Health management information prompting method and device based on wearable equipment

Abstract

The invention relates to the technical field of data processing, in particular to a health management information prompting method and device based on wearable equipment, and the method comprises the steps: collecting first-class information and second-class information of a target object through the wearable equipment, and obtaining basic health information of the target object; the first type of information is used for representing physiological signals of the target object; the second type of information is used for representing the pulse waveform and the body surface state of the target object; the basic health information is used for representing historical body health examination information of the target object; analyzing the first type of information and the basic health information to obtain a first analysis result; analyzing the second type of information and the basic health information to obtain a second analysis result; based on a pre-trained multi-modal feature fusion model, performing fusion analysis on the first analysis result and the second analysis result to generate a comprehensive analysis result representing the physiological state of the target object; and according to the comprehensive analysis result, the wearable device carries out health management information prompting on the target object.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a health management information prompting method and device based on a wearable device. BACKGROUND

[0002] With the development of the Internet of Things and sensor technology, intelligent wearable devices have been widely used in the field of health monitoring. For example, by collecting physiological signals such as photoplethysmography (PPG) and electrocardiogram (ECG) through an intelligent wearable device, and collecting motion signals through a motion sensor, real-time monitoring of parameters such as heart rate, blood oxygen saturation, and motion state can be achieved, thereby providing users with basic health data reference.

[0003] However, in the related art, intelligent wearable devices (such as smart watches) rely on Western medicine quantitative indicators for data presentation, and usually only output basic parameters such as heart rate values and blood oxygen saturation values. Then, a simple comparison is made with the preset threshold corresponding to the heart rate and blood oxygen saturation, and the health management information of the user is generated according to the single analysis basis of "whether it is out of the normal range". This analysis method relying on only a small number of quantitative indicators has low accuracy, and different users have different physical conditions. Using a unified threshold for judgment can easily lead to false positives of the prompted health management information in a specific group of people, and the applicability is also poor.

[0004] In summary, the health management information prompting method in the related art has the problems of low accuracy and poor applicability. SUMMARY

[0005] Embodiments of the present application provide a health management information prompting method and device based on a wearable device, to solve the problem of low accuracy and poor applicability of the health management information prompting method in the related art.

[0006] In a first aspect, the present application provides a health management information prompting method based on a wearable device, the method comprising: collecting first type information and second type information of a target object through a wearable device; wherein the first type information is used to represent the physiological signals of the target object; and the second type information is used to represent the pulse waveform and body surface state of the target object; obtaining the basic health information of the target object, wherein the basic health information is used to represent the historical physical examination information of the target object; analyzing the first type information and the basic health information to obtain a first analysis result, and analyzing the second type information and the basic health information to obtain a second analysis result; The first analysis result and the second analysis result are fused and analyzed based on a pre-trained multi-modal feature fusion model to generate a comprehensive analysis result representing a physiological state of the target object. According to the comprehensive analysis result, the wearable device provides health management information prompts to the target object.

[0007] In some embodiments, the first type of information includes part or all of a photoplethysmogram (PPG) signal and an electrocardiogram (ECG) signal; wherein the PPG signal and the ECG signal are physiological signals collected by the wearable device from a first body part of the target object. The second type of information includes part or all of pulse waveform feature data and body surface state feature data, wherein the pulse waveform feature data is a pulse kinetics parameter obtained by waveform morphology analysis of the PPG signal, and the body surface state feature data is image data collected by an image collection device of the target device from a second body part of the target object; the target device is the wearable device or a terminal device in communication with the wearable device.

[0008] In this way, based on the wearable device collecting PPG, ECG signals and body surface images, non-invasive and continuous multi-modal data acquisition is realized, providing high-quality input for subsequent accurate analysis.

[0009] In some embodiments, the first type of information includes a PPG signal and an ECG signal. The analysis of the first type of information and the basic health information to obtain a first analysis result includes: According to the PPG signal, a heart rate parameter and an oxygen saturation parameter of the target object are determined. According to the ECG signal, a heart rhythm fluctuation parameter of the target object is identified. Based on the synchronization relationship between the PPG signal and the ECG signal, a pulse wave velocity (PWV) is calculated, and based on the PWV, a vascular elasticity parameter of the target object is determined. According to the basic health information, threshold ranges of the heart rate parameter, the oxygen saturation parameter, the heart rhythm fluctuation parameter, and the vascular elasticity parameter corresponding to the target object are determined. After comparing the heart rate parameter, the oxygen saturation parameter, the heart rhythm fluctuation parameter, and the vascular elasticity parameter with the corresponding threshold ranges, the first analysis result is generated according to the comparison results; the first analysis result is a state identifier representing the degree of deviation of physiological signals.

[0010] In this way, by using PPG and ECG synchronous analysis, key physiological parameters such as heart rate, blood oxygen, heart rhythm fluctuation and blood vessel elasticity are extracted, and a state identifier is generated in combination with individual threshold values, thereby improving the accuracy of the first analysis result.

[0011] In some embodiments, the second type of information includes pulse waveform feature data and body surface state feature data. The analysis of the second type of information and the basic health information obtains a second analysis result, including: Waveform morphology analysis is performed on the pulse waveform feature data to extract target waveform parameters, and a body constitution category matching the target waveform parameters is determined according to a mapping relationship between preset waveform parameters and body constitution categories, and a body constitution score is determined; Image recognition is performed on the body surface state feature data to extract target body surface appearance parameters, and a body surface state score matching the target body surface appearance parameters is determined according to a mapping relationship between preset body surface appearance parameters and body surface state scores; According to the basic health information, the weight coefficients of the body constitution score and the body surface state score in the fusion process are determined; The body constitution score is multiplied by the corresponding weight coefficient to obtain a first weighted value, and the body surface state score is multiplied by the corresponding weight coefficient to obtain a second weighted value; The first weighted value and the second weighted value are summed to generate the second analysis result.

[0012] In this way, by scoring the body constitution category and the body surface state respectively, and dynamically weighting and fusing based on the user's basic health information, a second analysis result is generated, thereby enhancing the accuracy of the second analysis result.

[0013] In some embodiments, the multi-modal feature fusion model is trained in the following manner: A plurality of sample data are obtained, wherein each sample data includes first type of sample information, second type of sample information, basic health sample information, and corresponding physiological state label, and the physiological state label is reference information generated based on expert annotation or observable physiological event; For each sample data, the first type of sample information and the basic health sample information are analyzed to obtain a first sample analysis result, and the second type of sample information and the basic health sample information are analyzed to obtain a second sample analysis result; The first sample analysis result and the second sample analysis result of each sample data are taken as input, and the corresponding physiological state label of each sample data is taken as target to train the multi-modal feature fusion model.

[0014] In this way, the multi-modal feature fusion model is trained by using real sample data, so that the model can learn the mapping relationship from the first and second analysis results to the physiological state, and the prediction reliability of the comprehensive analysis result is improved.

[0015] In some embodiments, the wearable device prompts the target object with health management information according to the comprehensive analysis result, including: By querying the preset health management information library, the health management information prompt content matched with the comprehensive analysis result is determined; According to the health management information prompt content, the wearable device prompts the target object.

[0016] In this way, the matching prompt content is generated by querying the preset health management information library, and output by the wearable device, thereby improving the user experience.

[0017] In some embodiments, before the wearable device prompts the target object with health management information according to the comprehensive analysis result, the method further includes: The target device receives the subjective body state feedback information input by the user; the target device is a wearable device or a terminal device connected in communication with the wearable device; The wearable device prompts the target object with health management information according to the comprehensive analysis result, including: The wearable device prompts the target object with health management information in combination with the subjective body state feedback information and the comprehensive analysis result.

[0018] In this way, the user's subjective body state feedback and the objective comprehensive analysis result are combined for prompting, so as to avoid false positives and false negatives, and improve the accuracy of health management prompts and user trust.

[0019] In some embodiments, the method further includes: Real-time monitoring of the AC component amplitude of the PPG signal; When it is detected that the AC component amplitude decreases by more than a preset threshold within a preset time, the signal quality of the PPG signal is calculated, and the acquisition state abnormality level of the PPG signal is determined based on the calculated signal quality; The signal recovery response strategy corresponding to the acquisition state abnormality level is executed.

[0020] In this way, by real-time monitoring of the AC component change of the PPG signal, the signal acquisition abnormality is identified in time and responded in stages, thereby improving the system stability and monitoring continuity.

[0021] In some embodiments, the signal recovery response strategy corresponding to the acquisition state abnormality level is executed, including: if the abnormality level is a low level, controlling the wearable device to re-collect and verify the PPG signal; if the abnormality level is a medium level, determining whether the target object is continuously stationary for a first preset time length, if yes, controlling the wearable device to start red light enhanced detection and determining whether the PPG signal recovers to normal within a second preset time length; if the PPG signal does not recover to normal within the second preset time length, controlling the wearable device to enhance light source intensity and receiving gain and determining whether the PPG signal recovers to normal within a third preset time length; if the PPG signal does not recover to normal within the third preset time length, controlling the target device to output a wearing state adjustment prompt; if the abnormality level is a high level, controlling the wearable device to enhance light source intensity and receiving gain and determining whether the PPG signal recovers to normal within a third preset time length; if the PPG signal does not recover to normal within the third preset time length, controlling the target device to output a wearing state adjustment prompt.

[0022] In this way, corresponding recovery strategies are executed according to different abnormality levels, and the response opportunity is optimized in combination with the stationary state of the user, so that the probability of generating inaccurate or even misleading health management information prompts is effectively reduced.

[0023] In a second aspect, the present application provides a wearable device-based health management information prompting device, the device comprising: a collection module, configured to collect first type information and second type information of a target object through a wearable device; wherein the first type information is used to represent the physiological signal of the target object; and the second type information is used to represent the pulse waveform and body surface state of the target object; an acquisition module, configured to acquire the basic health information of the target object, wherein the basic health information is used to represent the historical physical examination information of the target object; an analysis module, configured to analyze the first type information and the basic health information to obtain a first analysis result; and analyze the second type information and the basic health information to obtain a second analysis result; a fusion module, configured to perform fusion analysis on the first analysis result and the second analysis result based on a pre-trained multi-modal feature fusion model, to generate a comprehensive analysis result representing the physiological state of the target object; a prompting module, configured to perform health management information prompting on the target object by the wearable device according to the comprehensive analysis result.

[0024] In a third aspect, the present application provides an electronic device, comprising: a memory, configured to store program instructions; A processor configured to invoke program instructions stored in the memory to perform the steps of the method of any of the first aspect.

[0025] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, the computer program comprising program instructions which, when executed by a computer, cause the computer to perform the method of any of the first aspect.

[0026] In a fifth aspect, the present application provides a computer program product comprising computer program code which, when run on a computer, causes the computer to perform the method of any of the first aspect.

[0027] The present application has the following beneficial effects: In the embodiments of the present application, the first type of information, the second type of information and the basic health information of the target object are acquired by the wearable device; the first type of information is used to represent the physiological signals of the target object; the second type of information is used to represent the pulse waveform and the body surface state of the target object; the basic health information of the target object is acquired, and the basic health information is used to represent the historical physical examination information of the target object; the first type of information and the basic health information are analyzed to obtain a first analysis result; and the second type of information and the basic health information are analyzed to obtain a second analysis result; based on the pre-trained multi-modal feature fusion model, the first analysis result and the second analysis result are analyzed and fused to generate a comprehensive analysis result representing the physiological state of the target object; and according to the comprehensive analysis result, the wearable device provides health management information to the target object. In this way, by independently analyzing the “physiological signals + historical physical examination information” (first path) and the “pulse waveform / body surface state + historical physical examination information” (second path), the characteristic specificity of different modal information is retained, and the individualization degree of the analysis is enhanced by combining the individual historical data. Through the deep fusion of the pre-trained multi-modal feature fusion model, the problem of single modal information being easily disturbed by noise and high misjudgment rate is effectively overcome, the accuracy of the health management information prompt is significantly improved, and the applicability of the health management information prompt is improved by introducing the historical physical examination information of the individual user.

[0028] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1This is a schematic diagram illustrating an application scenario of a health management information notification method provided in an embodiment of this application. Figure 2 A flowchart illustrating a method for displaying health management information, as provided in an embodiment of this application; Figure 3 A flowchart for generating a first analysis result is provided in an embodiment of this application; Figure 4 A flowchart for generating a second analysis result is provided in an embodiment of this application; Figure 5 A flowchart illustrating a training method for a multimodal feature fusion model provided in this application embodiment; Figure 6 This is a schematic diagram of a health management information prompt provided in an embodiment of this application; Figure 7 This is a schematic diagram of an anomaly detection process provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of a health management information prompting device provided in an embodiment of this application; Figure 9 This is a schematic diagram of the hardware structure of an electronic device for implementing a method for prompting health management information, provided in an embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0031] The terms "first" and "second" in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising" and any variations thereof are intended to cover non-exclusive protection. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. The term "multiple" in this application can mean at least two, for example, two, three, or more, and the embodiments of this application do not impose limitations.

[0032] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These embodiments should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that in the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solutions of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0033] In the technical solution of this application, all data is collected after obtaining user permission, and the acquisition, transmission, storage, and use of data all comply with the requirements of relevant national laws and regulations.

[0034] The following description, in conjunction with the accompanying drawings, illustrates some preferred embodiments of this application.

[0035] See Figure 1 , Figure 1 This is a schematic diagram of an application scenario for a health management information prompting method based on a wearable device, provided in an embodiment of this application. It includes a wearable device 100, a terminal device 110 and a server 120 that are communicatively connected to the wearable device 100.

[0036] In this application embodiment, the wearable device 100 may be a smartwatch, and the terminal device 110 may include, but is not limited to, mobile phones, tablets, laptops, desktop computers, smart voice interaction devices, etc. The wearable device 100 and the terminal device 110 may have health management related clients installed. The client may be health management software, or it may be a webpage, mini-program, etc. The server 120 is a backend server corresponding to the software, webpage, mini-program, etc., or a server specifically used for health management information prompts. This application does not make specific limitations.

[0037] Server 120 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0038] It should be noted that the health management information prompting method based on wearable devices in the various embodiments of this application can be executed by an electronic device, which can be a wearable device 100, a terminal device 110, or a server 120. That is, the method can be executed by the wearable device 100, the terminal device 110, or the server 120 alone, or it can be executed by the wearable device 100, the terminal device 110, and the server 120 together.

[0039] For example, when server 120 executes alone, the target object uses wearable device 100 to collect the first type of information and the second type of information of the target object, and uploads them to server 120 according to a preset upload cycle or in response to the acquisition request of server 120. Server 120 analyzes the first type of information and basic health information to obtain the first analysis result; and server 120 analyzes the second type of information and basic health information to obtain the second analysis result; further, server 120 performs fusion analysis on the first analysis result and the second analysis result based on a pre-trained multimodal feature fusion model to generate a comprehensive analysis result representing the physiological state of the target object. Finally, server 120 provides health management information prompts to the target object based on the comprehensive analysis results. The prompt content can be sent to the client of wearable device 100 and displayed by the visual interface of the client of wearable device 100.

[0040] Optionally, the wearable device 100, the terminal device 110, and the server 120 can communicate via a communication network.

[0041] Optionally, the communication network can be a wired network or a wireless network.

[0042] It should be noted that, Figure 1 This is merely an example of an application scenario and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0043] See Figure 2 , Figure 2 The flowchart illustrates a method for prompting health management information based on a wearable device, as provided in this application embodiment, and includes the following steps.

[0044] In step 201, a first type of information and a second type of information of the target object are collected through a wearable device; wherein, the first type of information is used to characterize the physiological signals of the target object; and the second type of information is used to characterize the pulse waveform and body surface state of the target object.

[0045] The first type of information includes some or all of the PPG and ECG signals. The PPG and ECG signals can be physiological signals collected from the first body part of the target object through a wearable device. For example, when the wearable device is a smartwatch, the first body part can be the wrist, and when the wearable device is a chest strap heart rate monitoring device, the first body part can be the area around the chest. The following description uses a smartwatch as an example.

[0046] In practice, wearable devices can be equipped with a variety of sensors, including pulse wave photoelectric sensors, electrocardiogram detection chips, accelerometers, body temperature sensors, Bluetooth modules, etc.

[0047] The PPG signal is collected by a pulse wave photoelectric sensor, typically located on the wrist of a smartwatch. This sensor typically includes a set of photodiodes (each set consisting of two independent photodiodes) and three types of light-emitting diodes (LEDs) (green, red, and infrared). It works by using LEDs to illuminate the skin; blood flow causes subtle changes in skin color, which in turn affect the reflected light signal. The photodiodes detect the light reflected from the skin and convert it into an electrical signal. By analyzing these subtle changes in the light signal, the pulse wave photoelectric sensor can acquire a signal synchronized with the heartbeat—the pulse waveform.

[0048] ECG signals are collected by an electrocardiogram detection chip, which is usually located between the crown and the back of the smartwatch, forming a loop. It can be synchronously sampled with the pulse wave photoelectric sensor to ensure accurate calculation of pulse wave conduction velocity.

[0049] An accelerometer can support the watch's motion detection function, enabling accurate step counting, calorie consumption, distance measurement, and other sports-related functions.

[0050] The body temperature sensor measures temperature by contact with the arterial area (accuracy ±0.1℃) and uses Bluetooth to achieve real-time / offline dual-mode data transmission.

[0051] Optionally, the first type of information may also include body composition-related parameters obtained based on bioelectrical impedance analysis (BIA), such as body mass index (BMI) and basal metabolic rate (BMR), to reflect the user's body fat level and energy metabolism trend.

[0052] The second type of information includes some or all of the pulse waveform characteristic data and body surface condition characteristic data. The pulse waveform characteristic data consists of pulse dynamic parameters obtained through waveform morphology analysis of PPG signals, such as at least one of the following: main wave rise time, dicrotic wave delay, pulse rate variability, pulse wave propagation time, rising limb slope, enhancement index, pulse wave amplitude, and waveform symmetry. These pulse dynamic parameters are quantifiable, engineered physiological indicators suitable for long-term trend tracking and health status assessment. The body surface condition characteristic data consists of image data acquired from a second body part of the target object using the image acquisition device of the target device. The target device is a wearable device or a terminal device (such as a smartphone) that communicates with a wearable device. When the target device is a smartwatch, the image acquisition device can be a miniature camera integrated into the smartwatch; when the target device is a smartphone, the image acquisition device can be a rear or front-facing camera, working in conjunction with the wearable device via Bluetooth or Wireless Fidelity (Wi-Fi). The second body part can be the target's face, neck, tongue, or other exposed skin areas, used to assist in analyzing changes in skin color, micro-expressions, and sweating.

[0053] It should be noted that the pulse waveform features in this application are pulse dynamic parameters extracted from physiological signals acquired by a PPG sensor, which are quantifiable engineering parameters and belong to general technical features in the field of biomedical signal processing. Its purpose is for non-medical health status monitoring and management, such as daily vitality assessment, stress level trend analysis, and exercise recovery suggestion generation, and it is not used for any clinical disease diagnosis, pathological judgment, or medical treatment decision-making.

[0054] In step 202, basic health information of the target object is obtained, which is used to characterize the target object's historical physical health examination information.

[0055] Among them, basic health information is used to characterize the target object's historical physical health examination information, including but not limited to historical physical examination reports, routine health examination data, and previous physical indicator records, to establish a user's individualized health baseline model and support the trend comparison of subsequent physiological data and the generation of personalized feedback.

[0056] In step 203, the first type of information and basic health information are analyzed to obtain the first analysis result; and the second type of information and basic health information are analyzed to obtain the second analysis result.

[0057] In practice, when the first type of information includes PPG and ECG signals, the first type of information and basic health information can be analyzed in the following ways to obtain the first analysis result.

[0058] See Figure 3 , Figure 3 A flowchart for generating a first analysis result is provided for an embodiment of this application, and includes the following steps.

[0059] In step 301, the heart rate parameters and blood oxygen saturation parameters of the target object are determined based on the PPG signal.

[0060] The PPG signal is generated by illuminating the skin with an LED light source and detecting light reflection fluctuations caused by changes in blood volume. Each pulse peak corresponds to one heartbeat, allowing for the calculation of heart rate. Furthermore, the difference in absorption between red and infrared light can be used to calculate blood oxygen saturation.

[0061] For example, after a user wears a smartwatch for 30 minutes of exercise, the device collects PPG signals: If the pulse wave frequency is 140 peaks per minute, then the heart rate is 140 bpm. Based on the ratio of red light to infrared light absorption, the blood oxygen saturation can be calculated to be 96%.

[0062] In step 302, the heart rhythm fluctuation parameters of the target object are identified based on the ECG signal.

[0063] Among them, the ECG signal reflects the electrical activity of the heart, with clear P waves, QRS complexes, and T waves. The heart rhythm fluctuation parameters include the standard deviation of normal-to-normal intervals (SDNN), heart rate variability, and abnormal waveform detection.

[0064] For example, when a user is at rest, a 30-second ECG signal is collected. The SDNN value is 45ms, the heart rate variability is 58s (indicating moderate parasympathetic activity), and no abnormal waveforms are found (stable heart rhythm fluctuations).

[0065] In step 303, based on the synchronization relationship between the PPG signal and the ECG signal, the pulse wave velocity (PWV) is calculated, and the vascular elasticity parameters of the target object are determined based on the PWV.

[0066] Pulse wave velocity (PWV) is an important indicator for measuring vascular elasticity. The R wave origin of an ECG represents the start of cardiac electrical excitation, while the pulse wave origin of a PPG represents the time it takes for blood to reach the periphery (such as the wrist). The time difference between the two is the pulse wave conduction time. Based on the user's height or arm length, the pulse wave velocity (PWV) can be calculated. A higher PWV indicates poorer vascular elasticity, while a lower PWV indicates better vascular elasticity.

[0067] For example, if user A's arm is approximately 75cm long, and the time difference between the R-wave onset time of the ECG and the pulse wave onset time of the PPG is calculated to be 0.12s, then the PWV is... Furthermore, based on the PWV range of healthy adults, such as less than 8 m / s, it can be determined that user A's vascular elasticity is within the normal range.

[0068] In step 304, based on basic health information, the threshold ranges for heart rate parameters, blood oxygen saturation parameters, heart rhythm fluctuation parameters, and vascular elasticity parameters corresponding to the target object are determined respectively.

[0069] In practice, personalized judgment criteria can be established for users based on basic health information to avoid misjudgment. For example, a low resting heart rate is normal for athletes, but may be misjudged as abnormal according to general standards.

[0070] For example, the statistics of user A's physical examination or daily monitoring data over the past 6 months are shown in Table 1.

[0071] Table 1

[0072] Therefore, 58–78 bpm can be defined as the threshold range for user A's heart rate parameter, 95%–99% as the threshold range for user A's blood oxygen saturation parameter, 40–60 ms as the threshold range for user A's heart rhythm fluctuation parameter, and 5.1–6.9 m / s as the threshold range for user A's vascular elasticity parameter.

[0073] In step 305, the heart rate parameter, blood oxygen saturation parameter, heart rhythm fluctuation parameter, and vascular elasticity parameter are compared with their respective threshold ranges, and a first analysis result is generated based on the comparison results; the first analysis result is a status indicator representing the degree of deviation of physiological signals.

[0074] The status indicator only reflects the deviation trend of physiological parameters from an individual's historical level, and is used to assist users in adjusting their lifestyle habits, such as increasing aerobic exercise, improving the sleep environment, and maintaining a regular schedule.

[0075] In practice, the degree of deviation can be determined based on the comparison results. For example, if all parameters are within the corresponding threshold range, the first analysis result is determined to be normal. If one parameter is not within the corresponding threshold range, the first analysis result is determined to be slightly deviated. If two parameters are not within the corresponding threshold range, the first analysis result is determined to be moderately deviated. If more than two parameters are not within the corresponding threshold range, the first analysis result is determined to be highly deviated, and a status indicator representing the degree of deviation is output.

[0076] For example, if user A's actual heart rate is 85 bpm, blood oxygen saturation is 93%, SDNN is 42 ms, and PWV is 6.5 m / s, based on the threshold range of user A's parameters, it can be known that the heart rate is greater than 78 bpm and the blood oxygen saturation is less than 95%. Therefore, it can be determined that two parameters are not within the corresponding threshold range. The first analysis result is that the physiological signal is moderately deviated, with the heart rate being greater than 78 bpm and the blood oxygen saturation being less than 95%.

[0077] In this way, personalized initial analysis results can be generated based on each user's Western medicine quantitative indicators corresponding to PPG and ECG signals, as well as their actual baseline health information. This can more accurately reflect an individual's health status and avoid the risk of misjudgment that may arise from a single standard. Based on the trend analysis of personalized threshold changes, the system can detect potential health problems in advance. Even if some indicators are still within the normal range, if a trend of continuous deviation from personal benchmarks is observed, the system can promptly remind users to adjust their lifestyle. By establishing and applying personalized judgment criteria, the effectiveness and reliability of health monitoring can be improved.

[0078] In practice, when the second type of information includes pulse waveform characteristic data and body surface condition characteristic data, the second type of information and basic health information can be analyzed in the following ways to obtain the second analysis result.

[0079] See Figure 4 , Figure 4 A flowchart for generating a second analysis result is provided in this application embodiment, and includes the following steps.

[0080] In step 401, waveform morphology analysis is performed on the pulse waveform feature data to extract target waveform parameters. Based on the mapping relationship between preset waveform parameters and body type, the body type that matches the target waveform parameters is determined and the body type score is determined.

[0081] In practice, pulse waveform characteristic data can include the upward slope, dicrotic wave delay time, waveform symmetry index, and the amplitude ratio of the main wave to the dicrotic wave. By performing waveform morphology analysis on the pulse waveform characteristic data, target waveform parameters can be obtained, and the level corresponding to each target waveform parameter can be determined according to the threshold range corresponding to each waveform parameter. For example, if the upward slope is in the range of 2.5-4.0 au / ms, the corresponding level is "fast rise"; if the dicrotic wave delay time is greater than 100 ms, the corresponding level is "long delay"; if the amplitude ratio of the main wave to the dicrotic wave (Augmentation Index, AIx) is less than 30%, the corresponding level is "low AIx"; and if the waveform symmetry index is between 0.9 and 1.1, it is judged as "symmetric".

[0082] Then, based on the level corresponding to each target waveform parameter, the target waveform parameters are combined to generate a target waveform parameter group. Next, based on the pre-defined mapping relationship between waveform parameters and constitution categories, the constitution category tending towards the target waveform parameter group is analyzed. These constitution categories include, but are not limited to: Qi deficiency, Yang deficiency, Phlegm-dampness, Damp-heat, Qi stagnation, Blood stasis, Yin deficiency, Special constitution, and Balanced constitution. The mapping relationship can be obtained through expert experience modeling, statistical regression, or training a pre-trained constitution category classification model, and stored in the system database. The pre-trained constitution category classification model can use Random Forest, XGBoost, Support Vector Machine, or shallow neural networks as classifiers.

[0083] For example, when a user's target waveform parameter set is "rapid rise, long delay, low AIx," the system identifies a high degree of match between their pulse characteristics and the "Qi deficiency constitution" category, thus determining that the user's constitution type tends to be Qi deficiency. Finally, by calculating the normalized membership degree of each target waveform parameter within its corresponding value range for the matched constitution category and taking their arithmetic mean, a comprehensive membership degree is obtained, generating a constitution score. The constitution score ranges from 0 to 100, with higher scores indicating a higher degree of match. This constitution score is used for subsequent fusion with other health data to generate a comprehensive health status assessment result, but is not used for clinical disease diagnosis, pathological judgment, or medical treatment decisions.

[0084] Assume that the mapping relationship between the preset waveform parameters and the body type is as shown in Table 2.

[0085] Table 2

[0086] As shown in Table 2, for example, based on the pulse waveform characteristic data of user A, the extracted target waveform parameters are: rising limb slope: 3.2 au / ms (indicating a rapid rise), dicrotic wave delay time: 120 ms (indicating a long delay), and the ratio of the main wave to the dicrotic wave amplitude (AIx) of 28% (indicating a low AIx), with a waveform symmetry index of 1.0 (indicating symmetry). The system compares this combination with the preset mapping relationship and determines that the most matching constitution type is "Qi deficiency".

[0087] Furthermore, for each waveform parameter value, the value is compared with the range of values ​​corresponding to the same waveform parameter in the corresponding physical constitution category, the membership degree is calculated, and the physical constitution score of user A is determined based on the average membership degree of each waveform parameter.

[0088] For example, according to calculations, the membership degrees of user A's waveform parameters are as follows: The ascending limb slope of 3.2 au / ms has a membership degree of 0.8 within the range of 2.5–4.0 au / ms corresponding to the Qi deficiency constitution; the diphthoplasty delay time of 120 ms has a membership degree of 1.0 within the range of >100 ms corresponding to the Qi deficiency constitution; the ratio of the main wave to the diphthoplasty amplitude, Aix 28%, has a membership degree of 0.9 within the range of <30% corresponding to the Qi deficiency constitution; and the waveform symmetry index of 1.0 has a membership degree of 1.0 within the range of 0.9–1.1 corresponding to the Qi deficiency constitution. Therefore, the overall membership degree is... Finally, the overall membership degree is linearly mapped to a 0-100 score to obtain the physical fitness score: .

[0089] For example, by directly inputting the waveform parameters of various targets of user A into the pre-trained constitution classification model, the matching probability of each constitution category output by the constitution classification model can be obtained. For example: Qi deficiency constitution: 85%, Yang deficiency constitution: 10%, Phlegm-dampness constitution: 5%. Then, based on the matching result with the highest probability, the constitution category that user A is most likely to match is "Qi deficiency constitution", and the matching probability is converted into a constitution matching degree score of 0-100 points (e.g., 85 points).

[0090] In this way, by leveraging the experience of traditional Chinese medicine's "palpation" to transform pulse waveform characteristic data into quantifiable constitution category scores, the accuracy of health management information prompts can be further enhanced. Furthermore, pulse waveform characteristic data can also serve as an assessment indicator for cardiovascular disease risk, enabling the synergistic application of traditional Chinese medicine identification and modern chronic disease early warning. Specifically, by simultaneously collecting the user's electrocardiogram (ECG) signal and photoplethysmography (PPG) signal, pulse wave velocity (PWV) is obtained. Further, the degree of arteriosclerosis is assessed based on the PWV value: when the PWV exceeds a preset threshold, an arteriosclerosis risk is identified. This arteriosclerosis degree information can be used as an auxiliary feature, input into the constitution classification model to further improve the accuracy of constitution category generation and generate personalized cardiovascular health assessment reports. In practical applications, the sampling rate of the PPG signal can also be intelligently controlled: during routine blood oxygen and heart rate monitoring, a low sampling rate mode is used to meet energy-saving needs; when the "extract pulse wave signal through PPG" function is activated, it will automatically switch to a high sampling rate mode to ensure PPG signal quality and improve the accuracy of pulse characteristic judgment. After sampling is complete, return to normal low sampling mode.

[0091] In step 402, image recognition is performed on the body surface state feature data to extract the target body surface appearance parameters, and the body surface state score that matches the target body surface appearance parameters is determined according to the mapping relationship between the preset body surface appearance parameters and the body surface state score.

[0092] Among them, the body surface state feature data is image data collected from a second body part (such as the face, neck, or back of the hand) of the target object by the image acquisition device of the target device (such as a smartwatch camera or a smartphone camera connected to it). Specifically, with the permission of the target object, corresponding image acquisition operations can be performed on the target object according to one or more interactive windows set on the target device. Optionally, the target device (wearable device or terminal device communicating with the wearable device) can provide a specified interactive window to the target object by installing a client related to image acquisition. The interactive window can be a software interface or a mini-program interface corresponding to the aforementioned client, or it can be one or more information display windows with interactive functions contained in the aforementioned software interface or mini-program interface.

[0093] Body surface appearance parameters refer to numerical features extracted from acquired image data through digital image processing, used to quantitatively describe the appearance of target body parts. These include, but are not limited to, skin color uniformity (based on RGB channel standard deviation), facial puffiness (based on facial key point distance variation analysis), and sweating level (based on skin area reflectivity variation). Finally, based on the mapping relationship between preset body surface appearance parameters and body surface status scores, a body surface status score matching the target body surface appearance parameters is determined. The score range can be set from 0 to 100 points; a higher score indicates a more stable and healthy body surface condition, suitable for non-medical health trend assessments.

[0094] For example, when user A takes a facial photo with a mobile phone camera and performs image recognition, the extracted surface appearance parameters are: a small standard deviation of skin color R / G / B indicates uniform color; slight puffiness under the eyes; and strong reflection in the forehead area indicates slight sweating.

[0095] Assume that the mapping relationship between the preset body surface appearance parameters and body surface status scores is as shown in Table 3.

[0096] Table 3

[0097] As shown in Table 3, the body surface condition score corresponding to the combination of physical appearance characteristics of user A (uniform color, slight edema, slight sweating) is 70-89 points. The median value of the interval can be taken as the final score. For example, the body surface condition score of user A is determined to be 80 points.

[0098] In practice, the scope of data collection for body surface condition features may also include the oral cavity and tongue region. For example, the tongue image can be captured with the user's cooperation using an image acquisition device of the target device (such as a smartphone camera or a wearable device with a front-facing camera). The tongue image should include the main area of ​​the tongue, and the lighting should be uniform and without obvious obstruction. Then, the tongue image is digitally processed to extract the tongue appearance parameters of the body surface appearance parameters, which are used as part of the body surface condition features.

[0099] Among them, the apparent parameters of the tongue include, but are not limited to: Tongue color index: Calculated based on the average value of the RGB channels in the region of interest on the tongue surface, used to quantify the changes in brightness and redness of the tongue color; Tongue coating color index: Based on the HSV or Lab color space features of the coating area, it distinguishes between light and dark coatings; Tongue coating coverage: The coating area is identified by a semantic segmentation model, and its proportion of the total tongue surface area is calculated. Tongue slenderness: Based on the geometric shape fitted to the edge contour of the tongue (such as ellipticity, area / perimeter ratio), it reflects whether there is local swelling or expansion trend; Crack / tooth mark detection: Using edge enhancement, morphological filtering and template matching algorithms, the system identifies whether there are linear cracks or periodic indentations on the tongue surface; Length of the under-tongue region: The length of the under-tongue region is estimated by locating the origin of the frenulum and the projection of the tip of the tongue at key points. Tongue color index: Extract the RGB mean or saturation index of the sublingual vein area to analyze the local color change trend.

[0100] For example, after taking a photo of user A's tongue, the system extracts that the tongue color is pale (low R value), the tongue coating is thin and white, the tongue body is slightly swollen and accompanied by slight teeth marks, which indicates that there is a certain deviation between the tongue shape and color.

[0101] Furthermore, the aforementioned tongue appearance parameters and other body surface appearance parameters (such as complexion, edema, sweating, etc.) can be input into a preset scoring model to generate a comprehensive body surface status score.

[0102] Assume that the pre-defined correspondence between physical appearance features and physical condition scores is as shown in Table 4.

[0103] Table 4

[0104] Thus, the combination of user A's features (pale tongue color, thin tongue coating, slightly enlarged tongue, slight teeth marks) combined with other facial features (even color, slight swelling, slight sweating) results in a body surface condition score of 70-89 points. The median of the range can be taken as the final score. For example, user A's body surface condition score can be determined to be 80 points.

[0105] In this way, the experience of "observation" in traditional Chinese medicine can be transformed into a quantifiable and repeatable image processing workflow. By fusing parameters such as facial features and tongue features in a multimodal manner, the accuracy of health management information prompts can be further enhanced.

[0106] In this way, by using the experience of traditional Chinese medicine's "inspection" to transform facial feature data, tongue feature data, and other data into quantifiable body surface status scores, the accuracy of health management information prompts can be further enhanced.

[0107] The tongue image analysis method described in this application is not intended for any purpose of traditional Chinese medicine diagnosis or disease identification. The extracted tongue appearance parameters are merely auxiliary reference indicators reflecting physiological states such as peripheral circulation, water metabolism, and autonomic nerve function. The generated "body surface status score" is a trend-based, non-diagnostic health indicator, suitable for personal daily health management.

[0108] In step 403, the weighting coefficients of the physical fitness score and the body surface status score in the fusion process are determined based on the basic health information.

[0109] In practice, different users have different sensitivities to different modal data. Therefore, the weighting coefficients of physical fitness scores and body surface status scores in the fusion process can be adjusted according to the user's basic health information. For example, the pulse waveform of people who exercise regularly is more stable, so the weight of physical fitness scores can be increased. People who are prone to edema are more sensitive to changes in their body surface, so the weight of body surface status scores can be increased.

[0110] For example, based on user A's basic health information, PPG waveform changes are highly correlated with fatigue level (correlation coefficient 0.8), and body surface edema is correlated with excessive water intake (correlation coefficient 0.6). Therefore, the weight of the physical fitness score can be set higher, such as 0.6, and the weight of the body surface condition score can be set slightly lower, such as 0.4.

[0111] In step 404, the physical fitness score is multiplied by the corresponding weighting coefficient to obtain the first weighted value; and the body surface condition score is multiplied by the corresponding weighting coefficient to obtain the second weighted value.

[0112] Taking user A's physical fitness score with a weighting factor of 0.6 and their physical appearance score with a weighting factor of 0.4 as an example, assuming user A's physical fitness score is 95 and their physical appearance score is 80, then the first weighting value is... The second weighted value is .

[0113] In step 405, the first weighted value and the second weighted value are summed to generate the second analysis result.

[0114] In practice, a comprehensive score can be generated by summing the first and second weighted values. The comprehensive score is used to characterize the overall trend of the user's current peripheral circulation and body surface status. Then, the second analysis result is determined based on the comprehensive score. For example, when the comprehensive score is above 90, the second analysis result is a good condition; when the comprehensive score is between 70 and 89, the second analysis result is a stable condition; and when the comprehensive score is below 70, the second analysis result is a poor condition.

[0115] For example, summing the first and second weighted values ​​for user A yields... Therefore, User A's second analysis result is in good condition.

[0116] In this way, by simultaneously analyzing the waveform morphological characteristics of PPG signals (such as the slope of the ascending branch, the delay time of the diphtheria wave, AIx, etc.) and the visual appearance characteristics of body surface images (such as skin color uniformity, facial edema, sweating status, pale tongue color, thin tongue coating, slight enlargement, slight teeth marks), multimodal data fusion from internal to external is achieved, overcoming the problem of assessment bias caused by relying on only a single physiological parameter, and improving the comprehensiveness and reliability of health status identification.

[0117] In practice, by long-term tracking and pattern recognition of the user's pulse dynamics parameters extracted from the physiological signals collected by the PPG sensor, and by combining information such as the user's body type and body surface condition, a personalized physiological state baseline model can be constructed.

[0118] Building upon this foundation, the system can dynamically identify subtle shifts or abnormal fluctuations in pulse dynamics parameters, providing users with early warnings of abnormal cardiovascular function. For example, detecting features such as a significantly prolonged pulse rise time, disappearance of the dicrotic wave, or a persistently shortened pulse tone (PTT) may indicate potential risks such as increased peripheral vascular resistance, progression of arteriosclerosis, or autonomic nervous system dysregulation. Furthermore, using pulse dynamics parameters as an important quantitative indicator for assessing cardiovascular health not only enables non-invasive, continuous, and home-based monitoring of diseases such as hypertension, arrhythmia, and heart failure, but also provides early warnings during sub-healthy stages, assisting users in timely lifestyle adjustments. This plays a positive role in chronic disease management, health promotion, and prevention of sudden cardiovascular events.

[0119] In step 204, based on the pre-trained multimodal feature fusion model, the first analysis result and the second analysis result are fused and analyzed to generate a comprehensive analysis result that characterizes the physiological state of the target object.

[0120] In practice, the multimodal feature fusion model can be trained in the following ways.

[0121] See Figure 5 , Figure 5 The flowchart of a training method for a multimodal feature fusion model provided in this application embodiment includes the following steps.

[0122] In step 501, multiple sample data are acquired, wherein each sample data includes first-type sample information, second-type sample information, basic health sample information, and corresponding physiological state labels.

[0123] Physiological state labels are: reference information generated based on expert annotations or observable physiological events.

[0124] In practice, each sample data includes: first-class sample information (such as PPG signal, ECG signal), second-class sample information (such as pulse waveform characteristics, body surface condition characteristics), basic health sample information (such as historical physical examination reports, routine health examination data, previous physical indicator records, etc.) and corresponding physiological status labels (generated based on expert annotation or observable physiological events). The labels can be discrete categories (such as "good", "stable", "needs attention") or continuous values ​​(such as a score of 0-100).

[0125] In step 502, for each sample data, the information of the first type of sample and the basic health sample information are analyzed to obtain the first sample analysis result; and the information of the second type of sample and the basic health sample information are analyzed to obtain the second sample analysis result.

[0126] In specific implementation, the process of analyzing the first type of sample information and basic health sample information to obtain the first sample analysis result can refer to the process of analyzing the first type of information and basic health information to obtain the first analysis result in steps 301 to 305. The process of analyzing the second type of sample information and basic health sample information to obtain the second sample analysis result can refer to the process of analyzing the second type of information and basic health information to obtain the second analysis result in steps 401 to 405. This application will not elaborate further here.

[0127] For example, for a sample A, after analyzing the information of the first type of sample and the basic health sample, the analysis result of the first sample is normal, with a heart rate of 72 bpm, blood oxygen saturation of 96%, heart rate fluctuation parameter (SDNN) of 48 ms, and vascular elasticity parameter (PWV) of 6.5 m / s. After analyzing the information of the second type of sample and the basic health sample, the physical fitness score is 85 points and the body surface condition score is 75 points. Based on multiplying by their respective weight coefficients, the comprehensive score is 81 points, and the analysis result of the second sample is determined to be a stable condition.

[0128] In step 503, the first sample analysis result and the second sample analysis result of each sample data are used as input, and the goal is to output the physiological state label corresponding to each sample data to train the multimodal feature fusion model.

[0129] In practice, the first and second sample analysis results of each sample data can be used as inputs, with the goal of outputting the physiological state label corresponding to each sample data, to train a multimodal feature fusion model.

[0130] Among them, the multimodal feature fusion model can be a variety of deep learning models, such as convolutional neural networks, long short-term memory networks, etc.

[0131] For example, the input features of sample A are [heart rate: 72 bpm, blood oxygen saturation: 96%, heart rate fluctuation parameter (SDNN): 48 ms, vascular elasticity parameter (PWV): 6.5 m / s, physical fitness score: 85 points, body surface condition score: 75 points], and the physiological state label of sample A is "stable".

[0132] In specific training, the model parameters can be initialized, then the input features can be fed into the model to obtain the predicted physiological state labels, the loss function between the predicted labels and the true labels can be calculated, and then the model parameters can be updated using the gradient descent method to gradually reduce the loss value. The above process is repeated until the model converges.

[0133] In practice, the comprehensive analysis results may only include status indicators such as good condition, stable condition, or condition requiring attention, or they may include specific analysis results of physiological signals, such as high heart rate or low blood oxygen saturation. They may also include specific analysis results of pulse waveform characteristic data and body surface condition characteristic data, such as obvious body surface edema or dark complexion. This application does not limit these aspects.

[0134] In step 205, based on the comprehensive analysis results, the wearable device provides health management information prompts to the target object.

[0135] In practice, a pre-set health management information database can be queried to determine the health management information prompts that match the comprehensive analysis results. Based on the health management information prompts, the wearable device can be controlled to send prompts to the target. Alternatively, a terminal device connected to the wearable device can also send prompts to the target.

[0136] Among them, the health management information database is a pre-built knowledge base that stores the mapping relationship between various comprehensive analysis result status indicators and corresponding non-medical health management information prompts. The prompts are general, trend-oriented, and lifestyle-oriented suggestions.

[0137] In practice, the content of the health management information database can be as shown in Table 5.

[0138] Table 5

[0139] In practice, the content of health management information prompts can be specifically set based on the user's basic health information or user profile (such as age, gender, exercise habits), and this application does not impose any restrictions on this.

[0140] In practice, after determining the health management information prompts that match the comprehensive analysis results, prompts can be sent to the target audience via wearable devices (such as smartwatches) or terminal devices connected to them (such as smartphones and tablets). These prompts can take various forms, such as vibration alerts, voice broadcasts, and text / graphical interface displays. For example, see... Figure 6 , Figure 6 This is a schematic diagram illustrating a health management information prompt provided in an embodiment of this application. A health prompt box pops up on a smartphone screen, displaying suggestions such as "Recent physiological fluctuations have been detected, with heart rate higher than usual and blood oxygen levels slightly lower. It is recommended to ensure sufficient sleep, increase outdoor activity time, and avoid high-pressure work for more than 2 consecutive hours." Two options are provided below the prompt box: "OK" and "View Details." Clicking "OK" closes the prompt box, while clicking "View Details" allows the user to view the corresponding test data.

[0141] Considering that physiological signals are affected by various factors in practical applications (such as emotions, environment, and measurement errors), relying solely on objective data may lead to false alarms or excessive alerts. For example, a user may have just completed strenuous exercise, at which point an elevated heart rate is normal, but the system may misjudge it as abnormal. A user may subjectively feel fatigued, even if their physiological parameters are still within the normal range. Therefore, before providing health management information alerts to the target individual based on comprehensive analysis results, it is also possible to receive subjective feedback on the user's physical state, achieving a "subjective-objective fusion" intelligent alert mechanism.

[0142] In practice, the subjective physical status feedback information input by the user can be received through the target device; the target device is a wearable device or a terminal device that communicates with and is connected to the wearable device; then, by combining the subjective physical status feedback information and the comprehensive analysis results, health management information prompts are given to the target object.

[0143] Subjective physical condition feedback information can include fatigue, sleep quality, emotional state, and the user's own health feelings, such as chest tightness, headache, dizziness, palpitations, weakness, shortness of breath, and decreased attention.

[0144] In practice, a lightweight interactive interface can be proactively popped up before providing health management information prompts to the target audience to inquire about their current feelings.

[0145] For example, a pop-up question interface on a smartwatch can ask whether the user is currently feeling tired, experiencing chest tightness or dizziness, feeling emotionally unwell, or how much exercise they did today. The user can then input subjective feedback about their physical condition via voice or by clicking on options.

[0146] For example, if the comprehensive analysis results show that the physiological parameters are normal, but the user's subjective physical state feedback information is fatigue, then based on the subjective physical state feedback information and the comprehensive analysis results, the output prompt could be: "A fluctuation in physiological state has been detected and you feel fatigued. It is recommended to stop high-intensity work and take 10 minutes of deep breathing or a walk to relax."

[0147] For example, if the comprehensive analysis results show that the physiological parameters are normal, but the user's subjective physical condition feedback information is chest tightness and palpitations, then based on the subjective physical condition feedback information and the comprehensive analysis results, the output prompt could be: "We have detected that you have reported chest discomfort. We suggest you rest and observe for the time being. If the discomfort persists, please seek medical attention promptly."

[0148] For example, if the comprehensive analysis results show that heart rate variability has decreased and the user reports anxiety, the output prompt could be: "Your autonomic nervous system regulation ability has decreased, which may be related to mood fluctuations. You can try deep breathing exercises."

[0149] In this way, by combining the experience of traditional Chinese medicine "diagnosis" with the user's subjective feedback on physical condition, the accuracy of health management information prompts can be further enhanced.

[0150] In practice, when a user inputs subjective feedback information about their physical state via voice, the user's voice signal can be collected through a wearable device, and voice features related to traditional Chinese medicine auscultation can be extracted based on acoustic feature analysis. These voice features include at least one of voice intensity, cough characteristics, and abnormal breath sounds. Additionally, volatile components of the user's body surface or exhaled air can be collected through a gas sensor integrated into the wearable device to generate odor feature data. Finally, the voice feature and / or odor feature data are input into a multimodal feature fusion model to enhance the diagnostic accuracy of the comprehensive analysis results.

[0151] For example, a user might voice input: "I've been sighing a lot lately, and my chest feels tight."

[0152] At this point, based on the user's voice information, acoustic feature analysis can be used to extract speech features, obtaining one or a combination of fundamental frequency parameters, sound intensity parameters, speech rate parameters, proportion of breath sounds, and wheezing frequency. The output prompt could be, "Frequent sighing and weakened autonomic nervous system regulation have been detected. Combined with chest tightness, there may be liver qi stagnation. It is recommended to perform 5 minutes of abdominal breathing and avoid emotional suppression."

[0153] For example, when a user inputs voice, a gas sensor collects volatile components from the user's skin or exhaled breath. If an increase in hydrogen sulfide concentration in the exhaled breath is detected, the output prompt could be: "Bad breath is detected and is related to digestive burden. It is recommended to reduce greasy food and take a 10-minute walk after meals to aid digestion."

[0154] In this way, by combining the experience of "auscultation" in traditional Chinese medicine with further analysis of users' voice characteristics and / or odor characteristics, the accuracy of health management information prompts can be further enhanced.

[0155] In this embodiment, given the user's known constitution type (such as Qi deficiency, Yang deficiency, phlegm-dampness, damp-heat, Qi stagnation, blood stasis, Yin deficiency, etc., from the TCM "palpation" dimension) and body surface condition information (such as tongue appearance, complexion, skin moisture, etc., from the TCM "inspection" dimension), the following multi-source information is further integrated: the user's subjective physical condition feedback information, including descriptions of symptoms such as fatigue, chest tightness, dizziness, and mood fluctuations, etc., from the TCM "inquiry" dimension; voice features extracted from speech signals (such as voice strength, cough characteristics, and wheezing patterns); and odor features collected through a miniature gas sensor (such as changes in volatile components in breath and body odor), etc., from the TCM "olfaction" dimension, thus constructing an intelligent health assessment mechanism that conforms to the holistic view and dialectical thinking of TCM. By performing multi-dimensional fusion analysis of four types of information—"inspection" (body surface condition), "olfaction" (voice and odor), "inquiry" (subjective feedback), and "palpation" (constitution type derived from pulse waveform)—the accuracy and personalization of health management information prompts are significantly improved.

[0156] Considering that the quality of PPG signals directly affects the accuracy of calculating key physiological parameters such as heart rate, blood oxygen, and vascular elasticity in wearable devices, it's important to understand that in actual use, factors such as user movement, loosening of the watch, and changes in ambient light can significantly reduce the amplitude of the AC component of the PPG signal, leading to signal distortion or acquisition failure. Failure to promptly identify and address such anomalies will result in poor accuracy of subsequent analysis results, leading to inaccurate or even misleading health management information. For example, when a watch is worn too loosely, poor contact between the light source and the skin weakens the AC component of the PPG signal, potentially causing the system to misinterpret it as "low blood oxygen," thus generating incorrect health management information.

[0157] Therefore, before providing health management information prompts based on comprehensive analysis results, the amplitude of the AC component of the PPG signal can be monitored in real time. The AC component amplitude directly reflects the signal strength and signal-to-noise ratio. When the AC component amplitude is detected to drop below a preset threshold within a preset time, the signal quality of the PPG signal is calculated, and the abnormal acquisition status level of the PPG signal is determined based on the calculated signal quality. Then, the signal recovery response strategy corresponding to the abnormal acquisition status level is executed.

[0158] For example, based on indicators such as AC / DC ratio, signal-to-noise ratio, and waveform stability, the quality score of the current PPG signal (e.g., 0-100 points) is calculated. Based on the score, the level of abnormality in the acquisition status is determined. For example, 0-30 points correspond to a high level of abnormality, which means that the signal is severely attenuated and is most likely due to detachment or severe looseness. 31-60 points correspond to a medium level of abnormality, which means that the signal is continuously decreasing and may be due to misalignment. 61-100 points correspond to a low level of abnormality, which means that there is slight fluctuation and may be due to transient interference.

[0159] In practice, the signal recovery response strategy corresponding to the abnormal level of the acquisition status is as follows: If the anomaly level is low, the wearable device will be controlled to re-acquire and verify the PPG signal.

[0160] When a user is in motion, the decrease in the AC component of the PPG signal may not be due to loose wear or signal acquisition issues, but rather artifacts caused by the motion itself. In this case, blindly increasing the light source or gain will not only fail to improve signal quality but may also introduce problems such as saturation and noise amplification. Therefore, if the anomaly level is medium, it is determined whether the target object has remained stationary for a first preset duration (e.g., 10 seconds). If so, it indicates that the user has entered a "stable state," and the wearable device is controlled to activate red light enhancement detection and determine whether the PPG signal returns to normal within a second preset duration (e.g., 20 seconds). If the PPG signal does not return to normal within the second preset duration, it is more likely due to a device wearing problem (e.g., a loose watch or obstruction) rather than motion interference. In this case, the wearable device is controlled to increase the light source intensity and receiving gain and determine whether the PPG signal returns to normal within a third preset duration. If the PPG signal does not return to normal within the third preset duration (e.g., 40 seconds), the target device is controlled to output a wearing status adjustment prompt.

[0161] If the anomaly level is high, the wearable device is controlled to increase the light source intensity and the receiving gain, and it is determined whether the PPG signal is detected to have returned to normal within the third preset time period; if the PPG signal does not return to normal within the third preset time period, the target device is controlled to output a wearing status adjustment prompt.

[0162] In practice, the execution process of the signal recovery response strategy corresponding to the above-mentioned anomaly levels can not only be based on the currently detected anomaly level, but can also be dynamically evolving. That is, during the execution of the response strategy corresponding to any anomaly level, if the signal quality is detected to meet the conditions for the next anomaly level, the system automatically switches to the response strategy corresponding to that next level until the signal is recovered or the highest-level strategy is completed. For example, the system can transition from the signal recovery response strategy corresponding to a low level to the signal recovery response strategy corresponding to a medium level, and from the signal recovery response strategy corresponding to a medium level to the signal recovery response strategy corresponding to a high level.

[0163] See Figure 7 , Figure 7 The present application provides a schematic diagram of an anomaly detection process, which includes three stages: a first stage, a second stage, and a third stage. The first stage is a signal recovery response strategy for low-level anomalies, the second stage is a signal recovery response strategy for medium-level anomalies, and the third stage is a signal recovery response strategy for high-level anomalies.

[0164] Take, for example, a user wearing a smartwatch during daily activities, where the watch gradually loosens, causing the PPG signal to continuously deteriorate.

[0165] First, during the initial detection, green light is used to detect the PPG signal. If the AC component amplitude drops by 60% within 5 seconds (the preset threshold is 30%), the PPG signal acquisition status is determined to be at a low level based on the calculated signal quality. In the first stage, the wearable device is controlled to re-acquire and verify the PPG signal. After re-acquisition, the re-acquired PPG signal and AC component are used to determine whether the conditions for the next abnormal level are met. If the PPG signal still has not recovered and the AC component continues to drop to 70%, the conditions for the next abnormal level are met, and the system automatically upgrades to a medium-level response and enters the second stage.

[0166] Then, in the second stage, when the user remains stationary for 10 seconds, the wearable device is controlled to activate red light enhancement detection. This is because infrared wavelengths have a greater skin penetration depth than green wavelengths, which can reduce the impact of motion artifacts. At the same time, the absorbance of skin pigment melanin at infrared wavelengths is lower than that at green wavelengths, and increasing red light detection can reduce the impact of differences in skin color among users. Afterwards, based on the re-acquired PPG signal and AC component, it is determined whether the conditions for the next anomaly level are met. If the PPG signal does not return to normal within 20 seconds and the AC component continues to drop to 80%, the conditions for the next anomaly level are determined to be met, and the device is automatically upgraded to a higher level response and enters the third stage.

[0167] Finally, in the third stage, the wearable device is controlled to enhance the light source intensity and the receiving gain. Finally, if the PPG signal does not return to normal within 40 seconds, the target device is controlled to output a wearing status adjustment prompt, such as "Please check if the watch has fallen off".

[0168] This avoids response gaps caused by rigid level settings. For example, if a low-level strategy fails, it does not need to wait for the next round of detection but can be upgraded immediately, significantly shortening the signal recovery time. This not only improves the success rate of signal recovery but also enhances the stability and intelligence of the system in real-world usage scenarios.

[0169] In practice, during the execution of the above signal recovery response strategy, if the PPG signal still fails to return to normal after the completion of the third stage (high-level anomaly), and continues to meet any of the following conditions: If the PPG signal quality score is lower than the preset score threshold and the duration exceeds the preset duration T1 (e.g., 2 minutes). The device did not detect user touch, raise-to-wake, voice wake-up, or other interactive behaviors for a duration exceeding T2 (e.g., 3 minutes). The accelerometer did not detect any significant limb activity (such as gait or gestures) for a duration exceeding T3 (e.g., 5 minutes). At this point, a combined state of "persistent lack of physiological signals and user unresponsiveness" can be identified, potentially indicating device detachment, severe wearing abnormalities, or a potential disability state where the user is unable to actively interact. The target device can be controlled to output emergency assistance prompts, such as a screen pop-up saying, "The device has not collected signals for an extended period and there is no operational feedback. Do you need to contact your emergency contact?" A voice prompt will also play. If the user does not respond within a preset countdown (e.g., 15 seconds), a reminder message containing location information will be automatically sent to the preset emergency contact, such as, "I may be currently unwell. I have authorized the device to send this message. My location is location A. Please confirm my safety." Furthermore, the user can interrupt this process at any time or disable the reminder afterward.

[0170] Furthermore, to avoid accidental triggering due to non-critical scenarios such as loose device wear, user meditation, or sleep, a joint risk assessment is conducted in conjunction with the Traditional Chinese Medicine (TCM) dimension features described in this application: In the "palpation" dimension: the waveform morphology analysis of the user's pulse waveform feature data is performed to extract the target waveform parameters, and the constitution category that matches the target waveform parameters is determined according to the mapping relationship between the preset waveform parameters and the constitution category (such as Qi deficiency constitution, Yang deficiency constitution, phlegm-dampness constitution, damp-heat constitution, Qi stagnation constitution, blood stasis constitution, Yin deficiency constitution, special constitution and peaceful constitution).

[0171] In the "inspection" dimension: image recognition is performed on the user's physical condition data to identify whether the user's tongue image shows a dark purple tongue body, tortuous sublingual veins, etc., and whether the face is pale, swollen, etc.

[0172] In the "Auscultation" dimension: whether the user's recent voice sounds low and weak, speech is intermittent, or abnormal breathing is detected.

[0173] In the "consultation" dimension: Based on the subjective physical state feedback information of the user's recent voice or clicked options, whether information such as palpitations, chest tightness, cold sweats, and extreme fatigue has been reported.

[0174] If two or more of the characteristics from the aforementioned Traditional Chinese Medicine (TCM) dimensions meet the warning criteria, the system raises the risk level, confirming that the current state is highly likely to be disability, fainting, or other dangerous conditions, rather than a normal device malfunction. At this time, the target device outputs an emergency assistance prompt, such as, "We have detected that you have been unresponsive for an extended period, and your health indicators suggest possible serious discomfort. Do you need to contact your emergency contact?" If the user does not respond within a preset countdown (e.g., 10 seconds), a reminder message containing location information is automatically sent to the preset emergency contact, for example: "I may be currently unwell. I have authorized the device to send this message. My location is location A. Please confirm my safety." Furthermore, the user can interrupt this process at any time or disable the reminder afterward.

[0175] In this way, providing a passive rescue channel when the signal is continuously abnormal and the user does not respond is beneficial to the safety of high-risk groups such as the elderly and patients with chronic diseases.

[0176] In this embodiment, a comprehensive, continuous, and non-invasive assessment of a user's health status is achieved through the fusion analysis of multi-source health information. Specifically, on the one hand, physiological signals from Western medicine indicators such as photoplethysmography (PPG) and electrocardiogram (ECG) are collected, and quantitative parameters such as heart rate, blood oxygen saturation, heart rhythm fluctuation parameters, vascular elasticity parameters, and pulse wave conduction velocity are extracted to reflect cardiovascular function and autonomic nervous system regulation. On the other hand, visual information is collected from the user's face, tongue, and other body surface areas using image recognition technology, and digitally representable parameters such as skin tone uniformity, facial edema, tongue color index, and tongue coating coverage are extracted. Simultaneously, waveform analysis of PPG or pressure sensor signals is used to extract "pulse-like" parameters such as the slope of the rising branch, the ratio of the main wave to the dicrotic wave, and rhythm stability, as well as subjective physical state feedback information input by the user, voice features extracted from speech signals, and odor features collected through a miniature gas sensor. This achieves engineered representation and objective collection in the dimensions of "observation, auscultation, inquiry, and palpation" in traditional Chinese medicine. All extracted features are transformed into standardized scoring indicators, and a weighted fusion model is used to generate a comprehensive health trend score, which is used to output non-diagnostic health management information prompts. Compared with related technologies that only use Western medicine quantitative indicators to generate user health management information, this application embodiment combines health characteristics based on traditional Chinese medicine concepts with Western medicine health characteristics for comprehensive analysis, breaking through the limitations of single-modality monitoring and improving the dimensional completeness and individual adaptability of health assessment.

[0177] Based on the same inventive concept, embodiments of this application provide a health management information prompting device, please refer to... Figure 8 The device includes: The acquisition module 801 is used to acquire a first type of information and a second type of information of the target object through a wearable device; wherein, the first type of information is used to characterize the physiological signals of the target object; and the second type of information is used to characterize the pulse waveform and body surface state of the target object. The acquisition module 802 is used to acquire the basic health information of the target object, wherein the basic health information is used to characterize the target object's historical physical health examination information; The analysis module 803 is used to analyze the first type of information and the basic health information to obtain a first analysis result; and to analyze the second type of information and the basic health information to obtain a second analysis result; The fusion module 804 is used to perform fusion analysis on the first analysis result and the second analysis result based on a pre-trained multimodal feature fusion model to generate a comprehensive analysis result characterizing the physiological state of the target object; The prompting module 805 is used to provide health management information prompts to the target object by the wearable device based on the comprehensive analysis results.

[0178] In some embodiments, the first type of information includes part or all of the photoplethysmography (PPG) signal and the electrocardiogram (ECG) signal; wherein the PPG signal and the ECG signal are physiological signals acquired by a wearable device from a first body part of the target object; The second type of information includes some or all of the pulse waveform feature data and body surface state feature data. The pulse waveform feature data is pulse dynamic parameters obtained by performing waveform morphology analysis on the PPG signal. The body surface state feature data is image data acquired by the image acquisition device of the target device from the second body part of the target object. The target device is the wearable device or a terminal device that is communicatively connected to the wearable device.

[0179] In some embodiments, the first type of information includes PPG signals and ECG signals; The analysis module 803 is specifically used for: Based on the PPG signal, the heart rate parameters and blood oxygen saturation parameters of the target object are determined; Based on the ECG signal, identify the heart rhythm fluctuation parameters of the target object; Based on the synchronization relationship between the PPG signal and the ECG signal, the pulse wave velocity (PWV) is calculated, and the vascular elasticity parameters of the target object are determined based on the PWV. Based on the basic health information, the threshold ranges for heart rate parameters, blood oxygen saturation parameters, heart rhythm fluctuation parameters, and vascular elasticity parameters corresponding to the target object are determined respectively. After comparing the heart rate parameter, blood oxygen saturation parameter, heart rhythm fluctuation parameter, and vascular elasticity parameter with their respective threshold ranges, the first analysis result is generated based on the comparison results; the first analysis result is a status indicator representing the degree of deviation of physiological signals.

[0180] In some embodiments, the second type of information includes pulse waveform feature data and body surface state feature data; The analysis module 803 is specifically used for: The pulse waveform feature data is analyzed for waveform morphology, target waveform parameters are extracted, and the body type matching the target waveform parameters is determined and the body type score is determined based on the mapping relationship between the preset waveform parameters and the body type. Image recognition is performed on the surface condition feature data to extract the target surface appearance parameters, and the surface condition score matching the target surface appearance parameters is determined according to the mapping relationship between the preset surface appearance parameters and the surface condition score. Based on the basic health information, determine the weighting coefficients of the physical fitness score and the body surface condition score in the fusion process; The physical fitness score is multiplied by the corresponding weighting coefficient to obtain a first weighted value; and the body surface condition score is multiplied by the corresponding weighting coefficient to obtain a second weighted value. The first weighted value and the second weighted value are summed to generate the second analysis result.

[0181] In some embodiments, the fusion module 804 trains the multimodal feature fusion model in the following manner: Multiple sample data are acquired, wherein each sample data includes first-type sample information, second-type sample information, basic health sample information, and corresponding physiological state labels, wherein the physiological state labels are reference information generated based on expert annotation or observable physiological events. For each sample data, the information of the first type of sample and the basic health sample information are analyzed to obtain the first sample analysis result; and the information of the second type of sample and the basic health sample information are analyzed to obtain the second sample analysis result. The multimodal feature fusion model is trained by taking the first and second sample analysis results of each sample data as input and outputting the physiological state label corresponding to each sample data as the goal.

[0182] In some embodiments, the prompting module 805 is specifically used for: By querying a pre-set health management information database, the health management information prompts that match the comprehensive analysis results are determined; Based on the health management information prompts, the wearable device is controlled to provide prompts to the target object.

[0183] In some embodiments, it also includes: The receiving module 806 is used to receive subjective physical state feedback information input by the user through the target device; the target device is a wearable device or a terminal device that is communicatively connected to the wearable device. The prompting module 805 is specifically used to provide health management information prompts to the target object by combining the subjective physical state feedback information and the comprehensive analysis results.

[0184] In some embodiments, it also includes: Detection module 807 is used to monitor the amplitude of the AC component of the PPG signal in real time; When the amplitude of the AC component is detected to drop more than a preset threshold within a preset time, the signal quality of the PPG signal is calculated, and the abnormal level of the acquisition status of the PPG signal is determined based on the calculated signal quality. Execute the signal recovery response strategy corresponding to the abnormal level of the acquisition status.

[0185] In some embodiments, the detection module 807 is specifically used for: If the anomaly level is low, the wearable device is controlled to re-acquire and verify the PPG signal; If the anomaly level is medium, it is determined whether the target object has remained stationary for a first preset duration. If so, the wearable device is controlled to activate red light enhancement detection and it is determined whether the PPG signal has returned to normal within a second preset duration. If the PPG signal has not returned to normal within the second preset duration, the wearable device is controlled to enhance the light source intensity and receiving gain and it is determined whether the PPG signal has returned to normal within a third preset duration. If the PPG signal has not returned to normal within the third preset duration, the target device is controlled to output a wearing status adjustment prompt. If the anomaly level is high, the wearable device is controlled to increase the light source intensity and the receiving gain, and it is determined whether the PPG signal is detected to have returned to normal within a third preset time period; if the PPG signal does not return to normal within the third preset time period, the target device is controlled to output a wearing status adjustment prompt.

[0186] Based on the same inventive concept, this application provides an electronic device that can realize the function of the health management information prompting device described above. Please refer to... Figure 9 The device includes a memory 901, one or more processors 902, and a bus 903.

[0187] The memory 901 is used to store computer programs executed by the processor 901. The memory 901 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0188] Memory 901 may be volatile memory, such as random-access memory (RAM); memory 901 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 901 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 901 may be a combination of the above-mentioned memories.

[0189] The processor 902 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 902 is used to implement the health management information prompting method described in the above embodiments when it invokes a computer program stored in the memory 902.

[0190] This application embodiment does not limit the specific connection medium between the memory 901 and the processor 902 described above. This application embodiment... Figure 9 The memory 901 and the processor 902 are connected via a bus 903, and the bus 903 is in Figure 9 The connections between other components are shown in bold lines only and are not intended to be limiting. The 903 bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0191] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium. The computer program product includes computer program code, which, when executed on a computer, causes the computer to perform a method for prompting any of the health management information described above. Since the principle by which the above-described computer-readable storage medium solves the problem is similar to the method for prompting health management information, the implementation of the above-described computer-readable storage medium can be found in the implementation of the method; repeated details will not be elaborated further.

[0192] Based on the same inventive concept, this application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, it causes the computer to execute any of the health management information prompting methods described above. Since the principle by which the above-described computer program product solves the problem is similar to that of the health management information prompting method, the implementation of the above-described computer program product can refer to the implementation of the method, and repeated details will not be described again.

[0193] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0194] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0195] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0196] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of user-operated steps to be executed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0197] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for providing health management information prompts based on wearable devices, characterized in that, The method includes: The first type of information and the second type of information of the target object are collected through wearable devices; wherein, the first type of information is used to characterize the physiological signals of the target object; and the second type of information is used to characterize the pulse waveform and body surface state of the target object. Acquire basic health information of the target object, wherein the basic health information is used to characterize the target object's historical physical health examination information; The first type of information and the basic health information are analyzed to obtain a first analysis result; and the second type of information and the basic health information are analyzed to obtain a second analysis result. Based on a pre-trained multimodal feature fusion model, the first analysis result and the second analysis result are fused and analyzed to generate a comprehensive analysis result characterizing the physiological state of the target object; Based on the comprehensive analysis results, the wearable device provides health management information prompts to the target object.

2. The method as described in claim 1, characterized in that, The first type of information includes part or all of the photoplethysmography (PPG) signal and the electrocardiogram (ECG) signal; wherein the PPG signal and the ECG signal are physiological signals obtained by the wearable device from a first body part of the target object; The second type of information includes some or all of the pulse waveform feature data and body surface state feature data. The pulse waveform feature data is pulse dynamic parameters obtained by performing waveform morphology analysis on the PPG signal. The body surface state feature data is image data acquired by the image acquisition device of the target device from the second body part of the target object. The target device is the wearable device or a terminal device that is communicatively connected to the wearable device.

3. The method as described in claim 2, characterized in that, The first type of information includes PPG signals and ECG signals; The analysis of the first type of information and the basic health information to obtain a first analysis result includes: Based on the PPG signal, the heart rate parameters and blood oxygen saturation parameters of the target object are determined; Based on the ECG signal, identify the heart rhythm fluctuation parameters of the target object; Based on the synchronization relationship between the PPG signal and the ECG signal, the pulse wave velocity (PWV) is calculated, and the vascular elasticity parameters of the target object are determined based on the PWV. Based on the basic health information, the threshold ranges for heart rate parameters, blood oxygen saturation parameters, heart rhythm fluctuation parameters, and vascular elasticity parameters corresponding to the target object are determined respectively. After comparing the heart rate parameter, blood oxygen saturation parameter, heart rhythm fluctuation parameter, and vascular elasticity parameter with their respective threshold ranges, the first analysis result is generated based on the comparison results; the first analysis result is a status indicator representing the degree of deviation of physiological signals.

4. The method as described in claim 2, characterized in that, The second type of information includes pulse waveform feature data and body surface condition feature data; The analysis of the second type of information and the basic health information to obtain a second analysis result includes: The pulse waveform feature data is analyzed for waveform morphology, target waveform parameters are extracted, and the body type matching the target waveform parameters is determined and the body type score is determined based on the mapping relationship between the preset waveform parameters and the body type. Image recognition is performed on the surface condition feature data to extract the target surface appearance parameters, and the surface condition score matching the target surface appearance parameters is determined according to the mapping relationship between the preset surface appearance parameters and the surface condition score. Based on the basic health information, determine the weighting coefficients of the physical fitness score and the body surface condition score in the fusion process; The physical fitness score is multiplied by the corresponding weighting coefficient to obtain a first weighted value; and the body surface condition score is multiplied by the corresponding weighting coefficient to obtain a second weighted value. The first weighted value and the second weighted value are summed to generate the second analysis result.

5. The method as described in claim 1, characterized in that, The multimodal feature fusion model is trained using the following method: Multiple sample data are acquired, wherein each sample data includes first-type sample information, second-type sample information, basic health sample information, and corresponding physiological state labels, wherein the physiological state labels are reference information generated based on expert annotation or observable physiological events. For each sample data, the information of the first type of sample and the basic health sample information are analyzed to obtain the first sample analysis result; and the information of the second type of sample and the basic health sample information are analyzed to obtain the second sample analysis result. The multimodal feature fusion model is trained by taking the first and second sample analysis results of each sample data as input and outputting the physiological state label corresponding to each sample data as the goal.

6. The method as described in claim 1, characterized in that, The step of providing health management information prompts to the target object by the wearable device based on the comprehensive analysis results includes: By querying a pre-set health management information database, the health management information prompts that match the comprehensive analysis results are determined; Based on the health management information prompts, the wearable device is controlled to provide prompts to the target object.

7. The method as described in claim 1, characterized in that, Before providing health management information prompts to the target object based on the comprehensive analysis results, the process further includes: The system receives subjective physical state feedback information input by the user through a target device; the target device is a wearable device or a terminal device that is communicatively connected to the wearable device. The step of providing health management information prompts to the target object based on the comprehensive analysis results includes: Based on the subjective physical state feedback information and the comprehensive analysis results, health management information prompts are provided to the target object.

8. The method as described in claim 2, characterized in that, The method further includes: Real-time monitoring of the AC component amplitude of the PPG signal; When the amplitude of the AC component is detected to drop more than a preset threshold within a preset time, the signal quality of the PPG signal is calculated, and the abnormal level of the acquisition status of the PPG signal is determined based on the calculated signal quality. Execute the signal recovery response strategy corresponding to the abnormal level of the acquisition status.

9. The method as described in claim 8, characterized in that, The execution of the signal recovery response strategy corresponding to the abnormal level of the acquisition status includes: If the anomaly level is low, the wearable device is controlled to re-acquire and verify the PPG signal; If the anomaly level is medium, it is determined whether the target object has remained stationary for a first preset duration. If so, the wearable device is controlled to activate red light enhancement detection and it is determined whether the PPG signal has returned to normal within a second preset duration. If the PPG signal has not returned to normal within the second preset duration, the wearable device is controlled to enhance the light source intensity and receiving gain and it is determined whether the PPG signal has returned to normal within a third preset duration. If the PPG signal has not returned to normal within the third preset duration, the target device is controlled to output a wearing status adjustment prompt. If the anomaly level is high, the wearable device is controlled to increase the light source intensity and the receiving gain, and it is determined whether the PPG signal is detected to have returned to normal within a third preset time period; if the PPG signal does not return to normal within the third preset time period, the target device is controlled to output a wearing status adjustment prompt.

10. A health management information prompting device based on a wearable device, characterized in that, The device includes: The acquisition module is used to acquire a first type of information and a second type of information of the target object through a wearable device; wherein, the first type of information is used to characterize the physiological signals of the target object; and the second type of information is used to characterize the pulse waveform and body surface state of the target object. The acquisition module is used to acquire the basic health information of the target object, wherein the basic health information is used to characterize the target object's historical physical health examination information; An analysis module is used to analyze the first type of information and the basic health information to obtain a first analysis result; and to analyze the second type of information and the basic health information to obtain a second analysis result; The fusion module is used to perform fusion analysis on the first analysis result and the second analysis result based on a pre-trained multimodal feature fusion model to generate a comprehensive analysis result characterizing the physiological state of the target object; The notification module is used to provide health management information notifications to the target object by the wearable device based on the comprehensive analysis results.

Images