Health monitoring and early warning method based on intelligent wearable device

Abstract

The invention discloses a health monitoring and early warning method based on intelligent wearable equipment, and particularly relates to the technical field of health monitoring. The method comprises the following steps: acquiring breathing depth, breathing frequency, electroencephalogram signals and heart rate variability data of a user, and preprocessing; constructing a respiration-viscera-nerve regulation model based on a six-character table respiration method, wherein the respiration-viscera-nerve regulation model is used for evaluating the autonomic nervous system state of the user; performing fusion analysis by using a random forest algorithm to generate a personalized health intervention strategy; establishing a health risk assessment model of the user in combination with the autonomic nervous system state of the user and the personalized health intervention strategy, and determining an early warning trigger condition; when the pressure index of the user is detected to exceed the preset threshold value, the Baihui acupoint, the Shenting acupoint and the temples are accurately positioned, and the wearable device is controlled to implement micro-current stimulation and red and blue light phototherapy of the corresponding acupoints, so that real-time evaluation and personalized early warning intervention of the health state of the user are realized, and the initiative and intelligent level of health management are improved.

Description

Technical Field

[0001] This invention relates to the field of health monitoring technology, and more specifically, to a health monitoring and early warning method based on smart wearable devices. Background Technology

[0002] Currently, most health monitoring smart wearable devices on the market focus on the collection of single physiological parameters and the display of static data. They lack the ability to fuse and analyze multidimensional physiological signals in real time and are difficult to effectively integrate traditional Chinese medicine regulation theories to build personalized intervention mechanisms. Existing devices generally use static atlases or fixed templates for acupoint positioning, which is difficult to adapt to individual differences in the head structure of different users. This results in insufficient intervention accuracy, an imperfect real-time response mechanism, and an inability to provide dynamic and personalized early warning and intervention plans when users are at critical health risk levels. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a health monitoring and early warning method based on smart wearable devices to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A health monitoring and early warning method based on smart wearable devices includes the following steps:

[0006] S1: Real-time acquisition of the user's respiratory depth, respiratory rate, EEG signals and heart rate variability data through a high-precision respiratory sensor, EEG alpha wave sensor and heart rate variability monitoring module, and preprocessing of such data;

[0007] S2: Construct a breathing-organ-nerve regulation model based on the Six Healing Sounds breathing method to assess the user's autonomic nervous system status;

[0008] S3: The random forest algorithm is used to perform multi-source fusion analysis on respiratory depth, respiratory rate, EEG signals and heart rate variability data to generate personalized health intervention strategies;

[0009] S4: Based on the user's autonomic nervous system status and personalized health intervention strategies, establish a health risk assessment model for the user and determine the early warning trigger conditions;

[0010] S5: Based on the warning trigger conditions, when the health risk assessment model detects that the user's stress index exceeds the preset threshold, it accurately locates acupoints through 3D head scanning technology and controls the wearable device to implement acupoint microcurrent stimulation and acupoint phototherapy.

[0011] In a preferred embodiment, S1 specifically refers to:

[0012] The device uses a high-precision respiratory sensor built into a smart wearable device to collect the user's breathing depth and respiratory rate in real time.

[0013] Real-time acquisition of users' brainwave alpha wave characteristic data via EEG alpha wave sensor;

[0014] The heart rate variability monitoring module collects the standard deviation of the interval between heartbeats during the user's normal sinus heartbeat cycle in real time.

[0015] The collected respiratory depth, respiratory rate, EEG alpha wave characteristic data, and standard deviation of heartbeat interval were denoised and filtered, respectively.

[0016] In a preferred embodiment, S2 specifically refers to:

[0017] By using the six-character breathing method to correspond to the user's breathing depth and breathing frequency data characteristics, a correlation rule is established between breathing depth, breathing frequency and specific organ functions.

[0018] Based on the association rules, the respiratory depth and respiratory rate characteristics of different breathing techniques are fused with EEG alpha wave characteristic data and the standard deviation of the heartbeat interval.

[0019] The fused data is matched with a pre-established respiratory-visceral-nervous regulation model to assess the user's current autonomic nervous system functional state and output the user's autonomic nervous system state assessment results.

[0020] In a preferred embodiment, S3 specifically refers to:

[0021] The random forest algorithm was used to perform data fusion analysis on the user's respiratory depth, respiratory rate, electroencephalogram (EEG) signal and heart rate variability data, and a training sample set for the random forest algorithm was constructed.

[0022] The random forest algorithm is used to perform feature importance analysis on the training sample set to determine the combination of feature parameters closely related to the user's health status;

[0023] Based on the combination of feature parameters, a mapping relationship library between a user's personal health status and intervention measures is established, and personalized health intervention strategies suitable for a user's personal health status are generated.

[0024] In a preferred embodiment, S4 specifically refers to:

[0025] Map the specific organ function state parameters corresponding to the user's autonomic nervous system state assessment results with the feature parameter combination in the personalized health intervention strategy to establish a matching rule between the user's autonomic nervous system state and the health intervention strategy.

[0026] Based on the matching rules between the user's autonomic nervous system state and health intervention strategies, the user's health risk level is classified and determined.

[0027] Based on the user's health risk level, determine the warning trigger conditions for the user's health risk.

[0028] In a preferred embodiment, S5 specifically refers to:

[0029] When the user's health risk assessment model determines that the user's stress index exceeds the pre-set health risk level threshold, the 3D head shape scanning technology is activated to perform a three-dimensional spatial scan of the user's head.

[0030] Based on the three-dimensional spatial information of the user's head shape obtained from the scan, the coordinates of the Baihui, Shenting, and Taiyang acupoints on the user's head are determined.

[0031] Based on the location coordinates, the microcurrent stimulation module and acupoint phototherapy module on the smart wearable device are respectively positioned at Baihui, Shenting and Taiyang acupoints;

[0032] Based on personalized health intervention strategies, microcurrent stimulation signals are output to Baihui, Shenting and Taiyang acupoints, while red or blue light waves corresponding to specific intervention strategies are irradiated.

[0033] The technical effects and advantages of the health monitoring and early warning method based on smart wearable devices of the present invention are as follows:

[0034] By integrating a high-precision respiratory sensor, an EEG alpha wave sensor, and a heart rate variability monitoring module, real-time acquisition and preprocessing of user breathing, EEG, and heart rate signals were achieved, ensuring the comprehensiveness and timeliness of data collection. A breathing-organ-nerve regulation model was constructed using the Six Healing Sounds breathing method, quantifying traditional Chinese medicine breathing theory and embedding it into the neural regulation analysis process, improving the accuracy of autonomic nervous system state assessment. A random forest algorithm was introduced to fuse multi-source data, enhancing the accuracy and adaptability of personalized intervention strategies. A health risk assessment model and dynamic early warning conditions were constructed to achieve graded identification and intervention decisions for user health status. Three-dimensional head scanning was used to dynamically locate acupoints and precisely control the coordinated execution of microcurrent stimulation and multi-band light illumination, significantly improving the targeting and intelligence of intervention responses. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of a health monitoring and early warning method based on a smart wearable device according to the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0037] Example

[0038] Figure 1 This invention discloses a health monitoring and early warning method based on a smart wearable device, which includes the following steps:

[0039] S1: Real-time acquisition of the user's respiratory depth, respiratory rate, EEG signals and heart rate variability data through a high-precision respiratory sensor, EEG alpha wave sensor and heart rate variability monitoring module, and preprocessing of such data;

[0040] S2: Construct a breathing-organ-nerve regulation model based on the Six Healing Sounds breathing method to assess the user's autonomic nervous system status;

[0041] S3: The random forest algorithm is used to perform multi-source fusion analysis on respiratory depth, respiratory rate, EEG signals and heart rate variability data to generate personalized health intervention strategies;

[0042] S4: Based on the user's autonomic nervous system status and personalized health intervention strategies, establish a health risk assessment model for the user and determine the early warning trigger conditions;

[0043] S5: Based on the warning trigger conditions, when the health risk assessment model detects that the user's stress index exceeds the preset threshold, it accurately locates acupoints through 3D head scanning technology and controls the wearable device to implement acupoint microcurrent stimulation and acupoint phototherapy.

[0044] S1: Real-time acquisition of the user's respiratory depth, respiratory rate, EEG signals, and heart rate variability data via a high-precision respiratory sensor, EEG alpha wave sensor, and heart rate variability monitoring module, followed by preprocessing, including:

[0045] The device uses a high-precision respiratory sensor built into a smart wearable device to collect the user's breathing depth and respiratory rate in real time.

[0046] Real-time acquisition of users' brainwave alpha wave characteristic data via EEG alpha wave sensor;

[0047] The heart rate variability monitoring module collects the standard deviation of the interval between heartbeats during the user's normal sinus heartbeat cycle in real time.

[0048] The collected respiratory depth, respiratory rate, EEG alpha wave characteristic data, and standard deviation of heartbeat interval were denoised and filtered, respectively.

[0049] High-precision respiratory sensors built into smart wearable devices refer to miniature pressure sensor arrays integrated into areas that conform to the skin, such as headbands, to detect the deformation of the chest or abdomen during respiration. Respiratory depth is defined as the displacement of the chest or abdomen during each respiratory cycle, and its physical quantity can be expressed as volume change or displacement change; for example, a chest expansion of 50 ml corresponds to a respiratory depth of 50 ml. Respiratory rate is defined as the number of complete respiratory cycles per unit time; for example, 12 complete inhalation-exhalation cycles per minute correspond to a respiratory rate of 12 breaths / min.

[0050] The respiratory sensor continuously reads pressure change signals at a 50 Hz sampling rate and converts the analog signals into digital signals via a 12-bit analog-to-digital converter. Real-time acquisition requires the total latency of sampling, processing, buffering, and transmission to be less than 100 milliseconds to ensure that changes in respiratory depth can reflect the current respiratory state in real time. For example, when a user is at rest, the average chest expansion is 400 ml, corresponding to a 0.2 volt change in the pressure sensor output voltage; the respiratory rate is 16 breaths / min. The sensor samples 50 times per second to accurately capture the two inflection points of expansion and contraction, thereby calculating the frequency value of 16 breaths / min.

[0051] High-precision respiratory sensors must have a volume measurement accuracy of ±5 ml and a frequency measurement accuracy of ±0.2 breaths / min during data acquisition. The respiratory depth and respiratory rate data output by the sensor serve as the raw input for signal preprocessing and multi-source fusion analysis, ensuring data consistency across steps.

[0052] An EEG alpha wave sensor refers to a metal electrode array integrated into the inner surface of a smart headband, used to detect potential changes on the user's scalp. The alpha wave frequency band is defined as brain electrical activity in the range of 8–13 Hz, and characteristic data include, but are not limited to, indicators such as signal amplitude, spectral power, and amplitude envelope.

[0053] The EEG electrodes are positioned according to the international 10-20 system, specifically O1, O2, P3, and P4. A 50-ohm impedance matching circuit sends microvolt-level electrical signals to a front-end amplifier, amplifying them by 10,000 times. The amplified signal is then acquired by a 16-bit analog-to-digital converter at a sampling rate of 250 Hz, and the alpha wave power spectral density is calculated in real time by a digital signal processing unit. For example, when the user is relaxed with their eyes closed, the average alpha wave power in the O1 channel is approximately 20 microwatts per square centimeter, with an amplitude envelope of approximately 50 microvolts; when the user concentrates, this power decreases to 10 microwatts per square centimeter.

[0054] Real-time acquisition ensures data latency of less than 200 milliseconds, facilitating subsequent fusion with respiratory and heart rate variability data. The power spectral density curve and amplitude envelope curve output from the EEG alpha wave characteristic data are directly used as input for multi-source fusion analysis.

[0055] Heart rate variability monitoring modules typically employ photoplethysmography (PPG) or patch-type ECG electrodes to detect the intervals between consecutive heartbeats in a normal sinus rhythm. The interval is defined as the time interval between two consecutive R-peaks, measured in milliseconds; the standard deviation (SDNN) reflects the level of heart rate variability.

[0056] Real-time acquisition requires the module to track ECG or pulse wave signals at a sampling rate of at least 250 Hz and calculate the RR interval sequence within a sliding window (e.g., within 5 minutes), then dynamically update the standard deviation value. For example, in a resting state, the average interval of 300 consecutive heartbeat intervals is 800 milliseconds. If the calculated standard deviation of this sequence is 50 milliseconds, then the SDNN will have a standard deviation of 50 milliseconds. In a state of exercise, the standard deviation of the RR interval may increase to 100 milliseconds.

[0057] The standard deviation value output by the heart rate variability monitoring module serves as a quantitative indicator of autonomic nervous activity and is synchronized with respiratory depth, respiratory rate, and EEG alpha wave characteristic data over time.

[0058] Denoising and filtering of respiratory depth and respiratory frequency signals: The respiratory components of 0.1 to 1 Hz were extracted using a bandpass filter and a fourth-order Butterworth digital filter was used. To remove motion artifacts and environmental vibration noise, a wavelet thresholding denoising algorithm was further applied. After the third-level Daubechies wavelet decomposition, the high-frequency detail coefficients were truncated by thresholding, and finally the denoised signal was reconstructed.

[0059] Denoising and filtering of EEG alpha wave feature data: First, a 50 Hz notch filter is applied to the raw EEG signal to remove power frequency interference; then, an 8-13 Hz bandpass filter is applied to extract the alpha wave frequency band; finally, an adaptive noise cancellation algorithm is used to remove electromyography and electrooculography artifacts using reference channel signals (such as earlobe or posterior neck electrodes) to output a clean alpha wave power spectral density curve.

[0060] Numerical denoising and filtering of standard deviation of heart interval: For continuous RR interval sequences, first apply a medium filter to remove outliers (the threshold can be set to the average RR ± 20%), and then use a first-order low-pass filter to smooth the standard deviation curve and filter out spike interference caused by sudden arrhythmias or measurement noise.

[0061] By denoising and filtering each data source, the respiratory depth curve and respiratory frequency time series diagram output by the high-precision respiratory sensor, the power spectral density curve output by the EEG alpha wave sensor, and the standard deviation curve output by the heart rate variability monitoring module can all achieve a signal-to-noise ratio of over 20 dB, meeting the signal quality requirements of subsequent multi-source fusion analysis steps.

[0062] S2: Based on the Six Healing Sounds breathing method, a breathing-internal organ-nerve regulation model is constructed to assess the user's autonomic nervous system status, including:

[0063] By using the six-character breathing method to correspond to the user's breathing depth and breathing frequency data characteristics, a correlation rule is established between breathing depth, breathing frequency and specific organ functions.

[0064] The Six-Syllable Breathing Method originates from Traditional Chinese Medicine's theory of breath regulation. It combines six vocalizations—Xu, He, Hu, Xi, Chui, and Xi—with breathing rhythms. Each syllable corresponds to a different breathing pattern and the regulatory needs of internal organs. Breathing depth refers to the extent of expansion and contraction of the chest or abdomen during one inhalation-exhalation cycle, as captured by a smart wearable device, expressed in milliliters or millimeters. Breathing frequency refers to the number of complete inhalation-exhalation cycles per unit time, expressed in breaths per minute.

[0065] The breathing depth and frequency range corresponding to each sound in the Six Healing Sounds were defined. For example, the breathing depth range for the "Xu" sound was set to 300–500 ml, and the corresponding breathing frequency range was set to 8–10 breaths / minute; the breathing depth range for the "He" sound was set to 200–300 ml, and the corresponding breathing frequency range was set to 10–12 breaths / minute; the breathing depth range for the "Chui" sound was set to 500–700 ml, and the corresponding breathing frequency range was set to 6–8 breaths / minute. The correspondence between each sound and the functions of the internal organs was determined based on classical Chinese medicine literature and modern experimental data. For example, the deep, low-frequency breathing guided by the "Xu" sound can soothe the liver and regulate qi; the moderate breathing guided by the "He" sound can harmonize the stomach and regulate the middle jiao; and the deep, slow breathing guided by the "Chui" sound can tonify the kidneys and strengthen essence.

[0066] When establishing association rules, the respiratory depth and respiratory rate ranges corresponding to each type of breathing technique are mapped one-to-one with the organ function category, forming a rule set. Example rule set format: "When respiratory depth is 300–500 ml and respiratory rate is 8–10 breaths / minute, associate with liver function regulation"; "When respiratory depth is 200–300 ml and respiratory rate is 10–12 breaths / minute, associate with spleen and stomach function regulation." When executing association rules, the actual collected respiratory depth and respiratory rate data are used as input. An interval matching algorithm determines the current breathing technique type and, based on this, the corresponding organ function category is determined.

[0067] For example, assuming the smart wearable device collects a respiratory depth of 320 ml and a respiratory rate of 9 breaths / minute during a certain respiratory cycle, the rule for matching the "shh" sound will generate an organ function category identifier of "liver function regulation". The association rule base retains a list of parameters related to liver function, such as indicators corresponding to decreased heart rate variability and increased EEG alpha wave power associated with liver qi stagnation, for use in subsequent analysis.

[0068] Based on the association rules, the respiratory depth and respiratory rate characteristics of different breathing techniques are fused with EEG alpha wave characteristic data and the standard deviation of the heartbeat interval.

[0069] After completing the association rule matching, it is necessary to aggregate the respiratory depth and respiratory rate data belonging to different organ function categories, along with the alpha wave power spectral density or amplitude envelope data obtained from the EEG sensor, and the standard deviation of the heart rate interval calculated from the heart rate variability module. This aggregation is achieved using a multi-dimensional data fusion framework, aligning the three types of data into feature vectors within the same time window. The time window length is typically a 5-second or 10-second sliding window, chosen based on physiological principles and real-time requirements.

[0070] Taking a 10-second sliding window as an example, the average respiratory depth, average respiratory rate, average alpha wave power, and standard deviation of the heartbeat interval are calculated within each time window. Assume that within a certain 10-second window, the average respiratory depth is 350 ml, the average respiratory rate is 9 breaths / min, the average alpha wave power is 18 microwatts / cm², and the standard deviation of the heartbeat interval is 45 milliseconds. These four scalar values ​​are organized into a four-dimensional vector [350, 9, 18, 45].

[0071] The fusion process employs a rule-based weighted fusion method: each indicator is assigned a weight to reflect its importance in the assessment of the autonomic nervous system. For example, EEG alpha wave power is assigned a weight of 0.3, respiratory depth a weight of 0.25, respiratory rate a weight of 0.2, and the standard deviation of heart rate intervals a weight of 0.25. The fused comprehensive feature values ​​are obtained by weighting the above example vectors.

[0072] Integration score = 0.25×350 + 0.2×9 + 0.3×18 + 0.25×45 = 87.5 + 1.8 + 5.4 + 11.25 = 105.95.

[0073] The fused data is matched with a pre-established respiratory-visceral-nervous regulation model to assess the user's current autonomic nervous system functional state and output the user's autonomic nervous system state assessment results.

[0074] The respiratory-visceral-neural regulation model is a mapping model constructed based on the theory of viscera and meridians in traditional Chinese medicine and the findings of modern neurophysiological research. It employs a combination of rule base and threshold judgment. The model includes setting the correspondence between the functional states of the five viscera (liver, heart, spleen, lungs, and kidneys) and the balance indicators of the autonomic nervous system, as well as setting judgment thresholds for the fusion score range. For example, if the fusion score falls within the range of 80-100 and the association rule points to liver function regulation, the assessment conclusion is "liver stagnation and irritability, slight sympathetic nerve excitation"; if the fusion score falls within the range of 100-120 and the association rule points to liver function, the assessment conclusion is "relatively balanced liver qi, moderate activation of the parasympathetic nervous system".

[0075] The matching process is as follows: taking the fusion score of 105.95 in the example and the organ category "liver function regulation" as input, the corresponding functional state entry is located in the model library, the numerical range of 105.95 is determined, and the corresponding functional state description is output. The output includes: classification of the functional state of the autonomic nervous system (such as sympathetic nerve excitation, parasympathetic nerve dominance); and a brief description of the function of the associated organ (such as "normal liver qi circulation" or "slightly insufficient liver qi").

[0076] The assessment results are output in a structured data format, including functional state type codes (e.g., 0 for sympathetic dominance, 1 for balance, 2 for parasympathetic dominance), organ codes (e.g., 1 for liver, 2 for heart), and corresponding state text descriptions. An example output would be: {Functional state code: 1, Organ code: 1, State description: "Liver Qi is stable, mild parasympathetic activation"}.

[0077] S3: A random forest algorithm is used to perform multi-source fusion analysis on respiratory depth, respiratory rate, EEG signals, and heart rate variability data to generate personalized health intervention strategies, including:

[0078] The random forest algorithm was used to perform data fusion analysis on the user's respiratory depth, respiratory rate, electroencephalogram (EEG) signal and heart rate variability data, and a training sample set for the random forest algorithm was constructed.

[0079] Random forest is an ensemble learning method that constructs multiple decision trees and generates the final prediction result through voting or averaging. It has advantages such as resistance to overfitting and compatibility with multidimensional data. Data fusion analysis refers to the comprehensive processing of multiple physiological signals from different sources within the same analytical framework.

[0080] After denoising and filtering, respiratory depth data is recorded in milliliters as the average expansion volume per respiratory cycle; respiratory rate data is recorded in breaths per minute as the number of respiratory cycles per minute; EEG signals refer to the alpha wave power spectral density value generated by the digital signal processing unit, in microwatts per square centimeter; heart rate variability data is measured in milliseconds as the standard deviation of heartbeat intervals. The multi-source time-series data obtained through the above preprocessing steps are then timestamped to construct a feature matrix. Each row of the feature matrix corresponds to an analysis time window, such as a 5-second or 10-second sliding window; each column corresponds to a feature quantity, including the average respiratory depth, average respiratory rate, average alpha wave power, standard deviation of heartbeat intervals, and optional additional features such as the rate of change of respiratory depth or the rate of change of alpha wave amplitude envelope.

[0081] Taking a 10-second sliding window as an example, assuming that at a certain moment, the average respiratory depth collected by the smart wearable device is 360 ml, the average respiratory rate is 9 breaths / minute, the average alpha wave power is 20 microwatts / cm², and the standard deviation of the heart rate interval is 50 milliseconds, then the feature vector of this time window is [360, 9, 20, 50]. Each feature vector, together with the corresponding health label (e.g., mild stress, normal state, high stress), constitutes a training sample record. The health label can be determined by a clinical assessment scale (such as the GAD-7 score) or reference laboratory indicators. For example, a GAD-7 score ≥ 8 is marked as "excessive stress," and a GAD-7 score < 5 is marked as "normal stress."

[0082] The training sample set is an ordered set of feature vector and label pairs within all time windows, organized into a matrix and label array according to standard machine learning formats. Assuming 1000 sliding window time periods are collected, and 1000 corresponding health labels are obtained from questionnaire assessments, the training sample set would consist of 1000 samples, 4 input features, and 1 output label column. This training sample set is input into the random forest model training module to build a set of decision trees. During training, each decision tree randomly selects a subset of features and a subset of samples for node splitting, ultimately forming a forest of several trees used for feature importance analysis and strategy generation.

[0083] The random forest algorithm is used to perform feature importance analysis on the training sample set to determine the combination of feature parameters closely related to the user's health status;

[0084] Feature importance analysis is used to evaluate the contribution of each input feature to the model output. Random forests measure importance by calculating the average contribution of each feature to Gini impurity or information gain during node splits across all decision trees. Feature parameter combinations refer to a set of highly important features used for subsequent mapping relationship construction.

[0085] In a trained random forest model, each decision tree's splitting node is visited, and the degree of reduction in node impurity is recorded each time a certain feature is used for splitting. The impurity reduction values ​​for the same feature across all splits are summed, and the average or normalized value is used as the feature's importance score. For example, the cumulative reduction in Gini impurity for the respiratory depth feature across all trees is 12.3, for respiratory rate it is 8.7, for alpha wave power it is 15.2, and for the standard deviation of heartbeat interval it is 14.1. Therefore, the importance ranking of these features is: alpha wave power > standard deviation of heartbeat interval > respiratory depth > respiratory rate.

[0086] Based on importance ranking, the combination of features most closely related to health status is selected by setting a threshold or using a strategy based on the top N features. The threshold can be set to a cumulative importance contribution of 90%, or the top 3 most important features can be selected. For example, in the example above, the calculated importance contribution percentages of alpha wave power, standard deviation of heartbeat interval, and respiratory depth are 35%, 32%, and 29%, respectively, with a cumulative contribution of 96%, which meets the threshold requirement. Therefore, these three features are determined to be the key feature parameter combination.

[0087] After the feature parameter combination is formed, it is labeled with the feature name and its statistical attributes, including the feature name (e.g., "average alpha wave power"), quantization unit (µW / cm²), calculation window (sliding window length), and normalization or standardization parameters (e.g., Z-score transformation parameters with a mean of 0 and a standard deviation of 1). This key feature parameter combination is used to reduce the size of the mapping library and improve the efficiency of personalized strategy generation.

[0088] Based on the combination of feature parameters, a mapping relationship library between a user's personal health status and intervention measures is established to generate personalized health intervention strategies suitable for a user's personal health status.

[0089] The mapping database is a data structure that establishes a one-to-one or one-to-many mapping between key physiological characteristics and specific interventions. Interventions include breathing guidance techniques, acupoint phototherapy plans, and the intensity and duration of microcurrent stimulation. Personalized health intervention strategies refer to generating precise and executable intervention plans for users based on their current health status, so that smart wearable devices can execute them automatically.

[0090] The possible value ranges of key characteristic parameter combinations are divided into intervals. For example, the average alpha wave power is 0–10 μW / cm², 10–20 μW / cm², and 20–30 μW / cm²; the standard deviation of the heartbeat interval is 0–30 ms, 30–60 ms, and 60–90 ms; and the respiratory depth is 0–300 ml, 300–500 ml, and 500–700 ml. This forms a three-dimensional interval grid, with each grid cell corresponding to a specific health status sub-type.

[0091] Combining a TCM intervention model and random forest training feedback, appropriate intervention programs are specified for each grid cell. For example, grid cell [A: 10-20 microwatts / cm², B: 30-60 milliseconds, C: 300-500 ml] corresponds to a slightly activated parasympathetic nervous system state, and can be recommended as "low-frequency deep breathing with the 'Xu' sound training program, 3 minutes of red light irradiation at Baihui acupoint, and 20 microamps of microcurrent stimulation for 30 seconds"; while grid cell [A: 20 microwatts / cm², B: 60 milliseconds, C: 500 ml] corresponds to an over-relaxed state, and can be recommended as "2 minutes of rapid shallow breathing with the 'He' sound training program, 1 minute of blue light irradiation at Fengchi acupoint, and 10 microamps of microcurrent stimulation for 15 seconds", etc.

[0092] The mapping database is stored in tabular or key-value pair format. The key is an identifier for a combination of feature parameter ranges, such as "P10-20_S30-60_D300-500". The value is an intervention protocol structure, including the identifier of the breathing guidance audio file, a list of LED light source wavelength parameters, and the amplitude and duration of the current pulses. The mapping database is maintained on a cloud platform or in local storage and can be dynamically updated based on long-term user feedback and new sample data.

[0093] When generating personalized health intervention strategies, the key feature combination values ​​calculated in real time are mapped to the corresponding intervals, and the mapping relationship library is retrieved to obtain the corresponding intervention plan. For example, if the key feature combination measured at a certain moment is alpha wave power of 15 microwatts / square centimeter, heart rate interval standard deviation of 45 milliseconds, and respiratory depth of 350 ml, it is mapped to the "P10-20_S30-60_D300-500" key, generating the plan "Xu-style breathing guidance audio, Baihui acupoint red light irradiation for 3 minutes, and 20 microamp microcurrent stimulation for 30 seconds".

[0094] The generated personalized health intervention strategy is encapsulated into an execution instruction set and sent to the control module of the smart wearable device via Bluetooth or near-field communication. The breathing guidance interface is activated, the LED light source is controlled, and microcurrent stimulation is initiated in a predetermined sequence, forming a complete closed-loop intervention process.

[0095] S4: Based on the user's autonomic nervous system state and personalized health intervention strategies, establish a health risk assessment model for the user and determine the early warning trigger conditions, including:

[0096] Map the specific organ function state parameters corresponding to the user's autonomic nervous system state assessment results with the feature parameter combination in the personalized health intervention strategy to establish a matching rule between the user's autonomic nervous system state and the health intervention strategy.

[0097] The user's autonomic nervous system status assessment results are classification information obtained from physiological feature vectors fused through a respiratory-visceral-neural regulation model. This classification information includes autonomic nervous system functional state category codes and corresponding visceral function category codes. For example, functional state code 1 represents mild parasympathetic nerve activation, and visceral function code 1 represents stable liver function. The feature parameter combination in the personalized health intervention strategy is a set of key physiological indicators and their numerical ranges selected in random forest analysis, such as an average alpha wave power of 10–20 microwatts / cm², a standard deviation of heart rate interval of 30–60 milliseconds, and an average respiratory depth of 300–500 ml.

[0098] Mapping operations refer to establishing a correspondence between the two types of information mentioned above. Mapping rules can be represented as a set of triplets: the first triplet is the autonomic nervous system functional state category code, the second is the organ function category code, and the third is the feature parameter combination identifier. The feature parameter combination identifier uses a format similar to "P10-20_S30-60_D300-500", where P represents the alpha wave power range, S represents the heart rate interval standard deviation range, and D represents the respiratory depth range. For example, the triplet [1,1,"P10-20_S30-60_D300-500"] indicates that when the parasympathetic nervous system is mildly activated and liver function is stable, a feature combination of alpha wave power 10-20, heart rate interval standard deviation 30-60, and respiratory depth 300-500 should be matched.

[0099] The matching rule base is stored in key-value pairs. The key is a concatenated string of functional state and organ function codes, such as "FS1_ZP1", and the value is a list of combinations of feature parameters, such as ["P10-20_S30-60_D300-500", "P8-15_S25-55_D280-480"]. When establishing mapping rules, a large amount of user data is collected, and multiple assessment results are automatically matched with personalized health intervention strategies. The most frequently occurring feature parameter combinations under each functional state and organ function code are statistically analyzed and used as recommended entries. For example, if data is collected and assessed from 100 subjects in a resting state, and 80 assessments are for functional state code 1 and organ code 1, and the corresponding feature combination "P10-20_S30-60_D300-500" has the highest frequency, then the key "FS1_ZP1" will correspond to this combination in the mapping rule base.

[0100] The mapping rule base can be implemented using a two-dimensional hash table structure. Row indices represent functional status codes, and column indices represent organ function codes. Each cell stores a list of feature parameter combinations and their frequencies. During the mapping operation, the current user's functional status code and organ function code are read first, then the corresponding cell is located, and the list of feature parameter combinations is extracted. This list is subsequently used to determine the health risk level assessment threshold and the early warning triggering strategy.

[0101] For example, when the user's functional status assessment result is coded as 2 (sympathetic nerve excitation) and the visceral function is coded as 3 (hyperactive spleen and stomach function), the feature combination corresponding to the "FS2_ZP3" key in the mapping rule base is ["P15-25_S20-40_D200-350", "P12-18_S25-45_D250-400"].

[0102] Based on the matching rules between the user's autonomic nervous system state and health intervention strategies, the user's health risk level is classified and determined.

[0103] Health risk grading is used to map a user's current physiological and functional state to an easily understood and manageable risk level. Health risk levels can be defined as three or five levels; for example, a three-level grading might include normal, low risk, and high risk. The grading is based on historical health intervention effect feedback and clinical research results from matching rule base combinations of characteristic parameters.

[0104] The determination process is as follows: First, obtain the list of feature parameter combinations obtained from the mapping rules, for example, ["P10-20_S30-60_D300-500", "P8-15_S25-55_D280-480"]. For each feature combination, read the historical risk feedback data associated with it, where each feedback data includes the feature combination, the user's actual health label (e.g., GAD-7 score range or clinical heart rate fluctuation range), and the corresponding risk level. Calculate the average risk level of each feature combination in the historical samples. For example, if out of 100 cases, this feature combination corresponds to 60 cases of mild risk, 30 cases of moderate risk, and 10 cases of high risk, then the average risk level is between mild and moderate risk.

[0105] The risk level reference value for this feature combination is generated using either a weighted average or the most frequent ranking method. Weighted average example: Assigning 1 to mild risk, 2 to medium risk, and 3 to high risk, the average value = (60×1 + 30×2 + 10×3) / 100 = 1.5, rounded down to 2, which is medium risk. The most frequent ranking method directly uses the risk level with the highest frequency. In this example, mild risk has the highest frequency, so it is classified as mild risk.

[0106] When multiple feature combination lists exist, the judgment results of each combination need to be summarized. The summarization method can be either a highest severity level priority principle (i.e., if any combination is judged as high risk, the overall risk level is high risk) or a majority voting principle (i.e., if one group is judged as medium risk and the other as low risk, the overall risk level is medium risk). The choice of principle can be preset in the mapping rule base. For example: if two feature combinations in the list are judged as low risk and medium risk respectively, using the majority voting principle will ultimately result in a medium risk judgment; using the priority principle will result in a medium risk judgment.

[0107] Based on the user's health risk level, determine the warning trigger conditions for the user's health risk;

[0108] Warning trigger conditions refer to the specific threshold settings that initiate device intervention and user alerts when a user's health risk level reaches a certain level. Warning trigger conditions are defined as a set of logical judgment conditions, including but not limited to rules such as the number of consecutive monitoring periods of risk level, the fusion feature score exceeding the upper limit threshold, or key physiological indicators exceeding limits.

[0109] Different trigger strategies are preset for each risk level. For example, the trigger conditions for the three risk levels are as follows:

[0110] Normal level: Risk level code 1, no warnings required, only continuous monitoring.

[0111] Mild risk: Risk level code 2. If the three consecutive assessment results are all mild risk, a low-intensity intervention will be triggered, including playing gentle breathing guidance sounds and short-term red light exposure.

[0112] Medium risk: Risk level code 3. If two consecutive classification results are both medium risk or any fusion score exceeds 120, a medium-intensity intervention will be triggered immediately, including breathing training with the "shh" sound, alternating red and blue light irradiation, microcurrent of 20 microamps for 30 seconds, and a prompt message will be pushed to the user terminal.

[0113] High risk: Risk level code 4. If any judgment result is high risk, a level 3 early warning response will be initiated, including playing an emergency breathing guidance sound, continuous red light irradiation of Baihui acupoint for 5 minutes, microcurrent of 50 microamps for 60 seconds, and sending an alarm message to the user's guardian or medical service platform.

[0114] S5: Based on the warning trigger conditions, when the health risk assessment model detects that the user's stress index exceeds a preset threshold, it accurately locates acupoints using 3D head scanning technology and controls the wearable device to implement acupoint microcurrent stimulation and acupoint phototherapy, including:

[0115] When the user's health risk assessment model determines that the user's stress index exceeds the pre-set health risk level threshold, the 3D head shape scanning technology is activated to perform a three-dimensional spatial scan of the user's head.

[0116] The health risk assessment model is a computational module that uses a random forest algorithm and a visceral-neural regulation model to jointly determine risk levels based on multi-source physiological characteristics such as respiratory depth, respiratory rate, EEG alpha wave power, and heart rate variability. The stress index is defined as the ratio of the GAD-7 scale score or multi-source physiological fusion score to historical clinical benchmarks. Health risk level thresholds are multiple numerical ranges determined based on clinical studies and population sample statistics. For example, the threshold for mild risk is set at GAD-7 ≥ 8 points or a fusion score ≥ 100, and the threshold for moderate risk is set at GAD-7 ≥ 12 points or a fusion score ≥ 120.

[0117] During real-time monitoring, each time the health risk assessment model generates a risk level output, it compares the risk level with a pre-stored threshold. For example, if the fusion score is 125, which is higher than the medium-risk threshold of 120, it is judged as medium to high risk, requiring intervention. If the fusion score is 135, and GAD-7 has not yet been updated, the stress index is still judged to exceed the medium-risk threshold based on the fusion score. Once the triggering conditions are met, the internal control unit of the smart wearable device calls the 3D head shape scanning technology interface.

[0118] The 3D head shape scanning technology combines a structured light or time-of-flight depth camera with a near-infrared laser stripe projection system to achieve high-density point cloud acquisition of the head's surface geometry. The scanning resolution is no less than 0.5 mm, the field of view covers the user's forehead and both temple areas, and the scanning speed is no less than 10 frames per second. During the scan, the head remains relatively still for five seconds to obtain complete head shape contour data; the data acquisition and processing delay does not exceed 200 milliseconds to ensure that the scan results are synchronized with subsequent acupoint location.

[0119] For example, when a user's breathing and heart rate are at high risk, the control unit sends a start command, the depth camera projects infrared coded stripes, and receives deformed stripe images sampled by a monocular camera. A phase-shift algorithm is then used to reconstruct a point cloud of the head surface. The reconstructed point cloud data generates a 3D triangular mesh model in the local processor and stores it in the device's memory, providing a basic model for acupoint coordinate extraction.

[0120] Based on the three-dimensional spatial information of the user's head shape obtained from the scan, the coordinates of the Baihui, Shenting, and Taiyang acupoints on the user's head are determined.

[0121] Baihui (GV20), Shenting (GV24), and Taiyang (EX-HN5) are key acupoints in Traditional Chinese Medicine (TCM) meridian theory used to regulate emotional and autonomic nervous system balance. They are located at the center of the top of the head, the center of the forehead, and the Taiyang points on both sides, respectively. The coordinates of the acupoints refer to the three-dimensional point coordinates (X, Y, Z) in the three-dimensional mesh model of the head shape, with the origin of the coordinate system selected as the highest point of the top of the head or the center of the cranial vault.

[0122] Acupoint location is first based on a standard cranial anatomy template, using a head-shaped mesh for non-rigid registration with the template. The template discretizes the location of Baihui (GV20) at 5 cm above the midpoint of the line connecting the top of the head and the two ears, Shenting (HT2) at 1.5 cm above the midpoint of the anterior hairline, and Taiyang (EX-HN5) approximately 1 cm posterior to the intersection of the lines connecting the ends of the eyebrows and the hairline. The iterative closest point (ICP) algorithm is used to align the template with the user's head-shaped mesh, completing the calculation of similarity transformation parameters, including rotation matrices and translation vectors.

[0123] After alignment, the acupoint markers in the template are mapped onto the user's head model using a similarity transformation to obtain the three-dimensional coordinates of the acupoints. For example, the coordinates of Baihui acupoint are assumed to be [X=0 mm, Y=+50 mm, Z=200 mm], Shenting acupoint [X=0 mm, Y=+30 mm, Z=180 mm], right temple [X=+40 mm, Y=+25 mm, Z=175 mm], and left temple [X=−40 mm, Y=+25 mm, Z=175 mm]. The coordinate accuracy is no less than ±2 mm to meet the precise alignment requirements of the microcurrent stimulation electrodes and phototherapy module.

[0124] For example, after completing the 3D head shape reconstruction, the registration module is called to align the user mesh with the head model template of the international 10-20 system. After a total of 30 iterations, the registration error is less than 1 mm. The coordinates of the Baihui acupoint markers on the template are mapped to the coordinates of the user's head shape, and then the position of each acupoint on the head model is displayed on the 3D visualization interface.

[0125] Based on the location coordinates, the microcurrent stimulation module and acupoint phototherapy module on the smart wearable device are respectively positioned at Baihui, Shenting and Taiyang acupoints;

[0126] The microcurrent stimulation module consists of a microelectrode array and a current-driven chip. The electrode array is installed inside the headband and in contact with the skin, communicating with the control unit via a flexible circuit board. The acupoint phototherapy module consists of a multi-band LED light source array and a lens assembly. The LED wavelengths cover red light (620–630 nm) and blue light (450–470 nm), with a light spot diameter of approximately 10 mm.

[0127] Based on the three-dimensional coordinates of each acupoint, the control unit completes positioning by aligning the coordinates with the servo micro-motors of the flexible electrode arm and LED arm. The positioning process includes: reading the acupoint coordinates, converting the coordinates into servo angle commands for the electrode arm and LED arm inside the headband, and sending them to the servo drive module via a serial bus; after the servo motor starts, it completes rotation and stops within 100 ms, with a positioning error of no more than ±1 mm.

[0128] For example, the coordinates of Baihui acupoint [X=0 mm, Y=50 mm, Z=200 mm] are converted into a horizontal rotation angle of 10° and a vertical rotation angle of 30° for the electrode arm; a horizontal rotation angle of 12° and a vertical rotation angle of 28° for the LED arm; and the positioning of Shenting acupoint and Taiyang acupoint is completed in the same way. After positioning, the electrode contact status is detected by capacitive contacts. If the contact resistance between the electrode and the scalp is confirmed to be less than 50 kΩ, stimulation is initiated.

[0129] Based on personalized health intervention strategies, microcurrent stimulation signals are output to Baihui, Shenting and Taiyang acupoints, while red or blue light waves corresponding to specific intervention strategies are irradiated simultaneously.

[0130] Personalized health intervention strategies are generated command sets that include parameters such as breathing guidance mantras, microcurrent stimulation intensity and frequency, phototherapy wavelength and irradiation duration, etc., tailored to the current health status. For example, the strategy corresponding to a moderate risk status is: "breathing guidance mantra, red light irradiation of Baihui acupoint for 180 s, microcurrent pulse of 20 μA at Baihui acupoint for 30 s, blue light irradiation of Shenting acupoint for 120 s, microcurrent pulse of 15 μA at Shenting acupoint for 20 s, alternating red and blue light irradiation of bilateral temples for 60 s."

[0131] The microcurrent stimulation signal waveform uses a DC pulse mode with a pulse frequency of 30 Hz and a pulse width of 100 μs. The current amplitude is precisely output and monitored in real time by the current drive chip, with an error controlled within ±1 μA. The signal output process is as follows: read the strategy parameters, and the control unit sequentially sends start, amplitude setting, frequency setting, and duration setting commands to each microcurrent stimulation module. After start-up, the output current and contact impedance are monitored in real time. If the impedance exceeds the set range, the process is interrupted and an error code is returned.

[0132] The phototherapy signal output process is as follows: The strategy wavelength and duration parameters are read, and the LED array is driven to illuminate according to wavelength, either in time-sharing or zone-sharing patterns. The red light wavelength is 620 nm ± 5 nm, and the blue light wavelength is 460 nm ± 5 nm. The light intensity is set to 5–20 mW / cm², and the light spot covers acupoints with a diameter of 10 mm. During irradiation, the phototherapy module's built-in photoelectric sensor measures the actual illuminance in real time and feeds it back to the control unit to adjust the LED drive current and maintain stable output.

[0133] For example, firstly, a 20 μA microcurrent pulse is initiated at the Baihui acupoint for 30 seconds, simultaneously with 620 nm red light irradiation at the Baihui acupoint for 180 seconds; after 30 seconds, the microcurrent is turned off, leaving only the red light irradiation; after this, 460 nm blue light irradiation is initiated at the Shenting acupoint for 120 seconds, along with a 15 μA microcurrent for 20 seconds; finally, alternating red and blue light irradiation is initiated at both temples for 60 seconds. All signal outputs complete mode switching within a 5 ms response time, ensuring that the intervention sequence and duration adhere to the strategy requirements.

[0134] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0135] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0136] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0137] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0139] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0140] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0141] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0143] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1.A health monitoring and early warning method based on a smart wearable device, characterized in that, Comprise the following steps: S1: Real-time collection of user's respiratory depth, respiratory rate, EEG signal and heart rate variability data by high-precision respiratory sensor, EEG alpha wave sensor and heart rate variability monitoring module, and preprocessing; S2: Based on the six-word formula breathing method, a respiratory-visceral nerve regulation model is constructed to evaluate the user's autonomic nervous system state; S3: Random forest algorithm is used to analyze the multi-source fusion of respiratory depth, respiratory rate, EEG signal and heart rate variability data, and generate personalized health intervention strategy; S4: Based on the user's autonomic nervous system state and personalized health intervention strategy, a user health risk assessment model is established to determine the warning trigger condition; S5: According to the warning trigger condition, when the health risk assessment model detects that the user's stress index exceeds the preset threshold, the acupoint is accurately positioned by 3D head shape scanning technology, and the wearable device is controlled to implement acupoint micro-current stimulation and acupoint light therapy. 2.The health monitoring and early warning method based on smart wearable device of claim 1, wherein, S1, specifically: Real-time collection of user's respiratory depth and respiratory rate by high-precision respiratory sensor built-in intelligent wearable device; Real-time collection of user's brain alpha wave feature data by EEG alpha wave sensor; Real-time collection of user's heart rate interval standard deviation value within normal sinus rhythm cycle by heart rate variability monitoring module; The collected respiratory depth, respiratory rate, EEG alpha wave feature data and heart rate interval standard deviation value are respectively denoised and filtered. 3.The health monitoring and early warning method based on smart wearable device of claim 2, wherein, S2, specifically: The word formula in the six-word formula breathing method corresponds to the data characteristics of the user's respiratory depth and respiratory rate, and the correlation rules between respiratory depth, respiratory rate and specific visceral function are established; According to the correlation rules, the respiratory depth, respiratory rate characteristics of different word formulas are fused with EEG alpha wave feature data and heart rate interval standard deviation value; The matched data and the pre-established respiratory-visceral nerve regulation model are matched to evaluate the user's current autonomic nervous system function state, and the user's autonomic nervous system state evaluation result is output. 4.The health monitoring and early warning method based on smart wearable device of claim 3, wherein, S3, specifically: Random forest algorithm is used to analyze the data fusion of user's respiratory depth, respiratory rate, EEG signal and heart rate variability data, and construct the training sample set of random forest algorithm; The feature importance analysis of the training sample set is performed by random forest algorithm to determine the feature parameter combination closely related to the user's health state; Based on the feature parameter combination, a mapping relationship library between user's personal health state and intervention measures is established to generate personalized health intervention strategy suitable for user's personal health state. 5.The health monitoring and early warning method based on smart wearable device of claim 4, wherein, S4, specifically: The specific visceral function state parameters corresponding to the user's autonomic nervous system state evaluation result and the feature parameter combination in the personalized health intervention strategy are mapped to establish the matching rules between the user's autonomic nervous system state and the health intervention strategy; According to the matching rules between the user's autonomic nervous system state and the health intervention strategy, the user's health risk level is graded; According to the user's health risk level, the warning trigger condition of user's health risk is determined. 6.The health monitoring and early warning method based on smart wearable device of claim 5, wherein, S5, specifically: When the user health risk assessment model determines that the stress index of the user exceeds a pre-set health risk level threshold, a 3D head shape scanning technology is started to perform a three-dimensional space scanning on the head of the user; According to the three-dimensional space information of the head shape of the user obtained by scanning, position coordinates of the Baihui acupoint, the Shenting acupoint and the temple acupoint of the user are determined; According to the position coordinates, a micro-current stimulation module and an acupoint light therapy module on the smart wearable device are respectively positioned at the Baihui acupoint, the Shenting acupoint and the temple acupoint; According to the individualized health intervention strategy, a micro-current stimulation signal is output to the Baihui acupoint, the Shenting acupoint and the temple acupoint, and a red light wave or a blue light wave corresponding to a specific intervention strategy is irradiated.

Images