Intelligent ring health monitoring method based on multi-sensor cooperation and related equipment

Abstract

The application relates to the physiological parameter monitoring technical field of intelligent wearable devices, in particular to an intelligent ring health monitoring method based on multi-sensor cooperation and related equipment. The method comprises the following steps: acquiring motion, optical and temperature sensing data, combining multi-dimensional rules and a user preset log to determine a scene, performing differentiated data acquisition, and generating a scene-matched health evaluation result after signal noise reduction, feature extraction and fusion analysis. The application can balance monitoring accuracy, low energy consumption and personalization, and improve the effectiveness of health monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of physiological parameter monitoring of intelligent wearable devices, and in particular to an intelligent ring health monitoring method based on multi-sensor cooperation and related equipment. BACKGROUND

[0002] With the improvement of the awareness of national health management and the rapid iteration of wearable technology, intelligent wearable devices have become the core carrier of personal health data monitoring. Among them, intelligent rings, with the advantages of small size, high wearing comfort, and no perception in daily use, are increasingly widely used in heart rate monitoring, body temperature tracking, motion state recording, and sleep quality analysis, providing important support for users to achieve convenient and normalized health management.

[0003] Although the current mainstream intelligent ring generally integrates motion sensing modules, optical sensing modules, and temperature sensing modules to cover the basic health parameter collection needs in the form of multi-sensor combination, the cooperative application of these sensors in existing technologies only stays at the level of independent operation of each sensor according to fixed parameters, and the collected data is directly used for single parameter calculation without adaptive adjustment according to the actual use scenario of the user.

[0004] This multi-sensor application mode lacking scene adaptability leads to core bottlenecks in the actual use of intelligent rings. On the one hand, in dynamic scenarios, the optical sensing module is easily disturbed by limb movement, while the fixed sampling frequency of the motion sensing module is difficult to capture the details of the action to assist in correcting the data, resulting in significant errors in physiological parameter monitoring such as heart rate. On the other hand, in static or sleep scenarios, the sensors still maintain a high sampling frequency, which not only consumes excessive power and shortens the device's battery life, but also may interfere with the effective information due to excessive collection of irrelevant data. At the same time, the multi-source data in different scenarios lacks relevant analysis logic matching the scene characteristics, and cannot form health evaluation basis that meets the scene needs. In summary, the multi-sensor cooperation mechanism of existing intelligent rings is not deeply combined with scene characteristics, and it is difficult to balance the accuracy of health monitoring data, effective control of device energy consumption, and actual effectiveness of monitoring data in complex use scenarios, ultimately leading to the health evaluation results output by the ring being unable to fully meet the user's demand for accurate and long-term health management. SUMMARY

[0005] The present application aims to provide an intelligent ring health monitoring method based on multi-sensor cooperation and related equipment, to solve the problem that the multi-sensor cooperation mechanism of existing intelligent rings is disconnected from the characteristics of static, dynamic, and sleep scenarios, and only runs the sensors independently according to fixed parameters, making it difficult to balance the accuracy of health monitoring data, device energy consumption control, and monitoring data effectiveness in complex scenarios, ultimately leading to health evaluation results that cannot meet the user's demand for accurate and long-term health management.

[0006] The application aims to provide an intelligent ring health monitoring method based on multi-sensor cooperation, comprising:

[0007] Obtaining sensing data, the sensing data including acceleration signals output by a motion sensing module, initial heart rate variability data collected by an optical sensing module, and basic body surface temperature detected by a temperature sensing module;

[0008] Based on the sensing data, a multi-dimensional scene determination rule is analyzed in combination with a preset activity log of a user to correct the determination result and determine a current monitoring scene type;

[0009] According to the determined scene type, each sensing module is controlled to execute a differentiated data acquisition strategy: a low-power sampling mode is adopted in a static scene; the sampling frequency of the optical sensing module and the motion sensing module is increased and the data of the two modules is kept synchronized in a dynamic scene; the sampling interval of the temperature sensing module and the body movement detection sensitivity of the motion sensing module are optimized in a sleep scene; and the corresponding module is controlled to enter a sleep state to reduce energy consumption in a non-acquisition period, so as to obtain multi-source monitoring data;

[0010] The multi-source monitoring data obtained through the differentiated acquisition strategy is sequentially subjected to signal-level noise reduction processing, feature-level parameter extraction, and decision-level fusion analysis steps to realize deep integration and accurate analysis of multi-dimensional data, so as to obtain data fusion analysis results;

[0011] Based on the data fusion analysis results, a health assessment result matched with the current scene type is generated: real-time physiological parameters and motion adaptation suggestions are output in a dynamic scene; sleep cycle analysis results are output in a sleep scene; and basic physiological parameter trend feedback is output in a static scene.

[0012] By adopting the technical scheme, the acceleration signal of the motion sensing module, the initial heart rate variability data of the optical sensing module and the basic body surface temperature of the temperature sensing module are synchronously acquired, thereby providing multi-dimensional and comprehensive original data support for subsequent scene judgment and data analysis; in combination with the preset multi-dimensional scene judgment rule and reference to the user preset activity log to correct the judgment result, the accuracy of the static, dynamic and sleep monitoring scene type identification is effectively improved, and the scene misjudgment problem caused by single feature judgment is avoided; then, a differentiated data acquisition strategy is executed according to different scene types, the sampling frequency of the optical and motion sensing modules is increased in the dynamic scene and kept synchronous to ensure the data timeliness and relevance, a low-power sampling mode is used in the static scene, the temperature sensing sampling interval and the motion sensing body motion detection sensitivity are optimized in the sleep scene, and the corresponding modules are put into sleep in the non-acquisition period, thereby solving the problem of insufficient data acquisition in the dynamic scene, greatly reducing the device energy consumption in the static and sleep scenes, prolonging the intelligent ring endurance time; meanwhile, the multi-source monitoring data are sequentially subjected to signal level noise reduction processing, feature level parameter extraction and decision level fusion analysis, the motion interference in the dynamic scene can be effectively filtered out, the time sequence alignment data of different sensors can be integrated, the deep correlation and accurate analysis of multi-dimensional data can be realized, and the analysis deviation caused by data isolation or distortion can be avoided; finally, the health evaluation result matched with the scene type is generated, the real-time physiological parameters and motion adaptation suggestions are output in the dynamic scene, the sleep cycle analysis result is output in the sleep scene, and the basic physiological parameter trend feedback is output in the static scene, thereby accurately meeting the health management needs of the user in different scenes, and the intelligent ring health monitoring is accurately, lowly and personalized, and the effectiveness of the health monitoring data and the user health management experience are significantly improved.

[0013] In a possible implementation of the present application, for the multi-feature conflict in scene judgment, the step of determining the current monitoring scene type comprises:

[0014] When different dimensional features in the sensing data point to different scene types, the conflict features are extracted;

[0015] The historical scene conversion data are called, and a feature weight model is constructed based on the scene contribution degree of the conflict features in the historical data;

[0016] According to the feature weight model, a dynamic weight associated with the current feature matching degree is given to the conflict features, a comprehensive scene score is obtained through weighted calculation, and the type with the highest comprehensive scene score is taken as the final scene type.

[0017] By adopting the technical scheme, for the multi-feature conflict in scene judgment, the conflict features pointing to different scene types are accurately extracted from the sensing data, the core contradiction point causing the ambiguity of scene judgment is effectively located, and the deviation caused by directly judging the scene due to ignoring the multi-dimensional feature difference is avoided; then, the historical scene conversion data is called, a feature weight model is constructed based on the scene contribution degree of the conflict features in the historical data, the weight setting no longer depends on subjective experience but relies on the actual scene conversion law, the objectivity and rationality of the weight distribution are ensured; finally, the conflict features are given dynamic weights associated with the current feature matching degree according to the feature weight model, and the final scene type is determined by calculating the comprehensive scene score through weighting, so that the quantitative analysis of the influence degree of the conflict features on the scene judgment is realized - for example, a certain conflict feature has a higher contribution degree in the historical dynamic scene conversion, so when the current feature matches the dynamic scene, a higher weight is given, and vice versa, which not only flexibly adapts the actual influence of different features in the current scene, but also quantifies the scene matching degree through the comprehensive score, completely solves the problem that the traditional fixed threshold or single feature judgment cannot cope with multi-feature conflict, significantly improves the accuracy and stability of the scene type judgment of static, dynamic, sleep and other scenes, and further provides a reliable scene basis for the accurate execution of the subsequent differentiated data collection strategy, avoids the problems of improper sampling frequency, energy waste or insufficient data effectiveness caused by scene misjudgment, and ensures the accuracy and efficiency of the overall health monitoring process.

[0018] In a possible implementation of the present application, in view of the optical signal distortion caused by motion interference in a dynamic scene, the steps of the signal level noise reduction processing include:

[0019] Extracting the acceleration signal time domain feature of the motion sensing module;

[0020] Establishing a mapping relationship between the acceleration signal time domain feature and the optical signal distortion mode; wherein the optical signal distortion mode is the baseline drift, peak attenuation and phase shift of the optical sensing module caused by motion interference;

[0021] When periodic motion interference is detected, based on the corresponding rule of action period and distortion period in the mapping relationship, the optical signal is periodically compensated through interference mode inverse operation to offset the baseline drift and phase shift synchronized with the action period;

[0022] When non-periodic motion interference is detected, based on the correlation level of amplitude peak value and distortion intensity in the mapping relationship, the distortion type and degree of the signal loss segment are identified, a temporary interpolation algorithm matched with the distortion degree is enabled to fill the signal loss segment, and the continuity of the heart rate variability data is ensured.

[0023] By adopting the technical scheme, the acceleration signal time domain features of the motion sensing module are extracted first, the motion law (such as period, amplitude, etc.) of the limb motion in a dynamic scene is accurately captured, a quantitative basis is provided for subsequent targeted compensation of optical signal distortion, and the blindness of traditional general filtering algorithms to motion interference "non-discriminative processing" is avoided; then the mapping relationship between the acceleration signal time domain features and the optical signal distortion modes (baseline drift, peak attenuation, phase shift) is established, the signal distortion type and degree corresponding to different motion features are determined, the interference processing is changed from "passive filtering" to "active adaptation", and the problem of incomplete noise reduction caused by weak correlation between motion interference and optical signal distortion is solved; for subsequent periodic motion interference, based on the corresponding law of motion period and distortion period in the mapping relationship, periodic compensation of the optical signal is realized through interference mode inverse operation, the baseline drift and phase shift caused by regular actions such as running and fast walking are accurately compensated, and the influence of periodic interference on heart rate variability data is significantly reduced; for non-periodic motion interference, according to the correlation level of amplitude peak value and distortion intensity in the mapping relationship, the distortion type and degree of the signal loss segment can be quickly identified, and then the appropriate temporary interpolation algorithm is matched and adapted to fill the signal gap, effectively avoiding the heart rate data rupture caused by sudden actions (such as jumping and turning).

[0024] In a possible implementation of the present application, in view of the contradiction between low power consumption and data integrity, the differentiated acquisition strategy includes:

[0025] In the non-acquisition period, a low-power wake-up signal output by the motion sensing module is monitored in real time, the low-power wake-up signal is a signal detected by the module in a sleep state and meeting a preset health-related action feature;

[0026] When the low-power wake-up signal is detected, the corresponding optical sensing module or temperature sensing module is awakened based on the action feature of the low-power wake-up signal, short-time high-frequency acquisition matching the action duration is performed, and the action-related physiological parameter change is ensured to be captured;

[0027] After the acquisition is completed, the related module is controlled to enter the sleep state again according to the health importance corresponding to the low-power wake-up signal.

[0028] By adopting the technical scheme, only the motion sensing module is allowed to run in a low-power mode and monitor the wake-up signal during a non-collection period, rather than allowing all sensing modules to continuously work, thereby greatly reducing unnecessary energy consumption from the source, and avoiding the short battery life caused by long-term high-frequency operation of the modules in the traditional fixed sampling mode. Meanwhile, the wake-up signal is limited to "complying with the preset health-related action characteristics", which accurately filters out redundant actions without health value (such as slight finger shaking), thereby avoiding the waste of electrical energy caused by invalid wake-up, and ensuring that key actions (such as abnormal body movement during sleep and sudden standing in a static scene) that may be accompanied by physiological parameter changes can be captured in time, and data meaningful for health assessment is prevented from being missed. In addition, after detecting the wake-up signal, only the optical or temperature sensing module (rather than all modules) that matches the action characteristics is woken up, and "short-time high-frequency collection matching the duration of the action" is adopted - by "short-time", the energy consumption overload caused by long-term high-frequency operation of the module is avoided, and by "high-frequency", the physiological parameter changes (such as heart rate fluctuations during abnormal body movement) associated with the action are completely captured, thereby realizing "on-demand collection" rather than "blind collection". Finally, after the collection is completed, the module is put into sleep according to the health importance of the wake-up signal, the sleep interval can be appropriately shortened for key actions (such as warm-up before exercise) that affect health assessment to ensure data continuity, and the sleep interval is extended for non-key actions to further save energy, thereby forming a closed-loop energy consumption control logic of "low-power monitoring - accurate wake-up - on-demand collection - intelligent sleep". The core contradiction of "low power consumption and data loss, or high energy consumption to preserve data" in the traditional mode is solved, the intelligent ring battery life is significantly prolonged (such as from 1-2 days to 5-7 days), key health-related data is completely preserved, and the effectiveness of subsequent health assessment results is ensured, thereby effectively meeting the dual core needs of users for "long battery life + accurate monitoring" of the intelligent ring.

[0029] In a possible implementation of the present application, in order to solve the fusion deviation caused by the different time sequences of multi-source data, the feature level parameter extraction includes:

[0030] A high-precision timestamp is added to the collected data of each sensing module;

[0031] Through the clock synchronization mechanism, the time offset between the motion sensing module, the optical sensing module and the temperature sensing module is calculated based on the high-precision timestamp, and the collected data of each sensing module is time calibrated according to the time offset, so that the time reference of the multi-source feature data collected by the three types of modules is consistent;

[0032] For the temperature sensing data with collection delay, based on the time difference between the high-precision time stamp of the temperature sensing data and the high-precision time stamp of the optical sensing module / motion sensing module data, a linear interpolation method is used to generate a temperature sensing time point matched with the optical sensing module / motion sensing module data time node on the time axis, ensuring that the multi-source feature parameters are aligned at the same time node.

[0033] By using the above technical solutions, high-precision time stamps are added to the data collected by each sensing module, and each piece of motion, optical, and temperature sensing data is given a precise time identifier, solving the problem of "no reference for time base" of multi-source data from the source, avoiding the deviation caused by the lack of time basis in subsequent calibration; then relying on the clock synchronization mechanism, the time offset between the three types of sensing modules is calculated based on the high-precision time stamp and calibrated, effectively offsetting the natural time difference between different modules due to differences in hardware response speed and sampling frequency, making the time base of the motion acceleration signal, heart rate variability data, and body temperature data completely unified, and eliminating the misalignment of "optical data has been collected for 5 seconds, and temperature data is synchronized to the corresponding time". For the collection delay problem that may exist in temperature sensing data, the matching time point is generated by linear interpolation method through the difference calculation of the high-precision time stamp of the temperature sensing data and the time stamp of the optical / motion data, which not only fills the time gap caused by temperature data delay, but also ensures the continuity and relevance of the interpolated data and the actual collected data, avoiding the "irrelevant physiological parameter changes" caused by direct splicing of delayed data. The overall scheme solves the core problem of multi-source data time sequence asynchronization through the progressive processing of "precise time marking → inter-module time calibration → delayed data time sequence completion", ensuring that the motion, optical, and temperature feature parameters are completely aligned at the same time node, providing a multi-dimensional data basis with consistent time sequence and close correlation for subsequent decision-level fusion analysis, effectively avoiding fusion deviations such as "heart rate fluctuation does not match body temperature change" and "motion intensity and physiological parameter correlation distortion" caused by data misalignment, significantly improving the accuracy of data fusion analysis results, and thus ensuring the reliability of the health assessment report (such as dynamic scene motion physiological correlation analysis and sleep scene body temperature and body movement coordination judgment) generated based on the fusion results, meeting the user's demand for accurate multi-dimensional health data comprehensive analysis.

[0034] In a possible implementation of the present application, for the data invalidation caused by abnormal sensor wearing state, the method further comprises:

[0035] Obtaining the micro-displacement signal output by the motion sensing module and the signal intensity distribution feature collected by the optical sensing module;

[0036] Based on the normal fluctuation threshold of the micro-displacement signal and the uniformity threshold of the signal intensity distribution characteristics, a wearing status evaluation model is constructed. The model is used to quantitatively determine whether the current wearing is stable.

[0037] When the wearing status assessment model detects that the signal strength of the optical sensing module drops sharply by more than a threshold compared to the normal reference, and the small displacement signal of the motion sensing module exceeds the normal fluctuation threshold, it is determined to be an abnormal wearing condition.

[0038] Based on the signal strength distribution characteristics of the optical sensing module, the sampling area is automatically adjusted to a sub-region where the signal strength meets the normal benchmark, and the confidence level of the data during the period of abnormality is marked according to the duration of the abnormality. Data with low confidence or below is not included in the core health assessment.

[0039] If the abnormal wearing condition persists for more than a preset time, the smart ring will trigger a vibration reminder or push a wearing adjustment prompt to the bound terminal to avoid long-term invalid data collection.

[0040] By adopting the technical scheme, the micro displacement signal of the motion sensing module and the signal intensity distribution characteristics of the optical sensing module are synchronously acquired, abnormality judgment basis is constructed from two core dimensions of "physical wearing position" (displacement) and "data acquisition quality" (signal intensity), the problem that temporary shielding (such as temporary water contact when washing hands) is mistakenly judged as wearing abnormality only by relying on a single dimension (such as only looking at signal intensity sudden drop) is avoided, and the accuracy of abnormality identification is improved; a wearing state evaluation model is constructed based on the normal fluctuation threshold of the micro displacement signal and the uniformity threshold of the signal intensity distribution, abstract "wearing stability" is converted into quantifiable threshold judgment standard, the traditional subjective experience type judgment is replaced, the abnormality judgment is more objective and consistent, and the problem that the judgment standards are different in different scenes is avoided; the wearing abnormality is only judged when the model detects that "the optical signal intensity sudden drop exceeds the threshold and the micro displacement exceeds the normal range", interference is further filtered through "double condition superposition", misjudgment caused by accidental signal fluctuation or slight body movement is reduced, and the abnormality judgment result is reliable; after the abnormality is judged, the sampling area is automatically adjusted to the signal normal sub-area based on the optical signal intensity distribution characteristics, active adaptation and remediation of data acquisition are realized, data loss caused by directly discarding all data due to local wearing deviation is avoided, low credibility data is excluded from the core health evaluation according to the abnormality duration, abnormal data interference is effectively prevented, and the evaluation accuracy is guaranteed; if the abnormality lasts for more than a preset duration, vibration reminding or terminal pushing is triggered, the user is guided to adjust the wearing state in time, monitoring interruption and resource waste caused by long-term invalid data acquisition are avoided, and the user experience is improved. The overall scheme forms a closed-loop processing logic of "abnormality accurate identification → data active remediation → invalid data filtering → user timely reminding", completely solves the problem of data invalidation caused by wearing abnormality of the traditional intelligent ring, guarantees the effectiveness and continuity of health monitoring data, improves the reliability of health evaluation results, optimizes the user convenience through active adaptation and reminding, and makes the intelligent ring more practical in daily wearing scenes.

[0041] In a possible implementation of the present application, in view of the fact that individual physiological differences lead to insufficient applicability of evaluation results, the step of generating the health evaluation result comprises:

[0042] When used for the first time, a user-specific physiological baseline including the following parameters is established through multi-scene data acquisition for a continuous preset number of days: resting heart rate baseline, basic body temperature fluctuation range, heart rate response threshold in dynamic scenes;

[0043] When the health evaluation result is generated subsequently, the real-time physiological data is compared with the user-specific physiological baseline;

[0044] When it is detected that the real-time data exceeds the preset deviation range of the user-specific physiological baseline for a continuous preset number of days, it is determined that the user's physiological baseline has a long-term drift, the corresponding parameter in the user-specific physiological baseline is automatically updated, and the subsequent health assessment result is regenerated based on the updated user-specific physiological baseline.

[0045] By using the above technical scheme, the user-specific physiological baseline including the resting heart rate baseline, the basic body temperature fluctuation range, and the dynamic scene heart rate response threshold is constructed through the continuous preset number of days of multi-scene data collection at the first use, which completely gets rid of the limitation of the traditional health assessment relying on the general physiological standard, that is, it avoids misjudging the resting heart rate of a professional athlete below 50 times per minute as abnormal, and does not give a false alarm due to the basic body temperature of an old user being slightly higher than the ordinary standard, so that the physiological baseline completely fits the individual characteristics such as age, physical fitness, and exercise habits of the user; when the health assessment result is generated subsequently, the slight change (such as an increase of 3 times per minute in the resting heart rate compared with the baseline) of the physiological parameter of the user is accurately captured based on the specific baseline as the comparison basis, rather than compared with the general range, so that the assessment result is more targeted and has more reference value, and the problem of "normal range covering individual abnormalities" under the general standard is effectively avoided; when it is detected that the real-time data exceeds the deviation range of the specific baseline for a continuous preset number of days, the baseline parameter is automatically updated, the long-term change (such as a steady decrease in the resting heart rate after regular exercise, and an adjustment of the body temperature fluctuation range during the recovery period after surgery) of the physiological state of the user is dynamically adapted, the assessment baseline is always synchronized with the current physiological state of the user, and the subsequent assessment result is not out of touch due to the solidification of the baseline. The overall scheme forms a health assessment system completely adapted to individual characteristics through the logic of "specific baseline establishment, real-time data comparison, and baseline dynamic update", completely solves the problem of insufficient applicability of the assessment caused by individual physiological differences, and enables different groups of people (such as athletes, the elderly, and patients with chronic diseases) to obtain accurate and self-adapted health assessment results, effectively improves the guiding value of health monitoring for personal health management, and meets the core needs of users for personalized health services.

[0046] The second object of the present application is to provide an intelligent ring health monitoring system based on multi-sensor cooperation, which comprises:

[0047] The sensing data acquisition module acquires sensing data, which includes an acceleration signal output by the motion sensing module, initial heart rate variability data collected by the optical sensing module, and a basic body temperature detected by the temperature sensing module.

[0048] The current monitoring scene type determination module analyzes based on the sensing data in combination with a preset multi-dimensional scene determination rule, and simultaneously corrects the determination result by referring to a user preset activity log to determine the current monitoring scene type.

[0049] The multi-source monitoring data acquisition module: according to the determined scene type, control each sensing module to execute the differentiated data collection strategy: the static scene adopts the low-power sampling mode; the dynamic scene improves the sampling frequency of the optical sensing module and the motion sensing module and keeps the data synchronization of the two; the sleep scene optimizes the sampling interval of the temperature sensing module and the body motion detection sensitivity of the motion sensing module; and in the non-collection period, control the corresponding module to enter the sleep state to reduce the energy consumption, so as to obtain the multi-source monitoring data;

[0050] The data fusion analysis result output module: for the multi-source monitoring data obtained through the differentiated collection strategy, sequentially execute the signal level noise reduction processing, the feature level parameter extraction and the decision level fusion analysis steps, to realize the deep integration and accurate analysis of multi-dimensional data, and obtain the data fusion analysis result;

[0051] The health assessment result generation module: based on the data fusion analysis result, generate the health assessment result matched with the current scene type: the dynamic scene outputs the real-time physiological parameter and the motion adaptation suggestion; the sleep scene outputs the sleep cycle analysis result; the static scene outputs the basic physiological parameter trend feedback.

[0052] By adopting the technical scheme, the acceleration signal of the motion sensing module, the initial heart rate variability data of the optical sensing module and the basic body surface temperature of the temperature sensing module are synchronously acquired, thereby providing multi-dimensional and comprehensive original data support for subsequent scene judgment and data analysis; in combination with the preset multi-dimensional scene judgment rule and reference to the user preset activity log to correct the judgment result, the accuracy of the static, dynamic and sleep monitoring scene type identification is effectively improved, and the scene misjudgment problem caused by single feature judgment is avoided; then, a differentiated data acquisition strategy is executed according to different scene types, the sampling frequency of the optical and motion sensing modules is increased in the dynamic scene and kept synchronous to guarantee the data timeliness and relevance, a low-power sampling mode is used in the static scene, the temperature sensing sampling interval and the motion sensing body motion detection sensitivity are optimized in the sleep scene, and the corresponding modules are put into sleep in the non-acquisition period, thereby solving the problem of insufficient data acquisition in the dynamic scene, greatly reducing the device energy consumption in the static and sleep scenes, prolonging the intelligent ring endurance time, simultaneously, the multi-source monitoring data are sequentially subjected to signal level noise reduction processing, feature level parameter extraction and decision level fusion analysis, the motion interference in the dynamic scene can be effectively filtered out, the time sequence alignment data of different sensors can be integrated, the multi-dimensional data deep correlation and accurate analysis can be realized, and the analysis deviation caused by data isolation or distortion is avoided; finally, the health evaluation result matched with the scene type is generated, the real-time physiological parameters and motion adaptation suggestions are output in the dynamic scene, the sleep cycle analysis result is output in the sleep scene, and the basic physiological parameter trend feedback is output in the static scene, thereby accurately meeting the health management needs of the user in different scenes, the intelligent ring health monitoring is accurately, lowly and personalized as a whole, and the effectiveness of the health monitoring data and the user health management experience are significantly improved.

[0053] The third object of the present application is to provide an intelligent ring health monitoring device based on multi-sensor cooperation, which comprises:

[0054] The memory and the processor, wherein the memory stores a computer program capable of being loaded and executed by the processor to execute the above-mentioned intelligent ring health monitoring method based on multi-sensor cooperation.

[0055] The fourth object of the present application is to provide a storage medium.

[0056] The fourth object of the present application is achieved by the following technical scheme:

[0057] A storage medium, wherein the storage medium stores a computer program capable of being loaded and executed by the processor to execute the above-mentioned intelligent ring health monitoring method based on multi-sensor cooperation.

[0058] In summary, the present application includes at least one of the following beneficial technical effects:

[0059] 1. Synchronously acquiring acceleration signals of a motion sensing module, initial heart rate variability data of an optical sensing module, and basic body surface temperature of a temperature sensing module, providing multi-dimensional and comprehensive raw data support for subsequent scene determination and data analysis; combined with pre-set multi-dimensional scene determination rules and reference to user pre-set activity logs to correct the determination results, effectively improving the accuracy of static, dynamic, sleep and other monitoring scene type identification, avoiding the scene misjudgment problem caused by single feature determination; then, according to different scene types, different data acquisition strategies are executed, in the dynamic scene, the sampling frequency of optical and motion sensing modules is improved and kept synchronous to ensure data timeliness and relevance, in the static scene, low-power sampling mode is used, in the sleep scene, temperature sensing sampling interval and motion sensing body motion detection sensitivity are optimized, and during non-acquisition period, the corresponding module is put into sleep, which not only solves the problem of insufficient data acquisition in dynamic scene, but also greatly reduces the energy consumption of static and sleep scenes, prolongs the endurance time of the intelligent ring; at the same time, the multi-source monitoring data is sequentially executed signal level noise reduction processing, feature level parameter extraction and decision level fusion analysis, which can effectively filter out motion interference in dynamic scene, integrate time sequence alignment data of different sensors, realize deep correlation and accurate analysis of multi-dimensional data, and avoid analysis deviation caused by data isolation or distortion; finally, the health evaluation results matched with the scene type are generated, the real-time physiological parameters and motion adaptation suggestions are output in the dynamic scene, the sleep cycle analysis results are output in the sleep scene, and the basic physiological parameter trend feedback is output in the static scene, which accurately meets the health management needs of users in different scenes, and overall realizes the precision, low consumption and personalization of the intelligent ring health monitoring, significantly improves the effectiveness of health monitoring data and user health management experience.

[0060] 2. For the scene judgment of multi-feature conflict, the conflict features pointing to different scene types are accurately extracted from the sensing data, effectively locating the core contradiction point causing the ambiguity of scene judgment, and avoiding the deviation caused by directly judging the scene due to ignoring the multi-dimensional feature difference; then the historical scene conversion data is called, and the feature weight model is constructed based on the scene contribution degree of the conflict features in the historical data, so that the weight setting no longer depends on subjective experience but relies on the actual scene conversion rule, ensuring the objectivity and rationality of the weight distribution; finally, the conflict features are given dynamic weights associated with the current feature matching degree according to the feature weight model, and the comprehensive scene score is obtained through weighted calculation to determine the final scene type, realizing the quantitative analysis of the influence degree of conflict features on scene judgment - for example, a conflict feature has a higher contribution degree in the historical dynamic scene conversion, so when the current feature matches the dynamic scene, it is given a higher weight, otherwise the weight is reduced, which not only flexibly adapts the actual influence of different features in the current scene, but also quantifies the scene matching degree through comprehensive score, completely solving the problem that the traditional fixed threshold or single feature judgment cannot cope with multi-feature conflict, significantly improving the accuracy and stability of the scene type judgment of static, dynamic, sleep and other scenes, and then providing a reliable scene basis for the accurate execution of the subsequent differentiated data collection strategy, avoiding the problems of improper sampling frequency, energy waste or insufficient data effectiveness caused by scene misjudgment, and ensuring the accuracy and efficiency of the overall health monitoring process. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a flow diagram of an intelligent ring health monitoring method based on multi-sensor cooperation provided by the embodiments of the present application;

[0062] Figure 2 is a virtual structure diagram of an intelligent ring health monitoring system based on multi-sensor cooperation provided by the embodiments of the present application. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme of the embodiments of the present application will be described clearly and completely below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0064] In addition, the term "and / or" in this paper is only to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the existence of A alone, the existence of A and B together, and the existence of B alone. In addition, the character " / " in this paper, unless otherwise specified, generally represents an "or" relationship between the associated objects before and after it.

[0065] The embodiments of the present application are further described in detail below with reference to the accompanying drawings.

[0066] The embodiments of the present application provide a kind of intelligent ring health monitoring method based on multi-sensor cooperation, refer to Figure 1 The main process of the method is described as follows:

[0067] S1: obtain sensing data, the sensing data include acceleration signal output by motion sensing module, initial heart rate variability data collected by optical sensing module and basic body temperature detected by temperature sensing module;

[0068] Wherein, the core of this step is to provide multi-dimensional original data support for subsequent scene determination, data processing and health assessment, through three functionally complementary sensing modules, the motion state, cardiovascular physiological characteristics and basic body temperature information of user are captured respectively, to ensure the comprehensiveness and relevance of original data.In specific implementation, the motion sensing module built-in intelligent ring adopts three-axis acceleration sensor, the initial sampling frequency of this sensor is set to 20Hz, can collect the acceleration signal (range ±4g, resolution 16 bits) of X, Y, Z three axial direction in real time, and the original acceleration data is transmitted to the processing unit of ring through I2C bus;Optical sensing module adopts the sensor integrated with green light and red light source, emits green light with wavelength 520nm to irradiate user's finger skin at 50Hz sampling frequency, calculates initial heart rate variability data (including RR interval, heart rate value, etc.) by receiving reflected light intensity change, while filtering out ambient light interference (such as through built-in ambient light sensor to compensate light intensity change);Temperature sensing module adopts sampling accuracy of ±0.1℃, initial sampling interval is set to 1 minute, directly detects the basic body temperature at the contact of finger skin, and transmits temperature data to processing unit after converting temperature data into digital signal;The original data of three sensing modules are temporarily stored in the on-chip RAM (capacity 64KB) of processing unit, to ensure the real-time of data transmission (delay <10ms).

[0069] S2: based on the sensing data, analysis is carried out in combination with preset multi-dimensional scene determination rules, and the determination result is corrected by synchronously referring to user's preset activity log, to determine the current monitoring scene type;

[0070] Wherein, by multi-dimensional data comparison and user log assistance, the current use scene (static, dynamic, sleep) of the user is accurately identified, which provides basis for subsequent execution of differentiated data collection strategy, and avoids scene misjudgment caused by single data dimension. In specific implementation, first, multi-dimensional scene judgment rules are pre-stored in the processing unit: the static scene judgment rule is "10-second variance of acceleration signal <0.1 m / s2, SDNN (standard deviation of all sinus rhythm RR intervals) of heart rate variability data >50 ms, body surface temperature fluctuation amplitude <0.2℃ (within 10 minutes)"; the dynamic scene judgment rule is "10-second variance of acceleration signal ≥0.1 m / s2, heart rate value is increased by ≥10 times / minute compared with the user's daytime resting heart rate (preset initial value 60-80 times / minute)"; the sleep scene judgment rule is "10-second variance of acceleration signal <0.05 m / s2 and lasts for more than 10 minutes, body surface temperature is reduced by 0.3-0.5℃ compared with the daytime average value (preset initial value 36.5℃)"; the processing unit compares the real-time sensing data acquired by S1 with the above rules every 5 seconds, and preliminarily outputs the scene judgment result; at the same time, the intelligent ring is bound with the user's mobile phone APP through Bluetooth, and the user can preset activity logs (such as "19:00-20:00 running" and "23:00-6:00 sleep") in the APP, and the processing unit reads the log information in real time, if the preliminary judgment result deviates from the log (such as S1 data showing acceleration variance 0.08 m / s2 close to the static rule at 19:10, but the log is "running"), the judgment result is corrected to "dynamic scene" by referring to the log, and finally the clear scene type is output (such as "static scene-office", "dynamic scene-running", and "sleep scene-pre-sleep").

[0071] S3: According to the determined scene type, control each sensing module to execute differentiated data collection strategy: low-power sampling mode is adopted in static scene; the sampling frequency of the optical sensing module and the motion sensing module is improved and the data of the two modules is kept synchronized in dynamic scene; the sampling interval of the temperature sensing module and the body motion detection sensitivity of the motion sensing module are optimized in sleep scene; and the corresponding modules are controlled to enter sleep state to reduce energy consumption in non-collection period, so as to acquire multi-source monitoring data.

[0072] Among them, according to the scene characteristics, the sensor parameters are dynamically adjusted, which can guarantee the accuracy of data collection in each scene, and at the same time, the energy consumption of the device is reduced to the greatest extent, and the balance between "monitoring effect" and "endurance ability" is balanced. In specific implementation, if it is determined that it is a static scene (such as a user sitting still, working), the processing unit controls the sampling frequency of the motion sensing module to be reduced from 20Hz to 5Hz, the sampling frequency of the optical sensing module to be reduced from 50Hz to 10Hz, and the temperature sensing module to keep a 1-minute sampling interval. In the non-collection period (that is, in addition to 10 times of sampling (each lasting 1 second) in each hour, the remaining time), a sleep instruction is sent to the three sensing modules, and only the processing unit is kept in a low-power mode (current consumption <10μA); if it is determined that it is a dynamic scene (such as running, fast walking), the processing unit increases the sampling frequency of the motion sensing module to 50Hz, the sampling frequency of the optical sensing module to 100Hz, and synchronizes the sampling time of the two through the internal clock signal (frequency 32MHz) of the processing unit (synchronization sampling is triggered once every 10ms), to ensure the time correspondence of the acceleration signal and the heart rate variability data, and the sampling interval of the temperature sensing module is shortened to 30 seconds. In the non-collection period (such as motion pause for more than 5 seconds), only the temperature sensing module is controlled to sleep; if it is determined that it is a sleep scene (such as after the user falls asleep), the processing unit adjusts the sampling interval of the temperature sensing module from 1 minute to 2 minutes, increases the body motion detection sensitivity of the motion sensing module (that is, the body motion trigger threshold is reduced from 0.1m / s² to 0.03m / s² to avoid missing the turning-over motion), and the sampling frequency of the optical sensing module is reduced to 15Hz. In the non-collection period (such as when the user is in deep sleep and the acceleration signal has no fluctuation for more than 5 minutes), the optical sensing module and the motion sensing module are controlled to sleep, and only the temperature sensing module is kept sampling at intervals; the multi-source monitoring data (acceleration, heart rate variability, temperature) collected in each scene are stored in the Flash memory of the ring, and are classified and archived according to the time stamp (accurate to seconds).

[0073] S4: The multi-source monitoring data obtained by the differential collection strategy are sequentially subjected to signal-level noise reduction processing, feature-level parameter extraction, and decision-level fusion analysis steps to realize deep integration and accurate analysis of multi-dimensional data, and obtain data fusion analysis results;

[0074] Among them, by layered data processing, filtering out the interference noise in the original data, extracting valuable feature parameters, and establishing the correlation of multi-dimensional data, laying the foundation for subsequent generation of accurate health assessment results. In specific implementation, first, signal level noise reduction processing is performed: for the acceleration signal of the motion sensing module, a sliding average filtering algorithm (filter window size is set to 5) is used to filter out high-frequency noise caused by slight finger shaking; for the initial heart rate variability data of the optical sensing module, a Kalman filtering algorithm (state equation is set to heart rate change rate model) is used to eliminate baseline drift interference caused by skin displacement in dynamic scenes, ensuring that the heart rate data error is <2 times / minute; for the data of the temperature sensing module, a median filtering algorithm (selecting the median value of 5 samples) is used to filter out abnormal values caused by environmental temperature fluctuations (such as water temperature influence when washing hands). Secondly, feature level parameter extraction is performed: the peak value, mean value, and variance (reflecting the intensity of movement) within 10 seconds are extracted from the denoised acceleration signal, SDNN and RMSSD (root mean square of adjacent RR interval differences, reflecting autonomic nervous function) are extracted from the denoised heart rate variability data, and the mean value and maximum fluctuation amplitude (reflecting body temperature stability) within 5 minutes are extracted from the denoised temperature data. All feature parameters are calculated and stored in 1-minute time windows. Finally, decision level fusion analysis is performed: the processing unit uses a weighted fusion algorithm to assign feature parameter weights according to scene types (dynamic scene: motion feature weight 40%, optical feature weight 40%, temperature feature weight 20%; static scene: optical feature weight 45%, temperature feature weight 45%, motion feature weight 10%; sleep scene: motion feature weight 30%, temperature feature weight 50%, optical feature weight 20%), and obtains scene-based fusion indicators (such as "motion physiological load index" for dynamic scenes, "basic physiological stability" for static scenes, and "sleep quality correlation value" for sleep scenes) through weighted calculation, which are the data fusion analysis results.

[0075] S5: Based on the data fusion analysis result, generate a health assessment result matching the current scene type: dynamic scene outputs real-time physiological parameters and motion adaptation suggestions; sleep scene outputs sleep cycle analysis results; static scene outputs basic physiological parameter trend feedback.

[0076] In the formula, the abstract fusion analysis result is converted into health information that can be understood by the user and has practical value, and the evaluation result is ensured to be highly adapted to the health management needs of the current scene of the user. In specific implementation, if it is a dynamic scene (such as running), the processing unit generates a health evaluation result based on a “exercise physiological load index”: real-time output of physiological parameters (current heart rate 130 times / minute, step frequency 180 steps / minute, body surface temperature 36.8℃), and judgment of exercise intensity (such as “current aerobic exercise interval”) according to the fusion index, generation of exercise adaptation suggestions (“suggest maintaining the intensity, remaining suitable exercise duration 20 minutes, avoiding heart rate exceeding 150 times / minute”), and real-time transmission of the evaluation result to the mobile phone APP through Bluetooth for display in the form of a pop-up window + digital list; if it is a sleep scene, the processing unit analyzes the sleep cycle (such as “2 hours of light sleep, 1.5 hours of deep sleep, 1 hour of REM sleep, and 2 times of wake-up”) based on the “sleep quality correlation value” during the night after the user wakes up (such as detection of acceleration signal variance > 0.1 m / s2 for 5 minutes), generates a sleep cycle analysis result (“deep sleep duration accounts for less than 25%, suggest going to sleep 30 minutes earlier and reducing the use of electronic devices before sleep”), and pushes it to the user through the APP; if it is a static scene, the processing unit aggregates “basic physiological stability” by hour to generate a basic physiological parameter trend feedback (such as “today 10:00-11:00 resting heart rate 62 times / minute, 2 times / minute lower than the same period yesterday; body surface temperature mean 36.5℃, fluctuation amplitude 0.1℃, within normal range”), display the trend of the day in the form of a line chart in the APP, and generate a trend report every week (such as “this week, the average resting heart rate is 61 times / minute, 3 times / minute lower than last week, and the autonomic nervous function is improved”).

[0077] Specifically, in some possible embodiments, for the conflict of multiple features in scene determination, the step of determining the current monitoring scene type comprises:

[0078] When different dimension features in the sensing data point to different scene types, the conflict features are extracted;

[0079] The historical scene transition data is called, and a feature weight model is constructed based on the scene contribution degree of the conflict features in the historical data;

[0080] According to the feature weight model, the conflict features are given dynamic weights associated with the current feature matching degree, a comprehensive scene score is obtained through weighted calculation, and the type with the highest comprehensive scene score is taken as the final scene type.

[0081] When different dimensional features in the sensing data point to different scene types, the conflict features are extracted first. Specifically, the processing unit monitors the matching of the feature parameters in the motion, optical, and temperature dimensions with the preset scene rules in real time. When the matching results of the features in at least two dimensions are inconsistent (for example, the 10-second acceleration variance of the motion sensing module is 0.12 m / s² (consistent with the dynamic scene rule), the 10-minute body surface temperature fluctuation amplitude of the temperature sensing module is 0.1°C (consistent with the static scene rule), and the SDNN of the optical sensing module is 55 ms (consistent with the static scene rule)), it is determined that there is a feature conflict, and the conflict feature parameters are extracted: the “acceleration variance 0.12 m / s²” in the motion dimension, the “fluctuation amplitude 0.1°C” in the temperature dimension, and the “SDNN 55 ms” in the optical dimension, and the original values of the features and the corresponding preliminary matching scenes (the motion feature matches the dynamic scene, and the temperature and optical features match the static scene) are recorded.

[0082] Then, the historical scene transition data is called to construct a feature weight model. The processing unit reads the historical scene transition records (each record contains the conflict feature parameters at the transition time, the final determined scene, and the scene duration) in the past 30 days from the Flash memory, and calculates the contribution degrees of the conflict features to different scenes in the historical data: for the acceleration variance feature in the motion dimension, the proportion of the number of times that the feature is used as a key feature in the historical dynamic scene determination (for example, the number of times that the feature causes the scene determination to be dynamic accounts for 65% of the total number of dynamic scene determination times in the past 30 days) is calculated, which is recorded as the dynamic contribution degree 65%; the proportion of the number of times that the feature is used as an interference feature in the static scene determination (for example, 15%) is calculated, which is recorded as the static contribution degree 15%; similarly, the static contribution degree of the temperature fluctuation amplitude feature is 70%, the dynamic contribution degree is 10%, the static contribution degree of the optical SDNN feature is 60%, and the dynamic contribution degree is 5%. Based on these contribution degrees, a weighted linear model is used to construct a feature weight model, and the model expression is: feature weight=(target scene contribution degree-non-target scene contribution degree) / 100, for example, the weight of the motion feature to the dynamic scene=(65%-15%) / 100=0.5, and the weight to the static scene=(15%-65%) / 100=-0.5; the weight of the temperature feature to the static scene=(70%-10%) / 100=0.6, and the weight to the dynamic scene=(10%-70%) / 100=-0.6.

[0083] Finally, the comprehensive scene score is calculated according to the feature weight model. The processing unit first calculates the matching degree of the current conflict feature and the historical same type feature (cosine similarity algorithm is adopted, the value range is 0-1), for example, the matching degree of the current acceleration variance 0.12 m / s2 and the typical acceleration variance (0.1-0.3 m / s2) in the historical dynamic scene is 0.9, and the matching degree with the typical value (<0.1 m / s2) in the static scene is 0.2; the matching degree of the temperature fluctuation amplitude 0.1℃ and the typical value (<0.2℃) in the historical static scene is 0.8, and the matching degree with the typical value (>0.2℃) in the dynamic scene is 0.3. Then the matching degree is multiplied by the feature weight of the corresponding scene to obtain the scene weighted value of each feature: the weighted value of the motion feature to the dynamic scene = 0.9 x 0.5 = 0.45, and the weighted value to the static scene = 0.2 x (-0.5) = -0.1; the weighted value of the temperature feature to the static scene = 0.8 x 0.6 = 0.48, and the weighted value to the dynamic scene = 0.3 x (-0.6) = -0.18; the weighted value of the optical feature to the static scene = 0.8 x 0.55 (optical feature static weight) = 0.44, and the weighted value to the dynamic scene = 0.2 x (-0.55) = -0.11. The sum of the weighted values of all features in the same scene is obtained to obtain the comprehensive scene score: the total score of the dynamic scene = 0.45-0.18-0.11 = 0.16, the total score of the static scene = -0.1+0.48+0.44 = 0.82, and the final selected static scene with the highest score is selected as the final determination result.

[0084] Specifically, in some possible embodiments, in view of the fact that the motion interference in the dynamic scene causes distortion of the optical signal, the step of signal level noise reduction processing includes:

[0085] extracting the acceleration signal time domain feature of the motion sensing module;

[0086] establishing a mapping relationship between the acceleration signal time domain feature and the optical signal distortion mode; wherein the optical signal distortion mode is the baseline drift, peak attenuation and phase shift of the optical sensing module caused by the motion interference;

[0087] when periodic motion interference is detected, based on the corresponding rule of action period and distortion period in the mapping relationship, the optical signal is periodically compensated by interference mode inverse operation to offset the baseline drift and phase shift synchronized with the action period;

[0088] when non-periodic motion interference is detected, based on the correlation level of amplitude peak value and distortion intensity in the mapping relationship, the distortion type and degree of the signal loss segment are identified, a temporary interpolation algorithm matched with the distortion degree is enabled to fill the signal loss segment, and the continuity of the heart rate variability data is ensured.

[0089] Wherein, when extracting the acceleration signal time domain characteristics of the motion sensing module, the processing unit pre-processes the three-axis acceleration signal in a dynamic scene: using a 1-second sliding window (step size 0.5 seconds) to frame the original acceleration data (sampling frequency 50 Hz), calculate the time domain characteristics in each frame, including the action period (extracted by the autocorrelation function, such as the step frequency corresponding to the period of about 0.5-1 seconds when running), amplitude peak value (the maximum value of the three-axis acceleration vector sum, unit m / s²), peak interval (the time difference between two consecutive amplitude peaks), and signal variance (reflecting the degree of motion intensity), and these characteristic parameters are stored in the cache area (capacity 100 frames) in real time.

[0090] When establishing the mapping relationship between the acceleration signal time domain characteristics and the optical signal distortion mode, the synchronous data of typical dynamic scenes (such as running, walking, and cycling) is collected through preliminary experiments to construct a mapping table: for periodic characteristics, record the corresponding relationship of action period T and optical signal baseline drift period T1 (experiments show that T1≈T), phase shift φ (related to the direction of action, such as φ≈90° when swinging arms); for amplitude characteristics, divide the amplitude peak level (0-2 m / s² for light, 2-5 m / s² for moderate, and >5 m / s² for heavy), corresponding to the peak attenuation rate of the optical signal (light 10%-20%, moderate 20%-50%, and heavy >50%) and the baseline drift amplitude (light 0.1-0.3V, moderate 0.3-0.6V, and heavy >0.6V, V is the voltage value of the optical signal); the mapping relationship is stored in the Flash of the processing unit, supporting dynamic updating (optimizing the mapping table every 10 hours of new data).

[0091] When periodic motion interference is detected (determination condition: the coefficient of variation of the action period of the continuous 5 frames of acceleration signal <10%), the processing unit calls the distortion parameters corresponding to the action period T (T1=T, φ=90°) from the mapping table, and generates a compensation signal through inverse operation of the interference mode: for baseline drift, generate a sine compensation wave with the same amplitude as the original drift signal and opposite phase based on T1 (such as a drift signal of 0.3sin(2πt / T1), and a compensation signal of -0.3sin(2πt / T1)); for phase shift, the optical signal is advanced by φ / (2π)×T1 length (such as T1 / 4 when φ=90°), so that the peak position is aligned with the non-interference time; the compensated optical signal is low-pass filtered (cutoff frequency 5 Hz) to remove high-frequency noise, ensuring the period accuracy of the heart rate variability data (error <5 ms).

[0092] When detecting non-periodic motion interference (decision condition: acceleration amplitude peak > 5 m / s² and motion period coefficient of variation > 30%), the processing unit first identifies the distortion type and degree based on the amplitude peak level (e.g., peak 8 m / s² corresponds to severe distortion, with a 300 ms signal loss segment); for mild distortion (loss segment < 100 ms), linear interpolation method (based on 3 valid data points before and after the loss segment) is used for filling; for moderate distortion (100-200 ms), cubic spline interpolation (retaining signal curvature characteristics) is used; for severe distortion (> 200 ms), prediction interpolation based on historical data is enabled (calling the heart rate variability trend under the same motion intensity in the past 10 seconds to predict the lost segment data); the filled data is processed by smoothing (moving average window of 3 points) to ensure continuity, so that the heart rate value calculation error is controlled within 3 times / minute, ensuring the effectiveness of subsequent feature extraction.

[0093] Specifically, in some possible embodiments, to resolve the contradiction between low power consumption and data integrity, the differentiated acquisition strategy includes:

[0094] During the non-acquisition period, a low-power wake-up signal output by the motion sensing module is monitored in real time, the low-power wake-up signal being a signal detected by the module in a sleep state and conforming to a preset health-related motion feature;

[0095] When the low-power wake-up signal is detected, the corresponding optical sensing module or temperature sensing module is awakened based on the motion feature of the low-power wake-up signal, short-time high-frequency acquisition matching the motion duration is performed, and motion-related physiological parameter changes are ensured to be captured;

[0096] After the acquisition is completed, the related module is controlled to enter the sleep state again according to the health importance corresponding to the low-power wake-up signal.

[0097] Wherein, when monitoring the low-power wake-up signal during the non-acquisition period, the motion sensing module switches to a low-power working mode: the sampling frequency is reduced from the regular 20 Hz to 1 Hz, only the acceleration detection of the X axis and the Y axis (main direction of finger movement) is retained, and a health-related motion feature threshold (e.g., displacement with an amplitude > 0.3 m / s² and a duration > 500 ms, excluding random jitter < 0.1 m / s²) is preset by a hardware comparator; when a signal conforming to the feature (e.g., a getting-up motion of a user changing from a sitting posture to a standing posture, a turning-over motion in sleep) is detected, the module sends a wake-up signal (high-level pulse) to the processing unit through an interrupt pin, and records the time-domain feature (amplitude peak, duration, direction change) of the motion; in this mode, the current consumption of the motion module is controlled to be within 5 μA, ensuring the low-power characteristic.

[0098] When the low-power wake-up signal is detected, the processing unit matches the corresponding sensing module according to the action feature: if it is a turning-over action in sleep (action amplitude 0.3-0.8 m / s², duration 1-3 seconds), it is determined that the body temperature and heart rate changes need to be monitored synchronously, and at the same time, the optical sensing module (sampling frequency is increased to 30 Hz) and the temperature sensing module (single sampling interval is shortened to 1 second) are awakened; if it is a short hand-raising action in the daytime (amplitude 0.5-1.2 m / s², duration <1 second), it is determined that the heart rate change is mainly related, and only the optical sensing module (sampling frequency 50 Hz) is awakened; the collection duration is bound to the action duration, and the strategy of "action duration + 1 second" (such as 3-second turning-over action corresponding to 4-second collection duration) is adopted, so as to ensure that the physiological parameter changes 0.5 seconds before the action, during the action and 0.5 seconds after the action are completely captured, and data truncation is avoided.

[0099] After the collection is completed, the processing unit controls the sleep according to the health importance classification of the wake-up signal: for the action with high health importance (such as continuous turning-over >3 times / minute at night, which may be related to sleep quality), a short sleep interval is set (the optical module is automatically awakened once for supplementary collection after sleeping for 5 minutes); for the action with medium importance (such as routine getting up in the daytime, <10 times per day), the control module enters deep sleep immediately after the collection is completed (no automatic awakening, only responding to the next wake-up signal); for the action with low importance (such as occasional finger shaking, amplitude <0.5 m / s²), the sleep interval is prolonged (the optical module sleeps for 30 minutes, and the temperature module maintains the original interval); the sleep control is realized by sending a sleep instruction through an I2C bus, and the switching time from awakening to re-sleeping of the module is <100 ms, which maximally reduces invalid energy consumption, while ensuring the integrity of the key health-related data.

[0100] Specifically, in some possible embodiments, in order to solve the fusion deviation caused by the time sequence of multi-source data being different, the feature-level parameter extraction includes:

[0101] A high-precision timestamp is added to the collected data of each sensing module;

[0102] Through the clock synchronization mechanism, the time offset between the motion sensing module, the optical sensing module and the temperature sensing module is calculated based on the high-precision timestamp, and the collected data of each sensing module is time calibrated according to the time offset, so that the time reference of the multi-source feature data collected by the three types of modules is consistent;

[0103] For the temperature sensing data with collection delay, based on the time difference between the high-precision timestamp of the temperature sensing data and the high-precision timestamp of the optical sensing module / motion sensing module data, a linear interpolation method is used to generate temperature sensing time points matching the optical sensing module / motion sensing module data time nodes on the time axis, ensuring that the multi-source feature parameters are aligned at the same time node.

[0104] When adding high-precision timestamps to the collection data of each sensing module, the 32 kHz real-time clock (RTC) built-in in the processing unit is used as the unified time reference, with a time accuracy of ±1 ms / day (regularly synchronized with the mobile phone network time through Bluetooth). After completing a single data collection (e.g., the motion module samples 1 frame of acceleration data, the optical module calculates 1 heart rate value, and the temperature module completes 1 temperature detection), the motion sensing module, optical sensing module, and temperature sensing module immediately send a data ready signal to the processing unit through the SPI interface. The processing unit responds to the signal and records the current RTC time (format: "year-month-day hour:minute:second.millisecond") as a high-precision timestamp, which is packaged and stored with the original data of the corresponding module (e.g., motion data entry: "2024-09-04 10:00:00.123, X:0.2 m / s², Y:0.1 m / s², Z:9.8 m / s²"). This ensures that each piece of data has a unique and accurate time identifier.

[0105] When calculating the time offset and calibrating the data through the clock synchronization mechanism, the processing unit triggers a multi-module clock synchronization every 10 seconds. First, it records the internal timing values of each module at the same time (e.g., RTC time t0) (motion module t1, optical module t2, temperature module t3), and calculates the offset of each module relative to the master clock (Δt1=t1-t0, Δt2=t2-t0, Δt3=t3-t0). If the absolute value of the offset exceeds 5 ms (determined as time out of sync), the historical collection data of each module is time-calibrated. For motion module data, the timestamp is corrected to "original timestamp-Δt1", optical module data to "original timestamp-Δt2", and temperature module data to "original timestamp-Δt3". This ensures that the time reference of the three types of module data is unified to the RTC master clock, and the time deviation of the calibrated data of different modules is controlled within ±3 ms.

[0106] When interpolating temperature sensing data with collection delay, first, the inherent collection delay of the temperature module is counted (about 200 ms measured by experiment, i.e., the actual collection time of the temperature data is 200 ms later than the marked timestamp), and the time difference with the optical / motion data is calculated (for example, the temperature data timestamp is t, and the optical data timestamp is t, so the actual time difference is 200 ms). For the low sampling interval of the temperature module (for example, 1 minute per time for a static scene), the sampling points (for example, 100 ms per time) of the optical / motion data are taken as the reference on the time axis, and the linear interpolation method is used to generate matching time points: if the temperature data is T1 at t1 and T2 at t2 (t2-t1=60 seconds), then at each optical / motion data time node t (t1<t<t2) between t1 and t2, the interpolated temperature value T=T1+(T2-T1)×(t-t1) / (t2-t1) is obtained, ensuring that each optical / motion data time point has corresponding temperature time sequence data, and finally achieving accurate alignment of multi-source feature parameters (such as acceleration peak value, heart rate value, and interpolated temperature at t) on the same time node, with a time axis error of ≤5 ms.

[0107] Specifically, in some possible embodiments, for the data invalidation caused by abnormal sensor wearing state, the method further includes:

[0108] Obtaining a micro-displacement signal output by the motion sensing module and a signal intensity distribution feature collected by the optical sensing module;

[0109] Based on a normal fluctuation threshold of the micro-displacement signal and a uniformity threshold of the signal intensity distribution feature, a wearing state evaluation model is constructed, which is used to quantitatively determine whether the current wearing is stable;

[0110] When the wearing state evaluation model detects that the signal intensity of the optical sensing module decreases by more than a threshold compared with the normal reference, and the micro-displacement signal of the motion sensing module exceeds the normal fluctuation threshold, it is determined that the wearing is abnormal;

[0111] Based on the signal intensity distribution feature of the optical sensing module, the sampling region is automatically adjusted to a sub-region where the signal intensity meets the normal reference, and the credibility level of the data in this period is marked according to the duration of the abnormality, and the data with low credibility and below are not included in the core health evaluation;

[0112] If the abnormal wearing state lasts for more than a preset duration, the vibration reminder of the intelligent ring is triggered or the wearing adjustment prompt is pushed to the bound terminal to avoid long-term invalid data collection.

[0113] Wherein, when the micro-displacement signal output by the motion sensing module and the signal intensity distribution characteristics collected by the optical sensing module are acquired, the motion sensing module adopts a three-axis micro-electromechanical acceleration sensor, switches to a low-power high-sensitivity mode (sampling frequency 10 Hz, range ±2g, resolution 12 bits), and specially captures the micro relative displacement between the finger and the ring (such as <0.1 m / s² acceleration change generated by ring sliding and loosening), outputs a set of micro-displacement signal data (including instantaneous acceleration values of X, Y, Z axes and displacement fluctuation standard deviation within 500 ms) every 500 ms; the optical sensing module divides the sampling area into 4 annular sub-areas (central area, inner ring area, middle ring area, outer ring area), independently collects reflected light signal intensity (unit: lux) in each sub-area, outputs signal intensity values of 4 sub-areas and overall average intensity value every 1 second, forms signal intensity distribution characteristics (such as intensity difference of each sub-area, overall intensity trend), and all data are transmitted to the processing unit cache in real time.

[0114] When the normal fluctuation threshold of the micro-displacement signal and the uniformity threshold of the signal intensity distribution characteristics are used to construct a wearing state evaluation model, first, the baseline parameters are determined through first wearing calibration: let the user wear the ring at the root of the commonly used finger (without loosening), continuously collect micro-displacement signals for 5 minutes, calculate the mean of the fluctuation standard deviation as 0.03 m / s², and set the normal fluctuation threshold as "±0.05 m / s²" (mean ± 2 times standard deviation); at the same time, collect optical signal intensity distribution data for 5 minutes, calculate the mean of the coefficient of variation (standard deviation / mean) of the intensity of 4 sub-areas as 12%, set the uniformity threshold as "coefficient of variation ≤15%", and set the overall average intensity baseline value as the average intensity during calibration period (such as 80 lux). The evaluation model adopts a quantitative scoring mechanism: the fluctuation standard deviation of the micro-displacement signal is within the normal threshold and gets 40 points, and exceeds the threshold and gets points according to the exceeding proportion (such as 0.06 m / s² gets 30 points, 0.1 m / s² gets 0 points); the coefficient of variation of the optical signal intensity is ≤15% and the overall average intensity is ≥80% (64 lux) of the baseline value, gets 60 points, the coefficient of variation exceeds 5% every time, gets 15 points, and the overall intensity is low by 10% every time, gets 20 points; the model total score = displacement score + optical score, 80-100 points for "stable wearing", 40-79 points for "mild abnormality", and 0-39 points for "serious abnormality", realizing quantitative judgment of wearing state.

[0115] When the wearing state evaluation model detects an anomaly, the decision logic is "both conditions are met at the same time": the overall average intensity of the optical sensing module drops by more than 30% (i.e. <56 lux) from the reference value (80 lux), and the standard deviation of the micro-displacement signal of the motion sensing module fluctuates continuously for 3 times (1.5 seconds) beyond the normal threshold (>0.05 m / s²), at which point the model total score drops to below 40 points, and it is determined that the wearing is abnormal (e.g. the ring slides towards the fingertip, causing insufficient contact between the optical module and the skin, while generating continuous micro-displacement). If only a single condition is met (e.g. the optical intensity drops but the displacement is normal, which may be temporary obstruction), it is determined as "interference" rather than "wearing abnormality" to avoid misjudgment.

[0116] When the signal intensity distribution characteristics of the optical sensing module are automatically adjusted to adjust the sampling area, the processing unit analyzes the real-time intensity values of the 4 sub-areas, filters out the sub-area with signal intensity ≥ 80% of the reference value (64 lux) and coefficient of variation ≤ 15% (e.g. inner ring area intensity 72 lux, coefficient of variation 10%), and controls the light source and photosensitive element of the optical module to switch to the sub-area through the GPIO pin, setting it as the new sampling area (switching response time < 200 ms); at the same time, the data reliability level is marked according to the duration of the anomaly: if the anomaly lasts for <10 seconds, it is marked as "high reliability" (the data can be used normally), if it lasts for 10-30 seconds, it is marked as "medium reliability" (the data needs to be verified in combination with other modules for use), and if it lasts for >30 seconds, it is marked as "low reliability" (the data is not included in the core health assessment, and is only used for anomaly recording). For example, if the wearing abnormality lasts for 40 seconds, the heart rate variability data during this period is marked as low reliability and does not participate in the calculation of the resting heart rate.

[0117] If the wearing abnormality state lasts for more than a preset duration (set to 60 seconds), the processing unit triggers the following reminder mechanism: the built-in micro-vibration motor of the smart ring vibrates at a frequency of 2 Hz, each vibration lasts for 0.5 seconds and is separated by 1 second, and the vibration is repeated for 3 times; at the same time, the smart ring pushes a wearing adjustment prompt to the bound mobile phone APP through Bluetooth, the prompt content is "detected wearing abnormality (offset / loose), please adjust to the root of the finger to fit the skin and ensure data accuracy", the APP displays the prompt in the form of a pop-up window, accompanied by a mobile phone notification bell sound, guiding the user to adjust the wearing state in time; if the wearing state has not returned to normal within 120 seconds after the reminder, the processing unit controls the optical sensing module and the temperature sensing module to enter sleep mode, only keeping the motion sensing module to monitor the displacement, to avoid long-term invalid data collection consuming power.

[0118] Specifically, in some possible embodiments, in view of individual physiological differences leading to insufficient applicability of the evaluation results, the step of generating health evaluation results comprises:

[0119] At first use, through multi-scene data collection for a preset number of consecutive days, a user-specific physiological baseline including the following parameters is established: resting heart rate baseline, basal body temperature fluctuation range, heart rate response threshold in dynamic scenes;

[0120] When generating the health assessment result subsequently, the real-time collected physiological data is compared with the user-specific physiological baseline;

[0121] When it is detected that the real-time data exceeds the preset deviation range of the user-specific physiological baseline for a preset number of consecutive days, it is determined that the user's physiological baseline has long-term drift, the corresponding parameter in the user-specific physiological baseline is automatically updated, and the subsequent health assessment result is regenerated based on the updated user-specific physiological baseline.

[0122] Wherein, after the smart ring is activated for the first time after pairing, a 7-day baseline collection period is automatically started, covering three typical scenes of static (such as morning rest, office), dynamic (such as walking, light exercise), and sleep. The resting heart rate baseline is collected 10 minutes after getting up every day (when the user is not active), and the optical sensing module continuously collects 5 minutes of heart rate data at a sampling frequency of 10 Hz. The first minute of transition data is removed, and the average heart rate of the last 4 minutes is taken as the resting heart rate of the day. After removing 1 maximum value and 1 minimum value from the 7-day data, the average value is taken as the user-specific resting heart rate baseline (such as 65 times / minute). The basal body temperature fluctuation range is collected for 24 hours a day, and the temperature sensing module samples once every 30 minutes. After 7 days, 336 data points are accumulated, and the mean value (such as 36.4°C) and the standard deviation (such as 0.2°C) are calculated. The "mean ± 2 x standard deviation" (36.0°C-36.8°C) is set as the basal body temperature fluctuation range. The heart rate response threshold in dynamic scenes is collected by guiding the user to complete 3 times of 10-minute moderate intensity exercise (such as brisk walking), and the difference between the maximum exercise heart rate and the resting heart rate baseline (such as 35 times / minute) is recorded. The average of the 3 differences is taken as the heart rate response threshold (i.e. if the heart rate in dynamic scenes increases by more than this value compared with the baseline, it is determined that the "exercise load is moderate"). All baseline parameters are stored in the encrypted Flash partition (capacity 512KB) of the ring and synchronized to the mobile phone APP backup.

[0123] When generating the health assessment result, the processing unit compares the real-time physiological data with the exclusive baseline every 5 minutes: in the resting state (during the fixed morning period every day), the real-time heart rate value is compared with the resting heart rate baseline, and if it is within the range of "baseline ± 5 times / minute" (such as 60-70 times / minute), it is marked as "normal"; the real-time value of the basal body temperature is compared with the fluctuation range, and if it is within 36.0°C-36.8°C, it is marked as "stable"; in the dynamic scene, the real-time heart rate increase value is compared with the heart rate response threshold, and if it is ≤80% of the threshold (such as ≤28 times / minute), it is marked as "low exercise intensity", 80%-120% (28-42 times / minute) is marked as "appropriate", and >120% (>42 times / minute) is marked as "high", and the comparison result is used as the core basis for health assessment, for example, the dynamic scene assessment result is "current heart rate 100 times / minute, increased by 35 times / minute compared with the resting baseline, and in the appropriate exercise intensity".

[0124] When it is detected that the real-time data exceeds the preset deviation range of the exclusive physiological baseline for a continuous preset number of days (set to 3 days), the baseline is automatically updated: the deviation range is set to the resting heart rate baseline ±10% (such as 65 times / minute corresponding to 58-72 times / minute), the basal body temperature fluctuation range ±0.3°C (such as 36.0°C-36.8°C corresponding to 35.7°C-37.1°C), and the heart rate response threshold ±20% (such as 35 times / minute corresponding to 28-42 times / minute). If the average value of the resting heart rate in the morning for 3 consecutive days is 58 times / minute (lower than the original baseline by 10%), it is determined that the resting heart rate baseline has long-term drift, and the new baseline is updated to the average value of 58 times / minute for the 3 days; if the average value of the basal body temperature for 3 consecutive days rises to 36.6°C and the fluctuation range expands to 36.2°C-37.0°C, the temperature fluctuation range is updated to 36.2°C-37.0°C; the updated baseline parameter covers the original data, and the update time and reason (such as "2024-09-11, the resting heart rate is low for 3 consecutive days, and the resting baseline is updated to 58 times / minute") are recorded in the mobile phone APP, and the subsequent health assessment is generated based on the updated baseline, for example, the new assessment result is adjusted to "current heart rate 95 times / minute, increased by 37 times / minute compared with the resting baseline (58 times / minute), and in the appropriate exercise intensity", to ensure that the assessment result always adapts to the current physiological state of the user.

[0125] Another embodiment of the present application provides an intelligent ring health monitoring system based on multi-sensor cooperation, wherein Figure 2 An intelligent ring health monitoring system based on multi-sensor cooperation, comprising:

[0126] The sensing data acquisition module 100 acquires sensing data, which includes acceleration signals output by the motion sensing module, initial heart rate variability data collected by the optical sensing module, and basal body temperature detected by the temperature sensing module;

[0127] The current monitoring scene type determination module 200: based on the sensing data, combined with the preset multi-dimensional scene judgment rule for analysis, and the judgment result is corrected by synchronously referring to the user preset activity log, the current monitoring scene type is determined;

[0128] The multi-source monitoring data acquisition module 300: according to the determined scene type, control each sensing module to execute the differentiated data acquisition strategy: the low-power sampling mode is adopted in the static scene; the sampling frequency of the optical sensing module and the motion sensing module is improved in the dynamic scene and the data of the two is kept synchronous; the sampling interval of the temperature sensing module and the body motion detection sensitivity of the motion sensing module are optimized in the sleep scene; and the corresponding module is controlled to enter the sleep state to reduce the energy consumption in the non-acquisition period, so as to obtain the multi-source monitoring data;

[0129] The data fusion analysis result output module 400: the multi-source monitoring data obtained through the differentiated acquisition strategy is sequentially executed signal level noise reduction processing, feature level parameter extraction and decision level fusion analysis steps, to realize the deep integration and accurate analysis of multi-dimensional data, and obtain the data fusion analysis result;

[0130] The health assessment result generation module 500: based on the data fusion analysis result, the health assessment result matched with the current scene type is generated: the real-time physiological parameters and the motion adaptation suggestion are output in the dynamic scene; the sleep cycle analysis result is output in the sleep scene; the basic physiological parameter trend feedback is output in the static scene.

[0131] The intelligent ring health monitoring system based on multi-sensor cooperation provided in the embodiment can realize the steps of the foregoing embodiment, and thus can achieve the same technical effects as the foregoing embodiment. The principle analysis can be referred to the related description of the steps of the foregoing method.

[0132] The embodiment of the application further provides an intelligent ring health monitoring device based on multi-sensor cooperation, comprising a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to execute the foregoing method.

[0133] The embodiment of the application further provides a storage medium, wherein the storage medium stores a computer program capable of being loaded and executed by the processor to execute the foregoing method.

[0134] The storage medium provided by the embodiment can achieve the same technical effects as the foregoing embodiments, and the principle analysis can be referred to the related description of the foregoing method steps, which will not be repeated here.

[0135] The storage medium includes, for example, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage medium capable of storing program codes.

[0136] The steps of the method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module can be stored in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0137] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples, without contradiction.

[0138] In addition, the term "first", "second" feature can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited, only for the purpose of description, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features.

[0139] Thus, any process or method described in a flowchart or otherwise described herein can be understood as representing a module, segment, or portion of code that includes one or more executable instructions for implementing the specified logical functions or steps, and the various embodiments of the application include additional or fewer steps or processes in alternative implementations. As will be understood by those skilled in the art, the steps or processes of the various embodiments of the application can be carried out in any order, including simultaneously, sequentially, or in reverse order, without departing from the scope of the application.

[0140] The embodiments of the present application are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, and therefore: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A multi-sensor collaboration based intelligent ring health monitoring method, characterized in that, The method comprises the following steps: acquiring sensing data, wherein the sensing data comprises acceleration signals output by a motion sensing module, initial heart rate variability data collected by an optical sensing module, and basal body temperature detected by a temperature sensing module; based on the sensing data, analyzing in combination with a preset multi-dimensional scene judgment rule, and synchronously referring to a user's preset activity log to correct the judgment result, to determine a current monitoring scene type; in view of multi-feature conflicts in scene judgment, the step of determining the current monitoring scene type comprises: when different dimensional features in the sensing data point to different scene types, extracting conflict features; calling historical scene conversion data, constructing a feature weight model based on scene contribution degrees of the conflict features in the historical data; according to the feature weight model, assigning dynamic weights associated with current feature matching degrees to the conflict features, obtaining a comprehensive scene score through weighted calculation, and taking the type with the highest comprehensive scene score as the final scene type; according to the determined scene type, controlling each sensing module to execute a differentiated data acquisition strategy: in a static scene, a low-power sampling mode is adopted; in a dynamic scene, the sampling frequency of the optical sensing module and the motion sensing module is increased and the data of the two modules is kept synchronized; in a sleep scene, the sampling interval of the temperature sensing module and the body motion detection sensitivity of the motion sensing module are optimized; and in a non-acquisition period, the corresponding module is controlled to enter a sleep state to reduce energy consumption, so as to acquire multi-source monitoring data; for the multi-source monitoring data obtained through the differentiated acquisition strategy, signal-level noise reduction processing, feature-level parameter extraction, and decision-level fusion analysis steps are sequentially executed to realize deep integration and accurate analysis of multi-dimensional data, and a data fusion analysis result is obtained; in view of motion interference leading to optical signal distortion in a dynamic scene, the signal-level noise reduction processing step comprises: extracting time domain features of the acceleration signals of the motion sensing module; establishing a mapping relationship between the time domain features of the acceleration signals and optical signal distortion modes; wherein the optical signal distortion modes are baseline drift, peak attenuation, and phase shift of the optical sensing module caused by motion interference; when periodic motion interference is detected, based on the corresponding rules of action period and distortion period in the mapping relationship, the optical signal is periodically compensated through interference mode inverse operation to offset the baseline drift and phase shift synchronized with the action period; when non-periodic motion interference is detected, based on the correlation level of amplitude peak value and distortion intensity in the mapping relationship, the distortion type and degree of the signal loss segment are identified, a temporary interpolation algorithm matched with the distortion degree is enabled to fill the signal loss segment, and the continuity of the heart rate variability data is ensured; based on the data fusion analysis result, a health assessment result matched with the current scene type is generated: in a dynamic scene, real-time physiological parameters and motion adaptation suggestions are output; in a sleep scene, sleep cycle analysis results are output; and in a static scene, basic physiological parameter trend feedback is output.

2. The multi-sensor collaboration based smart ring health monitoring method according to claim 1, wherein, in view of the contradiction between low power consumption and data integrity, the differentiated acquisition strategy comprises: In the non-collection period, a low-power wake-up signal output by the motion sensing module is monitored in real time, the low-power wake-up signal being a signal detected by the module in a sleep state and conforming to a preset health-related action feature; When the low-power wake-up signal is detected, the corresponding optical sensing module or temperature sensing module is woken up based on the action feature of the low-power wake-up signal, short-time high-frequency collection matching the action duration is performed, and physiological parameter changes associated with the action are ensured to be captured; After the collection is completed, the related module is controlled to enter the sleep state again according to the health importance of the low-power wake-up signal.

3. The multi-sensor collaboration based smart ring health monitoring method according to claim 1, wherein, To address fusion deviation caused by different time sequences of multi-source data, the feature level parameter extraction includes: A high-precision timestamp is added to the collection data of each sensing module; Through a clock synchronization mechanism, the time offset between the motion sensing module, the optical sensing module and the temperature sensing module is calculated based on the high-precision timestamp, and the collection data of each sensing module is time calibrated according to the time offset, so that the time bases of the multi-source feature data collected by the three types of modules are kept consistent; For the temperature sensing data with collection delay, a temperature sensing time sequence point matching the time node of the optical sensing module / motion sensing module data is generated on the time axis by using a linear interpolation method based on the time difference between the high-precision timestamp of the temperature sensing data and the high-precision timestamp of the optical sensing module / motion sensing module data, so as to ensure that the multi-source feature parameters are aligned at the same time node.

4. The multi-sensor collaboration based smart ring health monitoring method of claim 1, wherein, To address data invalidation caused by abnormal sensor wearing state, the method further includes: A micro-displacement signal output by the motion sensing module and a signal intensity distribution feature collected by the optical sensing module are obtained; A wearing state evaluation model is constructed based on a normal fluctuation threshold of the micro-displacement signal and a uniformity threshold of the signal intensity distribution feature, the model being used for quantitatively judging whether the current wearing is stable; When the signal intensity of the optical sensing module is detected by the wearing state evaluation model to be sharply reduced by more than a threshold value from a normal reference, and accompanied by the micro-displacement signal of the motion sensing module being beyond a normal fluctuation threshold, it is determined that the wearing is abnormal; Based on the signal intensity distribution feature of the optical sensing module, a sampling region is automatically adjusted to a sub-region with signal intensity conforming to the normal reference, and the credibility level of the data in the period is marked according to the abnormal duration, and data with low credibility and below is not included in the core health evaluation; If the abnormal wearing state lasts for more than a preset duration, vibration reminding of the smart ring is triggered or a wearing adjustment prompt is pushed to a bound terminal, so as to avoid long-term invalid data collection.

5. The multi-sensor collaboration based smart ring health monitoring method according to claim 1, wherein, To address insufficient applicability of evaluation results caused by individual physiological differences, the steps of generating the health evaluation result include: When used for the first time, a user-specific physiological reference including the following parameters is established through multi-scene data collection for a continuous preset number of days: resting heart rate baseline, basic body temperature fluctuation range, heart rate response threshold in a dynamic scene; When the health evaluation result is generated subsequently, the real-time collected physiological data is compared with the user-specific physiological reference. When it is detected that real-time data continuously exceeds the preset deviation range of the user-specific physiological baseline for a preset number of days, it is determined that the user's physiological baseline has long-term drift, the corresponding parameter in the user-specific physiological baseline is automatically updated, and the subsequent health assessment result is regenerated based on the updated user-specific physiological baseline.

6. An intelligent ring health monitoring system based on multi-sensor collaboration, characterized in that, Comprise: A sensing data acquisition module: acquire sensing data, the sensing data including acceleration signals output by a motion sensing module, initial heart rate variability data collected by an optical sensing module, and basic body surface temperature detected by a temperature sensing module; A current monitoring scene type determination module: based on the sensing data, analyze in combination with a preset multi-dimensional scene determination rule, and simultaneously correct the determination result by referring to a user preset activity log to determine a current monitoring scene type; To address multi-feature conflicts in scene determination, the step of determining the current monitoring scene type includes: when different dimensional features in the sensing data point to different scene types, extract conflict features; call historical scene conversion data, construct a feature weight model based on the scene contribution of the conflict features in the historical data; according to the feature weight model, assign a dynamic weight associated with the current feature matching degree to the conflict features, obtain a comprehensive scene score through weighted calculation, and take the type with the highest comprehensive scene score as the final scene type; A multi-source monitoring data acquisition module: according to the determined scene type, control each sensing module to execute a differentiated data acquisition strategy: for a static scene, use a low-power sampling mode; for a dynamic scene, increase the sampling frequency of the optical sensing module and the motion sensing module and keep their data synchronized; for a sleep scene, optimize the sampling interval of the temperature sensing module and the body motion detection sensitivity of the motion sensing module; and control the corresponding module to enter a sleep state to reduce energy consumption during a non-acquisition period, thereby acquiring multi-source monitoring data; A data fusion analysis result output module: for the multi-source monitoring data obtained through the differentiated acquisition strategy, sequentially execute signal level noise reduction processing, feature level parameter extraction, and decision level fusion analysis steps to realize deep integration and accurate analysis of multi-dimensional data, and obtain a data fusion analysis result; to address optical signal distortion caused by motion interference in a dynamic scene, the signal level noise reduction processing step includes: extracting time domain features of the acceleration signals of the motion sensing module; establishing a mapping relationship between the time domain features of the acceleration signals and optical signal distortion patterns; wherein the optical signal distortion patterns are baseline drift, peak decay, and phase shift of the optical sensing module caused by motion interference; when periodic motion interference is detected, based on the corresponding rules of action period and distortion period in the mapping relationship, periodically compensate the optical signal through interference pattern inverse operation to offset the baseline drift and phase shift synchronized with the action period; when non-periodic motion interference is detected, based on the correlation level of amplitude peak value and distortion intensity in the mapping relationship, identify the distortion type and degree of the signal loss segment, enable a temporary interpolation algorithm matching the distortion degree to fill the signal loss segment, and ensure the continuity of the heart rate variability data; The health assessment result generation module generates a health assessment result matched with the current scene type based on the data fusion analysis result: the dynamic scene outputs real-time physiological parameters and exercise adaptation suggestions; the sleep scene outputs sleep cycle analysis results; and the static scene outputs basic physiological parameter trend feedback.

7. An intelligent ring health monitoring device based on multi-sensor collaboration, characterized in that, Comprise: A memory and a processor, the memory has a computer program loaded and executed by the processor and capable of executing the health monitoring method of the smart ring based on multi-sensor cooperation according to any one of claims 1-5.

8. A storage medium, characterized by The memory has a computer program loaded and executed by the processor and capable of executing the health monitoring method of the smart ring based on multi-sensor cooperation according to any one of claims 1-5.

Images