sequence value, identify the missing values in the continuous data time axis, extract the adjacent data points of the missing values, and use the formula: ; Calculate the value of the interpolation point, perform data interpolation on the acquisition data sequence, and generate the interpolation compensation result; Among them, is the value corresponding to the interpolation point, is the value of the previous valid data point in the missing interval, is the value of the next valid data point in the missing interval, is the timestamp of the previous valid data point in the missing interval, is the timestamp of the next valid data point in the missing interval, is the timestamp of the current interpolation point, is the value of the fine-tuning factor; Call the time calibration sequence value. The system scans all standardized time series and marks the gap segments where the interval between consecutive time points exceeds the preset sampling interval threshold (such as 1 second). If it is found that there is a 2-second interval between the timestamps 09:15:11 and 09:15:13 of device D2, it is confirmed that there is a missing point 09:15:12 in the middle. The system locates the valid data points before and after the missing point as and , and at the same time defines the interpolation point as , the system sets the value of the fine-tuning factor to , and substitutes it into the following formula: ; ; The final interpolation value is 79.4. The system inserts this value into the original time series to construct continuous interpolation compensation data. Other devices perform the same operation, compensating for the missing data segments to form a complete sequence and generating the interpolation compensation result. Table 2 lists the actual data and results of interpolation calculations in multiple devices. The formula is used to automatically repair the null value data in the time series caused by disconnection or sampling failure. First, the value provided by the valid points before and after is used to judge the reasonable position of the current gap point in the state evolution process. For example, if the heart rate jumps from 78 to 80 and the missing point is exactly at the middle time point between the two, the system tends to think that the value of this point is a transitional state on the linear trend. The interpolation main body exactly simulates this linear transition process; at the same time, in reality, the device sampling process is affected by environmental factors or device state fluctuations, which may make the sampling trend slightly asymmetric or fluctuate up and down. Therefore, a fine-tuning factor is introduced to make a trend offset according to the up and down jump amplitude, so that the interpolation result can be dynamically adjusted without destroying the previous and subsequent trends, enhancing the fluency and stability of the overall data sequence.

[0027] Table 2 Interpolation Calculation Result Table As shown in Table 2, the results show that the values of the missing data points are restored after interpolation calculation and can be used for subsequent fusion and fluctuation analysis steps.

[0028] Please refer to Figure 4, the steps for obtaining the fused dataset are specifically as follows: S311: Call the interpolation compensation result, extract the signal interruption times parameter, signal abnormal fluctuation times parameter, and sampling interval offset parameter of each device, and generate a basic device stability parameter set; Call the interpolation compensation result. The system sequentially reads the continuous sampling sequences after interpolation for each device, and performs a signal analysis process on each record. First, identify the signal connection status flag of each segment of data. If the flag changes from "valid" to "disconnected" and then to "valid", then it is counted as one signal interruption within this segment, and the total number of interruptions in all data segments is counted as the signal interruption times parameter , secondly, analyze the amplitude change of each segment of sampling data. If the difference between the values of any two consecutive sampling points in the same segment is greater than the device-set threshold , then it is recorded as one signal abnormal fluctuation, and the cumulative abnormal times are used as the abnormal fluctuation times parameter , thirdly, extract the theoretical sampling interval from the sampling time series seconds, compare the actual time interval between each pair of adjacent sampling points and calculate the mean square error, and the result is recorded as the sampling interval offset , finally, count the time length of each continuous sampling segment in seconds, and record it as the signal duration , combine the above parameters and add the number of data segments divided by this round of acquisition of the device , and finally construct a complete basic device stability parameter set.

[0029] S312: According to the basic device stability parameter set, extract the signal interruption times and abnormal fluctuation times of each device in each segment of data, obtain the sampling interval offset and signal duration parameters, and call the device data segment number parameter, using the formula: ; Calculate the device stability score value; where, is the stability score value of the th device, is the signal interruption times of the th device in the th segment, is the signal abnormal fluctuation times of the th device in the th segment, is the sampling interval offset of the th device, is the signal duration of the th device, is the number of continuous data segments of the th device, is the sampling frequency, is the normalization reference value of the anomaly intensity, is the normalization reference value of the timing perturbation, represents the total number of data segments corresponding to each device, represents the device number index, represents the device data segment number index; According to the device stability basic parameter set, the system traverses each device number , and sequentially extracts the number of signal interruptions in each segment of it and the number of abnormal signal fluctuations , multiplies the elements at the corresponding positions and sums them to form a term . If the number of interruptions recorded in the 3 segments of device D1 is , and the number of abnormal fluctuations is , then the product sum is . Assuming the normalization reference value of the anomaly intensity , then the normalized value of this term is . Continue to extract the sampling interval offset and the signal duration of device D1, and the sampling frequency Hz, calculate its perturbation modulus as . Assuming the normalization reference value of the perturbation , then the normalized value of this term is . The absolute value of the difference between the two terms is . The number of device data segments is . Substitute into the formula: ; ; Among them, the stability score value of the device is a dimensionless value comprehensively calculated based on four indicators: signal interruption frequency, abnormal fluctuation degree, sampling time stability, and data continuous structure, and is used to represent the operation stability degree and data reliability of a certain device during the acquisition process. The higher the score value, the more interruptions, frequent fluctuations, or unstable sampling time the device has, and the worse the stability; the lower the score value, the more stable the device operation, the more complete the data, and the better the time synchronization. This score is used as the adjustment basis for the confidence weight in the subsequent data fusion stage, directly affecting the fusion ratio and credibility evaluation of data from different devices, and is an important basis for controlling data weight allocation and screening and fusing high-quality data. The formula aims to quantitatively describe the quality of device data from multiple dimensions, realize the identification of the behavioral stability of the device, and project it uniformly into a standardized score value, which is convenient for cross-device score ranking and subsequent weighted processing. In the design, the product is used to reflect the coupling strength, the square root is used to process the time influence, the normalization standard is introduced for dimensionless control, the absolute value is used to prevent directional errors, and the continuous segment correction is used to improve the score credibility. The calculation results of other devices in practical applications are shown in Table 3, and finally the device stability score value is obtained.

[0030] Table 3 Equipment Stability Score Calculation Table As shown in Table 3, different equipment yields different stability score results according to their degrees of abnormality and time stability differences.

[0031] S313: Call the equipment stability score value, adjust the corresponding data confidence weight according to the score value of each equipment, perform data fusion on similar data, and obtain a fused data set; Call the equipment stability score value, the system sets the initial confidence of each equipment data to 1.0, and reads its corresponding score value , set the confidence decay threshold table according to the score level. When the corresponding confidence weight is 0.9, set it to 0.8 when it is 0.7 when it is 0.6 when. The score of device D1 is 0.4255, which is between 0.4 and 0.5, so its confidence is adjusted to 0.7. The system traverses all devices and re - allocates the weighted weights in data fusion, performs weighted averaging on the device data of each type of data (such as heart rate, steps) according to the adjusted weights, and finally synthesizes a single data value stream at the same time point to obtain a fused data set.

[0032] Please refer to Figure 5 , the steps for obtaining the health data feature set are specifically as follows: S411: Call the fused data set, extract the walking volume parameter, sedentary duration parameter, sleep duration parameter, and heart rate fluctuation parameter within multiple time periods, and generate a periodic health parameter set; Call the fused data set, the system traverses all data points in the fused data, divides the data paragraphs by day or hour as a cycle, and defines each paragraph as a complete cycle , extract four types of health data dimensions within each cycle segment, including walking volume, sedentary duration, sleep duration, and heart rate fluctuation. Each dimension extracts the corresponding field from the original record. For example, the walking volume is statistically summed according to the cumulative step number field of the device, the sedentary duration is based on the posture sensor or action recognition module to mark the "stationary state" and sum the time, the sleep duration is judged and the effective sleep time is recorded according to the low movement at night + the lower limit of heart rate fluctuation, and the heart rate fluctuation extracts the difference between the maximum and minimum heart rates within this cycle segment. The system sequentially records the periodic data structures of each type of parameter, forming periodic parameter values such as (walking volume), (sleep duration), etc., and establishes a periodic health parameter set for multiple health parameters.

[0033] S412: Based on the set of periodic health parameters, extract the numerical changes and periodic time points of each type of health parameter in consecutive periods, construct a periodic number sequence and the total number of periods, and use the formula: ; Calculate the average change rate of multiple health parameters in the periodic sequence to obtain the periodic rate characteristic sequence value; Among them, represents the average change rate characteristic value of the th type of health parameter, represents at the th period, the value of the th type of health parameter, represents the value of this parameter in the next adjacent period, represents the rd period time point or sequence number, represents the time point or sequence number of the adjacent period, represents the total number of periods, represents the periodic number index, represents the health parameter type number; Based on the set of periodic health parameters, the system constructs the sequence of each type of health parameter in all periods , and at the same time defines the sequence time point for all periodic numbers. Taking the walking amount as an example, the three-period values are , and the periodic number is , then perform the following calculations: ; Substitute the walking amount parameter to get: ; The corresponding sleep duration data is , and the calculation gives: ; The heart rate fluctuation is , and the calculation gives: ; The eigenvalue of the average change rate of the final result is a change rate index composed of the numerical changes of various health parameters between adjacent periods in consecutive periods and the time difference. It is a numerical feature representing the activity of each type of health behavior output after average processing at the cycle level. The higher the value, the more frequent and rapid the fluctuations of the health behavior in the entire cycle sequence, reflecting higher behavioral instability or dynamics; a lower value means that the behavior is relatively stable and has strong stability. The parameter is an important quantitative basis for further identifying the trend of individual health status, constructing a portrait of behavior stability, and screening high-frequency abnormal behaviors. The formula is calculated continuously between periods, systematically revealing the change frequency and speed of behavior patterns such as walking volume, sedentary duration, sleep duration, and heart rate fluctuations. The absolute value is used to eliminate the direction interference, making it more suitable for measuring the stability trend, monitoring abnormal activities, or constructing a behavior fluctuation threshold. By averaging all periods, the full-cycle change activity characteristics of each type of health behavior are depicted, facilitating subsequent feature comparison, trend extraction, or behavior clustering.

[0034] Table 4 Table of eigenvalue of cycle rate of health parameters As shown in Table 4, different health parameters have different fluctuation amplitudes within the cycle, and the eigenvalue of the rate feature intuitively reflects their change intensity and stability.

[0035] S413: Call the sequence value of the cycle rate feature, arrange the rate feature values of each health parameter in cycle order, extract its time trend change pattern and judge the trend interval to obtain the health data feature set; Call the sequence value of the cycle rate feature, and the system will values are arranged in sequence one by one. A rate feature trend trajectory diagram is constructed according to the cycle sequence number. The change values at each time point are connected to form a trend line. The change direction of the trend line is judged. If the rate values of three consecutive cycles form a monotonically increasing sequence, the trend is marked as "rising interval"; if it is decreasing, it is "falling interval"; if the fluctuation amplitude is below 0.1, it is recorded as "stable interval". The upper and lower fluctuation interval boundary values are set respectively. The system divides the trend state section by section in the trend sequence of each parameter. For example, the walking volume trend line is 300 - 300, which belongs to stable; the sleep duration trend is from 0.2 rising to 0.5 and then falling back to 0.35, marked as oscillating; the heart rate fluctuation trend is from 1.7 falling to 0.4, belonging to an obvious downward trend. Finally, the trend labels and original values of each parameter are integrated to form the health data feature set.

[0036] Please refer to Figure 6 , the specific steps for obtaining the health data processing result are as follows: S511: calling the health data feature set, extracting the periodic change rate feature values ​​and time series fluctuation feature values ​​of multiple health parameters, and comparing them with the preset chronic disease risk factor feature thresholds, identifying and marking the activation state of each risk factor, and obtaining the activation state comparison results; Call the health data feature set. The system presets multiple chronic disease risk factors. Each risk factor corresponds to a multi-parameter associated feature set. The system extracts the change rate characteristics of each type of parameter in multiple cycles from the health data feature set. and volatility , and compare them item by item with the characteristic conditions in the risk factor definition to determine the activation status. For example, for the risk factor "metabolic decline risk", its definition characteristics are: , , , when the system detects that the corresponding parameters of the current user are , , If any two of the three indicator activation conditions are met, the system determines that the risk factor "metabolic decline risk" is activated. The system performs row-by-row judgment on multiple risk factors in turn according to the risk factor-parameter matching matrix, and finally generates a comparison result of the risk factor activation status.

[0037] S512: According to the activation status comparison result, by identifying the activation status of the risk factor of each health parameter, extracting the sensitivity parameter of each activated risk factor, evaluating the activation intensity of each parameter, and generating activation intensity data; Based on the activation status comparison results, the system traverses all risk factors marked as "activated" and extracts and defines sensitivity coefficients for their associated parameters. ,in represents the risk factor number, Indicates the factor on which the For example, the parameters and sensitivities corresponding to "metabolic decline risk" are: walking volume 0.6, sitting time 0.8, heart rate fluctuation 0.9. The system calculates the percentage of each parameter that actually exceeds the threshold and multiplies it by its sensitivity coefficient to form a single parameter activation intensity. , such as walking volume: , meditation duration: , heart rate fluctuations: .

[0038] Table 5 Composition of activation intensity of chronic disease risk factors As shown in Table 5, the system merges the activation strength of all relevant parameters of the current risk factor and generates the risk factor activation strength value using a weighted sum or average strategy. , the final activation intensity of the metabolic decline risk factor is , the other risk factors are calculated in turn, and finally a complete activation intensity matrix is constructed to form an activation intensity data set.

[0039] S513: Call the activation intensity data, obtain the chronic disease risk score of the user by calculating the risk score of each activation factor, and output the processing result of the health data; Call the activation intensity data, and the system integrates the activation intensities corresponding to each risk factor , generate the chronic disease risk score of the user. The score calculation method is to normalize the sum of the intensities of all activated risk factors to the upper limit of the total score of 1.0. In the current example, the activation intensity of "metabolic decline risk" is 0.38, and the activation intensity of "sleep quality disorder risk" is 0.28, then the total score is , the system segments the score according to the definition of the score level. The score <0.3 is "low risk", 0.3 - 0.6 is "medium risk", >0.6 is "high risk". The current score belongs to the "high risk" level. The system combines this level with information such as the activation factor name and the corresponding intensity to form a health risk score structure body, and outputs the processing result of the health data, including: risk score value, grading label, activation factor list, activation parameters of each factor, sensitivity and score weight mapping table, full life cycle label mapping and other dimensional field information to form the final chronic disease risk assessment result.

[0040] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. Cross-device health data fusion method, characterized in that It includes the following steps: S1: Monitor the device connection status parameters. By detecting the device reconnection request, call the data capture progress parameter and cache pointer position parameter in the device data cache, calculate the data missing interval amount in combination with the current data serial number, and send a data reissuance request to generate a breakpoint resumption compensation value; S2: Call the breakpoint resumption compensation value, obtain the acquisition data sequences of multiple devices. By comparing the local timestamp and reference timestamp of each device, calculate the time offset and perform calibration, identify the missing values in the data sequences and perform data interpolation to generate an interpolation compensation result; S3: Call the interpolation compensation result, extract the signal interruption times, signal abnormal fluctuation times parameter, sampling interval offset parameter of each device, calculate the stability score of the device, adjust the data confidence weight corresponding to each device and fuse the same type of data to obtain a fused data set; S4: Call the fused data set, extract the walking amount parameter, sedentary duration parameter, sleep duration parameter and heart rate fluctuation parameter, calculate the change rate characteristics of each parameter in multiple cycles, and analyze the time series change trend of the rate characteristics to generate a health data feature set.

2. The cross-device health data fusion method according to claim 1, wherein The breakpoint resumption compensation value is specifically the data missing interval amount, reissuance request instruction, data capture progress positioning information. The interpolation compensation result includes the time offset calibration value, data sequence missing segment identifier, interpolated generated data points. The fused data set is specifically the weighted fusion parameter value, device confidence mapping table, unified time series data structure. The health data feature set includes the cycle change rate matrix, trend change direction feature, continuous fluctuation range value.

3. The cross-device health data fusion method according to claim 1, wherein The specific steps for obtaining the breakpoint resumption compensation value are as follows: S111: Monitor the device connection status parameters, real-time identify the connection status of the device, including connection and disconnection, real-time detect the device reconnection request, obtain the reconnection request identifier and the current device connection status code, generate a device reconnection status identifier value; S112: Call the device reconnection status identifier value, extract the data capture progress parameter and cache pointer position parameter in the device data cache, calculate the offset between the data capture progress parameter and the cache pointer position parameter, generate a cache offset value; S113: Based on the cache offset value, call the current data serial number and the last valid data serial number, calculate the difference between the current data serial number and the last valid data serial number, obtain the data missing interval amount, and send a data reissuance request to generate a breakpoint resumption compensation value.

4. The cross-device health data fusion method according to claim 3, wherein The specific steps for obtaining the interpolation compensation result are as follows: S211: Call the breakpoint resumption compensation value, obtain the acquisition data sequences of multiple devices, extract the device number and timestamp corresponding to each device, generate a device acquisition sequence set value; S212: Based on the device acquisition sequence set value, calculate the difference between the local timestamp and the reference timestamp of each device, obtain the time offset and correct the acquisition data sequence of each device to generate a time calibration sequence value; S213: Call the time calibration sequence value, identify the missing values in the continuous data timeline, extract the adjacent data points of the missing values, calculate the values of the interpolation points, perform data interpolation on the collected data sequence, and generate an interpolation compensation result.

5. The cross-device health data fusion method according to claim 4, wherein The specific steps for obtaining the fusion data set are as follows: S311: Call the interpolation compensation result, extract the signal interruption count parameter, signal abnormal fluctuation count parameter, and sampling interval offset parameter of each device, and generate a device stability basic parameter set; S312: According to the device stability basic parameter set, extract the signal interruption count and abnormal fluctuation count of each device in each data segment, obtain the sampling interval offset and signal duration parameter, and call the device data segment quantity parameter to calculate the device stability score value; S313: Call the device stability score value, adjust the corresponding data confidence weight according to the score value of each device, perform data fusion on the same type of data, and obtain the fusion data set.

6. The cross-device health data fusion method according to claim 5, wherein, The specific steps for obtaining the health data feature set are as follows: S411: Call the fusion data set, extract the walking quantity parameter, sedentary duration parameter, sleep duration parameter, and heart rate fluctuation parameter within multiple time periods, and generate a periodic health parameter set; S412: Based on the periodic health parameter set, extract the numerical change situation and periodic time points of each type of health parameter in consecutive periods, construct a period number sequence and the total number of periods, and calculate the average change rate of multiple health parameters in the period sequence to obtain the period rate feature sequence value; S413: Call the period rate feature sequence value, arrange the rate feature values of each health parameter in chronological order, extract its time trend change pattern and judge the trend range, and obtain the health data feature set.

7. The cross-device health data fusion method according to claim 1, wherein, The method further includes: S5: Call the health data feature set, obtain the periodic change rate feature values and time series fluctuation feature values of multiple parameters, compare with the preset chronic disease risk factor feature thresholds, identify the activation status of the risk factors, and combine the sensitivity parameters of multiple risk factors to calculate the chronic disease risk score of the user and obtain the health data processing result; The health data processing result specifically refers to the chronic disease risk score value, the type of activated risk factors, and the combination of sensitivity impact parameters.

8. The cross-device health data fusion method according to claim 7, wherein The specific steps for obtaining the health data processing result are as follows: S511: Call the health data feature set, extract the periodic change rate feature values and time series fluctuation feature values of multiple health parameters, and compare with the preset chronic disease risk factor feature thresholds to identify and mark the activation status of each risk factor, and obtain the activation status comparison result; S512: According to the activation status comparison result, by identifying the activation status of the risk factors of each health parameter, extract the sensitivity parameters of each activated risk factor, evaluate the activation intensity of each parameter, and generate the activation intensity data; S513: Call the activation intensity data, obtain the chronic disease risk score of the user by calculating the risk score of each activated factor, and output the health data processing result.

Images