Real-time health monitoring and intelligent diagnosis method applied to wearable device

Abstract

The invention relates to the field of wearable equipment, in particular to a real-time health monitoring and intelligent diagnosis method applied to the wearable equipment. The method comprises the following steps: collecting human body bioelectrical impedance data, and constructing a spectrum data matrix; preprocessing the spectrum data matrix to obtain a standardized spectrum matrix; dynamic spectrum features are extracted from the standardized spectrum matrix and analyzed, and a health state vector is generated; and based on the health state vector, calculating a health risk assessment value, and analyzing and assessing the health condition of the user. The problem that the prediction accuracy and sensitivity are limited due to the fact that complex physiological state changes cannot be comprehensively captured is solved; generally, a linear model is used for analyzing data, a complex non-linear relation between physiological data is neglected, and the real-time adaptive capacity to dynamic change of the data is lacked; the problem that the risk assessment result is not accurate enough due to the lack of comprehensive consideration of historical data, time decay, feature influence and other factors in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of wearable devices, and in particular to a real-time health monitoring and intelligent diagnosis method applied to wearable devices. Background Art

[0002] In recent years, the application of wearable devices in the field of health monitoring has made significant progress. The wearable devices are usually equipped with a variety of sensors, such as heart rate monitors, blood oxygen saturation sensors, body temperature sensors, etc., which can collect the user's physiological data in real time. The integration of the sensors allows individuals to perform continuous health monitoring in their daily lives. In addition, with the development of data processing capabilities and machine learning algorithms, intelligent diagnostic systems are able to analyze the collected physiological data to predict and detect possible health problems. For example, through machine learning technologies such as neural networks, the user's health status change pattern is learned to provide personalized health advice and warnings.

[0003] Although wearable health monitoring technology has made significant progress, there are still some challenges and shortcomings: first, traditional wearable health monitoring methods mainly monitor single or a few physiological data, which may not be able to fully capture complex changes in physiological status, resulting in limited prediction accuracy and sensitivity; second, linear models are usually used to analyze data, ignoring the complex nonlinear relationship between physiological data, which may lead to misjudgment and delayed response to health status; third, the lack of real-time adaptive capabilities to dynamic changes in physiological data makes it difficult to cope with individual differences and environmental changes; in addition, current health risk assessment models often simply process physiological data linearly, lacking comprehensive consideration of historical data, time decay, feature influence and other factors, resulting in inaccurate risk assessment results, and thus unable to issue health warnings to users in a timely and effective manner. In summary, traditional wearable health monitoring technology still needs further breakthroughs in improving the comprehensiveness, accuracy and timeliness of health monitoring. Summary of the invention

[0004] The present invention provides a real-time health monitoring and intelligent diagnosis method applied to wearable devices to solve the problems that traditional wearable health monitoring methods mainly monitor a single or a few physiological data, which may not be able to fully capture complex changes in physiological states, resulting in limited prediction accuracy and sensitivity; linear models are usually used to analyze data, ignoring the complex nonlinear relationship between physiological data, and lacking real-time adaptive capabilities to dynamic changes in data, making it difficult to cope with individual differences and environmental changes; current health risk assessment models often simply perform linear processing on physiological data, lack comprehensive consideration of multiple factors such as historical data, time decay, and feature influences, resulting in inaccurate risk assessment results and inability to issue health warnings to users in a timely and effective manner.

[0005] The real-time health monitoring and intelligent diagnosis method for wearable devices of the present invention specifically includes the following technical solutions:

[0006] The real-time health monitoring and intelligent diagnosis method applied to wearable devices includes the following steps:

[0007] S1: Collect human bioelectrical impedance data and construct a spectrum data matrix; preprocess the spectrum data matrix to obtain a standardized spectrum matrix; extract dynamic spectrum features from the standardized spectrum matrix through spectrum analysis method;

[0008] S2: Build a dynamic spectrum model, analyze the dynamic spectrum characteristics, and generate a health status vector; calculate the health risk assessment value based on the health status vector; and analyze and evaluate the user's health status based on the health risk assessment value.

[0009] Preferably, the S1 specifically includes:

[0010] In the implementation process of preprocessing, the spectrum data matrix is ​​subjected to noise filtering to obtain a filtered spectrum data matrix; an adaptive spectrum enhancement algorithm is used to perform adaptive spectrum enhancement on the filtered spectrum data matrix to obtain an enhanced spectrum matrix; the specific formula of the adaptive spectrum enhancement algorithm is as follows:

[0011]

[0012] in, is the enhanced spectrum matrix; At time point t j and frequency f i Enhanced spectrum value under ; is the time point t in the filtered spectrum data matrix j and frequency f i Human bioelectrical impedance data below; is the filtered spectrum data matrix at frequency f i The standard deviation of human bioelectrical impedance data at all time points below; is the filtered spectrum data matrix at frequency f i is the mean of the human bioelectrical impedance data at all time points; exp is the exponential operation; i is the frequency index; n is the total number of frequencies.

[0013] Preferably, the S1 specifically includes:

[0014] The enhanced spectrum matrix is ​​processed by data normalization, and the standardized spectrum matrix is ​​generated by combining exponential transformation and normalized fraction.

[0015] Preferably, the S1 specifically includes:

[0016] An exponential weight factor is introduced to perform exponential weighting on the standardized spectrum values ​​in the standardized spectrum matrix to obtain dynamic spectrum feature components; based on the dynamic spectrum feature components, dynamic spectrum features are constructed.

[0017] Preferably, the S2 specifically includes:

[0018] The dynamic spectrum model parameters are introduced to construct a dynamic spectrum model. The health state vector is obtained by combining the dynamic spectrum characteristics of multiple frequency bands. The specific calculation formula is as follows:

[0019] S(t j )=log(||tanh(W·F(t j ))|| 2 +exp(V·sin(F(t j ))))

[0020] Among them, W and V are the parameters of the dynamic spectrum model, W is the linear weight matrix, and V is the nonlinear weight matrix; S(t j ) is the time point t j The health status vector of ||·|| 2 is the second norm; tanh is the hyperbolic tangent function; sin is the sine function; F(t j ) is the time point t j Dynamic spectrum characteristics.

[0021] Preferably, the S2 specifically includes:

[0022] A nonlinear dynamic spectrum optimization algorithm is introduced, and an optimization function is defined to optimize the dynamic spectrum model parameters. The specific definition of the optimization function is as follows:

[0023]

[0024] Where L is the target value of the optimization function; It's at the time A historical health state vector calculated based on historical dynamic spectrum characteristics; It's at the time Historical dynamic spectrum characteristics; Indicates a point in history; is the spectrum feature mapping value, which is obtained based on the historical dynamic spectrum features using the regression model; is the square operation of the two-norm; is the regularization coefficient.

[0025] Preferably, the S2 specifically includes:

[0026] Based on the objective value of the optimization function, the dynamic spectrum model parameters are optimized and updated using the gradient descent method in combination with the improved update formula to obtain updated dynamic spectrum model parameters; based on the updated dynamic spectrum model parameters, the health state vector is updated.

[0027] Preferably, the S2 specifically includes:

[0028] Based on the health state vector, the time decay factor is introduced to predict the health state vector at a future time point. The specific formula is as follows:

[0029]

[0030] in, At a future time The health status vector of is the prediction step length, which means from the current time point t j The time step for future prediction; α is the current health status weight; β is the influence weight of historical dynamic spectrum characteristics; γ is the time attenuation factor; F(t j -τ) is at time point t j -τ dynamic spectrum characteristics; T is the lag step, indicating the time from the current time point t j The number of time steps to trace back.

[0031] Preferably, the S2 specifically includes:

[0032] Based on the health status vector at a future time point, the health risk assessment value is calculated; the health risk assessment value is used to analyze and evaluate the user's health status. When the health risk assessment value is greater than or equal to the preset safety threshold, it indicates that the user's current health status has potential risks, and a real-time early warning notification is sent to the user; when the health risk assessment value is less than the safety threshold, no early warning notification is sent, and monitoring and recording continue.

[0033] The beneficial effects of the technical solution of the present invention are:

[0034] 1. By collecting human bioelectrical impedance data at different time points and multiple frequencies, a spectrum data matrix is ​​constructed, and the spectrum data matrix is ​​subjected to noise filtering, adaptive spectrum enhancement and standardization processing to extract dynamic spectrum features. This can capture changes in the electrical properties of human biological tissues and accurately identify subtle changes in the composition and structure of human tissues. The dynamic spectrum features enable the intelligent diagnostic system to more accurately identify changes in physiological states and potential abnormalities, significantly improving the sensitivity and accuracy of health monitoring.

[0035] 2. Use dynamic spectrum features to build a dynamic spectrum model, and identify the change pattern of health status by learning dynamic spectrum features and health status vectors. Optimize the dynamic spectrum model parameters through nonlinear dynamic spectrum optimization algorithm to further improve the prediction ability and realize adaptive adjustment of the dynamic spectrum model, making health status prediction more accurate and providing a scientific basis for users' health management.

[0036] 3. Combining the current health status vector, historical dynamic spectrum characteristics, time attenuation and other factors, by reasonably setting the prediction step, the current health status weight and the historical dynamic spectrum characteristic influence weight, the health status vector at future time points is predicted and the health risk assessment value is calculated; the health risk assessment value is compared with the preset safety threshold. When the health risk assessment value exceeds the safety threshold, an early warning notification is sent to the user in a timely manner, so that the user can take timely measures to improve the effectiveness of overall health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the real-time health monitoring and intelligent diagnosis method applied to wearable devices described in the present invention. DETAILED DESCRIPTION

[0038] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0040] The specific scheme of the real-time health monitoring and intelligent diagnosis method for wearable devices provided by the present invention is described in detail below with reference to the accompanying drawings.

[0041] See attached Figure 1 , which shows a flow chart of a real-time health monitoring and intelligent diagnosis method for a wearable device provided by an embodiment of the present invention, the method comprising the following steps:

[0042] S1: Collect human bioelectrical impedance data and construct a spectrum data matrix; preprocess the spectrum data matrix to obtain a standardized spectrum matrix; extract dynamic spectrum features from the standardized spectrum matrix through spectrum analysis method;

[0043] Wearable devices measure the bioelectrical impedance of the human body through built-in multi-band current sensors. A series of alternating currents with preset frequencies (such as 1kHz, 10kHz, 100kHz, 500kHz, 1MHz, etc.) are used to stimulate human tissues, and human bioelectrical impedance data is collected at the frequencies. Electrical impedance data is a parameter that reflects the electrical properties of biological tissues. Human bioelectrical impedance data at different frequencies can provide multidimensional information about the composition and structure of human tissues, such as water content, cell density, and tissue health. Wearable devices continuously collect human bioelectrical impedance data at fixed time intervals (such as once per second) to capture the dynamic changes in the physiological state of the human body. The selection of specific frequencies and time points can be specifically set according to the measurement objectives (such as heart rate monitoring, body fat analysis, muscle mass assessment, etc.) and application scenarios (such as exercise, rest, clinical environment, etc.).

[0044] Using human bioelectrical impedance data Z(fi,t j ) constructs the spectrum data matrix, the spectrum data matrix expression is as follows:

[0045]

[0046] Where Z is the spectrum data matrix; Z(fi,t j ) represents the time point t j and frequency f i The human bioelectrical impedance data under the condition of 1 is shown in Figure 1; i is the frequency index, ranging from 1 to n; j is the time point index, ranging from 1 to m; f i is the ith frequency for collecting human bioimpedance data; t j is the jth time point at which the human bioelectrical impedance data is collected; n is the total number of frequencies; and m is the total number of time points.

[0047] The spectrum data matrix is ​​preprocessed, and the preprocessing includes noise filtering, adaptive spectrum enhancement and data standardization. The specific implementation process of the preprocessing is as follows:

[0048] First, a specific technology (such as adaptive Kalman filtering or wavelet transform) is used to filter the noise of the spectrum data matrix to obtain a filtered spectrum data matrix; noise filtering is an existing technology and will not be described in detail here;

[0049] Furthermore, an adaptive spectrum enhancement algorithm is used to perform adaptive spectrum enhancement on the filtered spectrum data matrix to obtain an enhanced spectrum matrix; the adaptive spectrum enhancement algorithm is a method for processing spectrum data, which can highlight key features in spectrum data and improve the accuracy and effectiveness of data analysis. By analyzing the statistical characteristics (mean and standard deviation) of human bioelectrical impedance data at different frequencies, the filtered spectrum data matrix is ​​adaptively adjusted to ensure that the degree of data enhancement at each frequency is dynamically changing, so that the significant features of human bioelectrical impedance data can be effectively highlighted under different conditions. At the same time, the adaptive spectrum enhancement algorithm compresses the range of human bioelectrical impedance data at each frequency by logarithmic transformation, and amplifies the signal characteristics by combining exponential operation, so that smaller changes can also be identified and analyzed;

[0050] The formula of adaptive spectrum enhancement algorithm is as follows:

[0051]

[0052] in, is the enhanced spectrum matrix; At time point t j and frequency f i Enhanced spectrum value under ; is the filtered spectrum data matrix; is the time point t in the filtered spectrum data matrix j and frequency f i Human bioelectrical impedance data below; is the filtered spectrum data matrix at frequency f i The standard deviation of human bioelectrical impedance data at all time points below; is the filtered spectrum data matrix at frequency f i The mean of the human bioelectrical impedance data at all time points; exp is the exponential operation;

[0053] Finally, the enhanced spectrum matrix is ​​subjected to data standardization, and combined with exponential transformation and normalized fraction, a standardized spectrum matrix is ​​generated, which enhances the signal contrast and eliminates the influence of frequency differences, can more effectively enhance signal features, and improve the accuracy of subsequent feature extraction and analysis. The specific formula for data standardization is as follows:

[0054]

[0055] in, is the normalized spectrum matrix; At time point t j and frequency f i The normalized spectrum value under 0is the reference frequency, used for frequency normalization. The middle value of the preset frequencies can be selected, or the frequency that can reflect the changes in physiological state can be selected through experimental experience and literature, and it is set according to the implementation scenario; ∈ is a positive number used to avoid the denominator being zero, and it is set according to the specific implementation scenario.

[0056] Furthermore, the spectrum analysis method is used to extract dynamic spectrum features from the standardized spectrum matrix, and the standardized spectrum values ​​at multiple frequencies are analyzed to reflect the tissue status and change trend at a specific time point. By extracting dynamic spectrum features, the changes in electrical impedance data of human biological tissues at different frequencies and time points can be effectively identified and analyzed, helping to identify potential physiological abnormalities and assess changes in health status.

[0057] The extraction formula of dynamic spectrum features is as follows:

[0058] F(t j )=[F 1 (t j ), F 2 (t j ), …, F n (t j )]

[0059]

[0060] Among them, F(t j ) is at time t j Dynamic spectrum characteristics of F i (t j ) is at time t j and frequency f i Dynamic spectrum characteristic components under F n (t j ) is at time t j and frequency f n Dynamic spectrum characteristic components under ; is the frequency f in the normalized spectrum matrix i The standard deviation of human bioelectrical impedance data at all time points below; is an exponential weighting factor that emphasizes the impact of different frequencies by exponentially weighting the normalized spectrum values.

[0061] S2: Build a dynamic spectrum model, analyze the dynamic spectrum characteristics, and generate a health status vector; calculate the health risk assessment value based on the health status vector; and analyze and evaluate the user's health status based on the health risk assessment value.

[0062] A dynamic spectrum model is constructed. The dynamic spectrum model is a mathematical model for health monitoring. The dynamic spectrum model predicts the user's health status by analyzing the dynamic spectrum features extracted from human bioelectrical impedance data. The dynamic spectrum features are processed in a nonlinear manner using logarithmic and exponential operations, so as to better handle the complex patterns and relationships in the dynamic spectrum features and provide early warning of potential health risks.

[0063] The dynamic spectrum model combines the dynamic spectrum characteristics of multiple frequency bands to obtain a health status vector that reflects the user's overall physiological state. It is used to describe the user's comprehensive health status at a certain point in time, understand the patient's health trends, and evaluate and predict future health risks.

[0064] The specific calculation formula of the health state vector is as follows:

[0065] S(t j )=log(||tanh(W·F(t j ))|| 2 +exp(V·sin(F(t j ))))

[0066] Among them, W and V are the parameters of the dynamic spectrum model, W is the linear weight matrix, and V is the nonlinear weight matrix; S(t j ) is the time point t j The health status vector of ||·|| 2 is the second norm (Euclidean norm); tanh is the hyperbolic tangent function, which is used to increase the nonlinear characteristics of the dynamic spectrum characteristics and compress the data range; sin is the sine function.

[0067] Furthermore, a nonlinear dynamic spectrum optimization algorithm is used to optimize the parameters of the dynamic spectrum model to improve the prediction ability of the dynamic spectrum model on the relationship between the dynamic spectrum characteristics and the health state vector;

[0068] The nonlinear dynamic spectrum optimization algorithm uses an optimization function to adjust the dynamic spectrum model parameters. The goal of the optimization function is to minimize the error between the health state vector and the spectrum feature map value. At the same time, a regularization term is introduced to control the complexity of the dynamic spectrum model to avoid overfitting. During the optimization process, the dynamic spectrum model parameters are adjusted according to the gradient information to improve the generalization ability.

[0069] The specific implementation steps of the nonlinear dynamic spectrum optimization algorithm are as follows:

[0070] First, define the optimization function. The specific formula is as follows:

[0071]

[0072] Among them, L is the target value of the optimization function, and the summation range is selected according to the expert experience method; It's at the time A historical health state vector calculated based on historical dynamic spectrum characteristics; It's at the time The historical dynamic spectrum characteristics are from an existing database. represents a historical time point selected based on expert experience; is a spectrum feature mapping value, which is obtained based on historical dynamic spectrum features using a regression model. The regression model is an existing technology and is set according to a specific implementation scenario; is the square operation of the two-norm; is the regularization coefficient, which is used to control the influence of the regularization term to avoid overfitting and is obtained through experimental tuning;

[0073] Then, based on the objective value of the optimization function, the dynamic spectrum model parameters W and V are updated using the gradient descent method combined with the improved update formula. The improved update formula is as follows:

[0074]

[0075]

[0076] Wherein, W(k+1) is the linear weight matrix at the k+1th iteration, i.e., the updated linear weight matrix; W(k) is the linear weight matrix at the kth iteration; V(k+1) is the nonlinear weight matrix at the k+1th iteration, i.e., the updated nonlinear weight matrix; V(k) is the nonlinear weight matrix at the kth iteration; the initial values ​​of W(k) and V(k) are obtained by random initialization; η is the learning rate, which can be obtained by learning rate scheduling technology or hyperparameter optimization method (such as grid search or Bayesian optimization). The specific method used can be set according to the implementation scenario, which will not be described here; T represents transpose; is the gradient of the optimization function with respect to the linear weight matrix W(k); is the gradient of the optimization function with respect to the nonlinear weight matrix V(k). Gradient calculation is a well-known method for those skilled in the art and will not be described in detail here.

[0077] When the preset maximum number of iterations is reached or the rate of change of the optimization function is less than the set rate of change threshold, the iteration is stopped to obtain the updated linear weight matrix and nonlinear weight matrix. The maximum number of iterations and the rate of change threshold are set according to the specific implementation scenario and will not be described in detail here.

[0078] The health state vector is updated based on the updated dynamic spectrum model parameters.

[0079] Furthermore, based on the health status vector, the health risk of the user is assessed to obtain a health risk assessment value. The specific formula is as follows:

[0080]

[0081]

[0082] in, At a future time The health status vector of is the prediction step length, which means from the current time point t j The time step of the future prediction is the specific time interval of the prediction; α is the current health state weight, which is used to control the influence of the current health state vector; β is the historical dynamic spectrum feature influence weight, which is used to adjust the influence of the historical dynamic spectrum feature; γ is the time attenuation factor, which is used to control the degree of decrease in the influence of the historical dynamic spectrum feature; F(t j -τ) is at time point t j -τ dynamic spectrum characteristics; T is the lag step, indicating the time from the current time point t j The time step to trace back; R(t j ) is at time t j The health risk assessment value; λ is the total time period of the prediction, which determines how long the health risk assessment will be performed; In order to evaluate the standard deviation parameters of the dynamic spectrum model, the historical dynamic spectrum characteristics of a specific time range are selected according to the expert experience method, and the standard deviation of the historical dynamic spectrum characteristics of the time range is calculated to obtain the volatility and variation range of the data.

[0083] Use the health risk assessment value to analyze and evaluate the user's health status. Set the safety threshold according to the specific implementation scenario. When the health risk assessment value is greater than or equal to the safety threshold, it means that the user's current health status has potential risks, and a real-time warning notification is sent to the user through a wearable device or mobile application. The warning notification informs the user of the current health status and provides corresponding health management suggestions, such as suggesting that the user adjust their lifestyle, increase exercise frequency, or undergo further medical examinations, so that the user can take timely measures to reduce health risks and improve the effectiveness of overall health management; when the health risk assessment value is less than the safety threshold, the wearable device does not send a warning notification, and continues to monitor and record human bioimpedance data for long-term health trend analysis.

[0084] In summary, the real-time health monitoring and intelligent diagnosis method applied to wearable devices has been completed.

[0085] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0087] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A real-time health monitoring and intelligent diagnosis method for wearable devices, characterized in that: The following steps are involved: S1: Collect human bioelectrical impedance data and construct a spectrum data matrix; preprocess the spectrum data matrix to obtain a standardized spectrum matrix; extract dynamic spectrum features from the standardized spectrum matrix through spectrum analysis method; S2: Build a dynamic spectrum model, analyze the dynamic spectrum characteristics, generate a health state vector; calculate the health risk assessment value based on the health state vector; Based on the health risk assessment value, analyze and evaluate the user's health status.

2. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 1, characterized in that: The S1 specifically includes: In the implementation process of preprocessing, the spectrum data matrix is ​​subjected to noise filtering to obtain a filtered spectrum data matrix; an adaptive spectrum enhancement algorithm is used to perform adaptive spectrum enhancement on the filtered spectrum data matrix to obtain an enhanced spectrum matrix; the specific formula of the adaptive spectrum enhancement algorithm is as follows: in, is the enhanced spectrum matrix; At time point t j and frequency f i Enhanced spectrum value under ; is the filtered spectral data matrix at time point t j and frequency f i Human bioelectrical impedance data below; is the filtered spectrum data matrix at frequency f i The standard deviation of human bioelectrical impedance data at all time points below; is the filtered spectrum data matrix at frequency f i is the mean of the human bioelectrical impedance data at all time points; exp is the exponential operation; i is the frequency index; n is the total number of frequencies.

3. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 2, characterized in that: The S1 specifically includes: The enhanced spectrum matrix is ​​processed by data normalization, and the standardized spectrum matrix is ​​generated by combining exponential transformation and normalized fraction.

4. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 3 is characterized in that: The S1 specifically includes: An exponential weight factor is introduced to perform exponential weighting on the standardized spectrum values ​​in the standardized spectrum matrix to obtain dynamic spectrum feature components; based on the dynamic spectrum feature components, dynamic spectrum features are constructed.

5. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 1, characterized in that: The S2 specifically includes: The dynamic spectrum model parameters are introduced to construct a dynamic spectrum model. The health state vector is obtained by combining the dynamic spectrum characteristics of multiple frequency bands. The specific calculation formula is as follows: S(t j )=log(||tanh(W·F(t j ))||2+exp(V·sin(F(t j )))) Among them, W and V are the parameters of the dynamic spectrum model, W is the linear weight matrix, and V is the nonlinear weight matrix; S(t j ) is the time point t j The health state vector of ||·||2 is the 2-norm; tanh is the hyperbolic tangent function; sin is the sine function; F(t j ) is the time point t j Dynamic spectrum characteristics.

6. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 5, characterized in that: The S2 specifically includes: A nonlinear dynamic spectrum optimization algorithm is introduced, and an optimization function is defined to optimize the dynamic spectrum model parameters. The specific definition of the optimization function is as follows: Where L is the target value of the optimization function; It's at the time A historical health state vector calculated based on historical dynamic spectrum characteristics; It's at the time Historical dynamic spectrum characteristics; Indicates a point in history; is the spectrum feature mapping value, which is obtained based on the historical dynamic spectrum features using the regression model; is the square operation of the two-norm; is the regularization coefficient.

7. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 6, characterized in that: The S2 specifically includes: Based on the objective value of the optimization function, the dynamic spectrum model parameters are optimized and updated using the gradient descent method in combination with the improved update formula to obtain updated dynamic spectrum model parameters; based on the updated dynamic spectrum model parameters, the health state vector is updated.

8. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 7, characterized in that: The S2 specifically includes: Based on the health state vector, the time decay factor is introduced to predict the health state vector at a future time point. The specific formula is as follows: in, At a future time The health status vector of is the prediction step length, which means from the current time point t j The time step for future prediction; α is the current health status weight; β is the influence weight of historical dynamic spectrum characteristics; γ is the time attenuation factor; F(t j -τ) is at time point t j -τ dynamic spectrum characteristics; τ is the lag step, which means from the current time point t j The number of time steps to trace back.

9. The real-time health monitoring and intelligent diagnosis method for wearable devices according to claim 8, characterized in that: The S2 specifically includes: Based on the health status vector at a future time point, the health risk assessment value is calculated; the health risk assessment value is used to analyze and evaluate the user's health status. When the health risk assessment value is greater than or equal to the preset safety threshold, it indicates that the user's current health status has potential risks, and a real-time early warning notification is sent to the user; when the health risk assessment value is less than the safety threshold, no early warning notification is sent, and monitoring and recording continue.

Images