Intelligent wearing health monitoring method and system based on biological impedance analysis

Abstract

The invention discloses an intelligent wearing health monitoring method and system based on biological impedance analysis, and relates to the technical field of biomedical engineering. The method comprises the following steps: constructing an original impedance spectrum of a biological tissue, and synchronously collecting an original impedance signal; a two-channel impedance model is constructed, and quantitative evaluation of the electrode contact quality is achieved; constructing a time-varying transfer function to obtain a de-noising impedance time-frequency signal; establishing a Cole impedance model, extracting characteristic parameters through a tensor decomposition method, and constructing a dynamic tissue parameter matrix; physiological feature vectors are generated through a multi-layer perceptron; constructing a metabolic impedance coupling differential equation to obtain a body component estimation vector, and realizing dynamic tissue parameter weight distribution and sensitivity adjustment; and carrying out grading evaluation on the calibrated body component vector. The stability and adaptability of the system are improved by optimizing impedance measurement, constructing a high-dimensional tissue parameter model, dynamically estimating body components, enhancing the interpretability of health assessment, cooperatively processing physiological state parameters and fusing multi-source information.

Description

Technical Field

[0001] The present invention relates to the field of biomedical engineering technology, and in particular to a smart wearable health monitoring method and system based on bioimpedance analysis. Background Art

[0002] Bioelectrical impedance analysis (BIA) is a noninvasive measurement technology based on the response characteristics of human tissue to electrical current. It is widely used in body composition measurement, nutritional assessment, and health management. Traditional BIA methods often use single- or dual-frequency signals to simplify the modeling of total body impedance. This makes it difficult to accurately reflect the specific response characteristics of different tissues, such as muscle and fat, under multi-frequency conditions. Furthermore, factors such as poor electrode contact and human movement can easily introduce significant artifacts, affecting the stability and accuracy of impedance measurements.

[0003] In recent years, with the development of wearable devices, multi-frequency excitation, real-time signal processing, and artificial intelligence modeling have become important means to improve the accuracy of BIA measurements. Some studies have introduced the Cole model to parametrically model tissue electrical response characteristics and combined it with machine learning methods to improve body composition estimation. However, existing methods generally suffer from the following shortcomings: a lack of dynamic assessment mechanisms for electrode quality, making it difficult to effectively suppress motion artifacts in mobile scenarios; and a lack of high-dimensional modeling and dynamic extraction capabilities for tissue impedance parameters, which limits the precise quantification of health status. Summary of the Invention

[0004] Based on the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide a smart wearable health monitoring method and system based on bioimpedance analysis to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: a smart wearable health monitoring method based on bioimpedance analysis, comprising:

[0006] Acquire multi-frequency human electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously collect the original impedance signal in the time domain;

[0007] A dual-channel impedance model is constructed based on the original impedance spectrum to dynamically model the skin electrode contact artifact, thereby achieving a quantitative assessment of the electrode contact quality.

[0008] The inertial measurement unit information is introduced to construct a time-varying transfer function, and the motion artifacts are suppressed in the frequency domain to obtain the denoised impedance time-frequency signal.

[0009] The Cole impedance model was established, and the impedance characteristic parameters of muscle and fat tissues were extracted by tensor decomposition method to construct a dynamic tissue parameter matrix.

[0010] The dynamic tissue parameters are fused with the user's static features and input to generate physiological feature vectors through a multi-layer perceptron.

[0011] Based on the physiological eigenvector, a metabolic impedance coupled differential equation is constructed to obtain the body composition estimation vector, and the hyperbolic tangent activation function is used to achieve dynamic tissue parameter weight distribution and sensitivity adjustment.

[0012] The body composition estimation vector is calibrated to output a body composition vector, and the calibrated body composition vector is graded and evaluated.

[0013] The present invention is further configured such that the original impedance spectrum is a representation of the original impedance signal in the frequency domain, and the specific expression of the original impedance spectrum is: Z raw (f k ), where f k is the kth frequency point;

[0014] The original impedance signal refers to the electrical impedance data measured at a specific time point and a specific frequency. The specific expression of the original impedance signal is: Z raw (f k ,t real ), where t real is the sampling time, f k is the kth frequency point;

[0015] Based on the original impedance spectrum, a dual-channel model is constructed to extract the pure biological impedance spectrum related to the intrinsic characteristics of physiological tissue. The construction logic of the dual-channel model is as follows: is the original impedance spectrum from measurement channel A, is the original impedance spectrum from measurement channel B, Z b (f k ) is the intrinsic impedance component of biological tissue, and is the coupling impedance caused by the skin electrode contact interface;

[0016] Perform differential operation on the signals of channels A and B to eliminate the bioimpedance term Z b (f k ), extract the contact impedance differential spectrum. The calculation logic for extracting the contact impedance differential spectrum is: ΔZ c (f k ) is the contact impedance difference spectrum, is the contact impedance spectrum of channel A, is the contact impedance spectrum of channel B, f k is the kth frequency point, κ is the skin dielectric loss factor, θ skin is the skin polarization angle;

[0017] When ΔZ c (f k ) value is within the preset threshold range, indicating that the electrode contact quality is good and there is no abnormality in the signal measurement;

[0018] When ΔZ c (f k ) value is higher or lower than the preset threshold range, and when κ exceeds the preset threshold, it indicates that the electrode contact quality is poor, resulting in measurement error, and an early warning signal is issued to inform the user.

[0019] The present invention is further configured such that the construction logic of the time-varying transfer function is: H(f k ,t real ) is the time-varying transfer function, Z raw (f k ,t real ) is the original impedance signal, IMU (t real ) is the dynamic response strength signal from the inertial measurement unit, τ motion is the motion delay response time, f k is the kth frequency point, t real is the sampling time, j is the imaginary unit;

[0020] Based on the time-varying transfer function H(f k ,t real )Calculate the modulus of the transfer time-varying function |H(f k ,t real )|, according to the transfer time-varying function modulus|H(f k ,t real )|Build a spectrum suppression mask. The construction logic of the spectrum suppression mask is: M(f k ,t real )=1-|H(f k ,t real )|,M(f k ,t real ) is the spectrum suppression mask;

[0021] Perform short-time Fourier transform on the original impedance signal, map the original impedance signal to the frequency domain, and combine it with the spectrum suppression mask to obtain the weighted spectrum signal. The calculation logic of the weighted spectrum signal is: is the weighted spectrum signal;

[0022] Perform inverse short-time Fourier transform on the weighted spectrum signal to obtain the denoised impedance time-frequency signal after suppressing motion artifacts. The calculation logic of the denoised impedance time-frequency signal is: Z bio (fk ,t real ) is the denoised impedance time-frequency signal.

[0023] The present invention is further configured such that the construction logic of the Cole impedance model is: i∈{m,f},Z i (f) is the impedance of the i-th type of organization at frequency f, R ∞,i is the high-frequency limit impedance of the i-th type of organization, R 0,i is the low-frequency limit impedance of the i-th type of tissue, τ i is the relaxation time constant of the i-th tissue, φ i is the phase delay factor of the i-th tissue, j is the imaginary unit, i is the index variable, m is muscle, and f is fat;

[0024] The denoised impedance time-frequency signal is reorganized into a four-dimensional impedance tensor data structure according to the time series, tissue type, and Cole impedance model parameter dimensions to form a unified impedance parameter tensor input format. The specific dimensional expression of the four-dimensional impedance tensor is: Z∈C K×2×4×T , K is the number of frequency points, 2 is the number of tissue types, 4 is the number of Cole impedance model parameters, and T is the number of time points. The index expression of each element in the four-dimensional impedance tensor is: Z[f,o,p,t], f is the corresponding frequency, o is the tissue, p is the Cole impedance model parameter, and t is the time point index. Tissue types include m and f, and Cole impedance model parameters include R0, R ∞ , τ and φ, where R0 is the low-frequency resistance, R ∞ is the high-frequency resistance, τ is the relaxation time constant, and φ is the phase delay factor;

[0025] Tucker decomposition is used to reduce the dimension of the four-dimensional impedance tensor and extract its features. The specific decomposition logic is: Z≈G×1U freq ×2U tissue ×3U param ×4U time ,in, is the core tensor, U freq ∈R K×k1 、U tissue ∈R 2×2 、U param ∈R 4×2 and is a factor matrix;

[0026] Project the core tensor G along the tissue parameter dimension and factor matrix to the physical parameter space to obtain the tissue parameter tensor and extract the characteristic parameters of muscle and fat tissue, that is, execute Get the organizational parameter tensor P∈R 4×2×T , 4 is the number of Cole impedance model parameters, 2 is the number of tissue categories, and T is the number of time points; the specific index expression of the tissue parameter tensor is: P[p,i,t], where p∈{R0,R ∞ ,τ,φ} is the parameter dimension of Cole impedance model, i∈{m,f} is the tissue type, and t is the time point index;

[0027] Muscle activity index and fat metabolic rate were calculated based on the temporal gradient changes of tissue parameter tensors.

[0028] The present invention is further configured such that the calculation logic of the muscle activity index is: MAI(t) is the muscle activity index, Δτ m is the relaxation time difference of muscle tissue, is the spatial gradient of the muscle tissue phase angle, Δt is the time step, and the muscle tissue relaxation time difference Δτ m The calculation logic is: Δτ m =τ m (t)-τ m (t-1), τ m (t) is the relaxation time constant of the muscle tissue at the current time point t, τ m (t-1) is the relaxation time constant of the muscle tissue at the previous time point t-1, and the spatial gradient of the muscle tissue phase angle The calculation logic is: φ m (t) is the phase angle of the muscle tissue at time point t, φ m (t-1) is the phase angle of the muscle tissue at the previous time point t-1, φ m (t+1) is the phase angle of the muscle tissue at the next time point t+1;

[0029] The calculation logic of fat metabolic rate is: FMR(t) is fat metabolic rate, R 0,f (t) is the low-frequency resistance of fat tissue, R ∞,f (t) is the high-frequency resistance of fat tissue, R 0,m (t) is the low-frequency resistance of muscle tissue, Δφ f is the time change rate of the phase angle of fat tissue, the time change rate of the phase angle of fat tissue Δφ f The calculation logic is: Δφ f =φ f (t)-φ f (t-Δt), φ f (t) is the phase angle of fat tissue at the current time point t, φ f(t-Δt) is the phase angle of adipose tissue at the previous time point t-Δt;

[0030] The muscle activity index, fat metabolic rate, tissue relaxation time constant and tissue phase angle are fused as the dynamic tissue parameter matrix output. The expression of the dynamic tissue parameter matrix is: T(t) is the dynamic tissue parameter matrix, where

[0031] The present invention is further configured to perform multimodal fusion of dynamic tissue parameters and user static features to construct a physiological feature vector, and the specific steps include:

[0032] The dynamic organizational parameter matrix and the user's static features are concatenated to form a joint vector. The feature concatenation logic is as follows: is the joint vector, u is the user's static feature, age is age, sex is gender, height is height, weight is weight, and d is the user's static feature dimension;

[0033] Construct a multi-layer perceptron network and combine the vector As input, a physiological feature vector is output, and the expression of the physiological feature vector is: F = [f1, f2, f3, f4, f5, f6], where F is the physiological feature vector, f1 is the muscle metabolic state, f2 is the fat distribution coefficient, f3 is the hydration balance index, f4 is the skin conductivity, f5 is the respiratory variability index, and f6 is the heart rate variability index.

[0034] The present invention is further configured such that the construction logic of the metabolic impedance coupled differential equation is: is the rate of change of the body composition vector y(t) with respect to time t, f(t,y) is the function of the body composition vector y(t) and time t, F(t)∈R 6 is the physiological feature vector at time t, A∈R 3×6 is the coupling weight matrix, B∈R 3×4 is the sensitivity matrix, W∈R 6×6 is the weight mapping matrix, T(t)∈R 4 is the dynamic organizational parameter at time t;

[0035] The fourth-order Runge-Kutta numerical solution is used to solve the metabolic impedance coupled differential equation. The calculation logic of the solution is: Among them, y t ∈R 3 is the estimated body composition vector at time t, corresponding to body fat percentage, muscle mass and total body water respectively, Δt is the time step, k1, k2, k3 and k4 are the estimated values ​​of the intermediate step length;

[0036] Upon detection When it is greater than the preset threshold θ, the sensitivity matrix is ​​updated and adjusted, and the update logic is: γ is the attenuation coefficient.

[0037] The present invention is further configured to: t The body composition vector is output after calibration. The specific steps include:

[0038] The body composition estimation vector is normalized by the Sigmoid mapping function. The processing logic is as follows: y * (t) is the normalized volume component vector, BF * is the original predicted value of body fat percentage, MM * is the raw predicted value of muscle mass, TBW * is the original predicted value of total body water;

[0039] The normalized output volume component vector y * (t) is further linearly mapped to a range with physiological significance: BF final is the body fat percentage and MM after mapping final is the mapped muscle mass, TBW final is the total body water after mapping, BF min and BF max They are the physiological upper and lower limits of body fat percentage, MM min and MM max They are the physiological upper and lower limits of muscle mass, TBW min and TBW max They are the physiological upper and lower limits of total body water, respectively.

[0040] The present invention is further configured to input the mapped body composition vector into a preset health threshold discrimination module, analyze the degree of deviation between the body fat percentage, muscle mass, and total body water parameters and the corresponding preset physiological acceptable ranges, and then perform health status classification and output secondary health status labels, wherein the labels include normal state, marginal state, and abnormal state;

[0041] A graded warning response is triggered based on the health status tag. When the judgment result is a marginal state, a first-level warning is triggered, and the device is driven to vibrate to prompt the user's attention; when the judgment result is an abnormal state, a second-level warning is triggered, and the abnormal information is synchronously sent to the associated medical service terminal.

[0042] The present invention also provides a smart wearable health monitoring system based on bioimpedance analysis, the system comprising:

[0043] Multi-frequency impedance acquisition module: used to obtain multi-frequency human body electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously acquire the original impedance signal in the time domain;

[0044] Electrode quality assessment module: used to build a dual-channel impedance model based on the original impedance spectrum to model skin electrode contact artifacts and achieve quantitative assessment of electrode quality;

[0045] De-noising impedance time-frequency signal acquisition module: used to introduce inertial measurement unit information to construct a time-varying transfer function, suppress motion artifacts in the frequency domain, and obtain a de-noising impedance time-frequency signal;

[0046] Tissue impedance modeling and tensor feature extraction module: used to establish the Cole impedance model, extract the impedance characteristic parameters of muscle and fat tissues through tensor decomposition method, and construct a dynamic tissue parameter matrix;

[0047] Physiological feature vector generation module: used to fuse dynamic tissue parameters with user static features and generate physiological feature vectors through a multi-layer perceptron;

[0048] Metabolic impedance coupling modeling module: used to construct metabolic impedance coupling differential equations based on physiological eigenvectors to obtain body composition estimation vectors, and use hyperbolic tangent activation functions to achieve dynamic tissue parameter weight allocation and sensitivity adjustment;

[0049] Body composition calibration and grading evaluation module: used to calibrate the body composition estimation vector to output the body composition vector, and perform grading evaluation on the calibrated body composition vector.

[0050] The present invention provides a smart wearable health monitoring method and system based on bioimpedance analysis. The method obtains multi-frequency human electrical response data through multi-frequency excitation signals, constructs the original impedance spectrum of biological tissue in the frequency domain, and synchronously collects the original impedance signal in the time domain; constructs a dual-channel impedance model based on the original impedance spectrum, dynamically models the skin electrode contact artifact, and thus realizes the quantitative evaluation of the electrode contact quality; introduces inertial measurement unit information to construct a time-varying transfer function, suppresses motion artifacts in the frequency domain, and obtains a denoised impedance time-frequency signal; establishes a Cole impedance model, extracts the impedance characteristic parameters of muscle and fat tissue through a tensor decomposition method, and constructs a dynamic tissue parameter matrix; fuses the dynamic tissue parameters with the user's static features, and generates a physiological feature vector through a multi-layer perceptron; constructs a metabolic impedance coupled differential equation based on the physiological feature vector to obtain a body composition estimation vector, and uses a hyperbolic tangent activation function to realize dynamic tissue parameter weight distribution and sensitivity adjustment; calibrates the body composition estimation vector to output a body composition vector, and performs a graded evaluation on the calibrated body composition vector, resulting in the following beneficial effects:

[0051] 1. Improved impedance measurement accuracy and robustness: By building a dual-channel impedance model and introducing an electrode contact artifact modeling mechanism, dynamic assessment and quantitative determination of electrode quality are achieved, significantly improving the reliability of measurement data.

[0052] 2. Effectively suppress motion artifact interference: Inertial measurement unit information is introduced to construct a time-varying transfer function, and combined with a spectrum suppression mask for short-time Fourier domain weighting to achieve frequency domain elimination of motion artifacts and ensure signal stability in dynamic scenarios;

[0053] 3. Achieve high-dimensional modeling and compressed representation of tissue impedance parameters: By constructing the Cole impedance model and using the Tucker tensor decomposition method to extract tissue impedance characteristic parameters, a unified dynamic tissue parameter tensor structure is constructed, improving the ability to refine the expression of muscle and fat tissue parameters.

[0054] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without inventive efforts. In the drawings:

[0056] Figure 1 This is a flowchart of a smart wearable health monitoring method based on bioimpedance analysis according to an exemplary embodiment of the present invention;

[0057] Figure 2 The figure is a schematic structural diagram of a smart wearable health monitoring system based on bioimpedance analysis, showing an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following describes the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art will readily appreciate the other advantages and benefits of the present invention from the disclosure herein. The present invention may also be implemented or applied through various other specific embodiments, and the various details in this specification may be modified or altered based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are intended only to illustrate the present invention and are not intended to limit the scope of protection of the present invention.

[0059] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0060] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0061] Example 1

[0062] A smart wearable health monitoring method based on bioimpedance analysis, such as Figure 1 Shown, including:

[0063] Acquire multi-frequency human electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously collect the original impedance signal in the time domain;

[0064] A dual-channel impedance model is constructed based on the original impedance spectrum to dynamically model the skin electrode contact artifact, thereby achieving a quantitative assessment of the electrode contact quality.

[0065] The inertial measurement unit information is introduced to construct a time-varying transfer function, and the motion artifacts are suppressed in the frequency domain to obtain the denoised impedance time-frequency signal.

[0066] The Cole impedance model was established, and the impedance characteristic parameters of muscle and fat tissues were extracted by tensor decomposition method to construct a dynamic tissue parameter matrix.

[0067] The dynamic tissue parameters are fused with the user's static features and input to generate physiological feature vectors through a multi-layer perceptron.

[0068] Based on the physiological eigenvector, a metabolic impedance coupled differential equation is constructed to obtain the body composition estimation vector, and the hyperbolic tangent activation function is used to achieve dynamic tissue parameter weight distribution and sensitivity adjustment.

[0069] The body composition estimation vector is calibrated to output a body composition vector, and the calibrated body composition vector is graded and evaluated.

[0070] The present invention is further configured such that the original impedance spectrum is a representation of the original impedance signal in the frequency domain, and the specific expression of the original impedance spectrum is: Z raw (f k ), where f kis the kth frequency point;

[0071] The original impedance signal refers to the electrical impedance data measured at a specific time point and a specific frequency. The specific expression of the original impedance signal is: Z raw (f k ,t real ), where t real is the sampling time, f k is the kth frequency point;

[0072] Based on the original impedance spectrum, a dual-channel model is constructed to extract the pure biological impedance spectrum related to the intrinsic characteristics of physiological tissue. The construction logic of the dual-channel model is as follows: is the original impedance spectrum from measurement channel A, is the original impedance spectrum from measurement channel B, Z b (f k ) is the intrinsic impedance component of biological tissue, and is the coupling impedance caused by the skin electrode contact interface;

[0073] Perform differential operation on the signals of channels A and B to eliminate the bioimpedance term Z b (f k ), extract the contact impedance differential spectrum. The calculation logic for extracting the contact impedance differential spectrum is: ΔZ c (f k ) is the contact impedance difference spectrum, is the contact impedance spectrum of channel A, is the contact impedance spectrum of channel B, f k is the kth frequency point, κ is the skin dielectric loss factor, θ skin is the skin polarization angle;

[0074] When ΔZ c (f k ) value is within the preset threshold range, indicating that the electrode contact quality is good and there is no abnormality in the signal measurement;

[0075] When ΔZ c (f k ) value is higher or lower than the preset threshold range, and when κ exceeds the preset threshold, it indicates that the electrode contact quality is poor, resulting in measurement error, and an early warning signal is issued to inform the user;

[0076] Specifically, a dual-channel impedance measurement model was constructed to accurately extract the intrinsic impedance spectrum of physiological tissue from the interfered original electrical impedance signal. The dual-channel structure uses the contact difference between the two electrodes to construct a redundant information channel, making the contact artifact identifiable, thereby improving the measurement accuracy; ΔZc (f k ) is expressed as a function of frequency and skin polarization characteristics to explain the contact impedance difference spectrum ΔZ c (f k ) indicates that the difference is mainly determined by the skin polarization characteristics and frequency response; according to ΔZ c (f k ) value to judge the threshold value and evaluate the contact quality of the electrode. c (f k ) value is within the preset threshold range, indicating good contact. c (f k ) value exceeds the preset threshold range and κ exceeds the preset threshold, indicating poor electrode contact and an early warning. κ is used to reflect the ability of skin tissue to store or dissipate energy and is affected by skin thickness, water content, and temperature, with a value range of [0.01, 1]. Differential spectrum modeling is used to reveal artifacts caused by electrode contact problems, making up for the shortcoming that traditional single-channel cannot judge contact quality.

[0077] The present invention is further configured such that the construction logic of the time-varying transfer function is: H(f k ,t real ) is the time-varying transfer function, Z raw (f k ,t real ) is the original impedance signal, IMU (t real ) is the dynamic response strength signal from the inertial measurement unit, τ motion is the motion delay response time, f k is the kth frequency point, t real is the sampling time, j is the imaginary unit;

[0078] Based on the time-varying transfer function H(f k ,t real )Calculate the modulus of the transfer time-varying function |H(f k ,t real )|, according to the transfer time-varying function modulus|H(f k ,t real )|Build a spectrum suppression mask. The construction logic of the spectrum suppression mask is: M(f k ,t real )=1-|H(f k ,t real )|,M(f k ,t real ) is the spectrum suppression mask;

[0079] Perform short-time Fourier transform on the original impedance signal, map the original impedance signal to the frequency domain, and combine it with the spectrum suppression mask to obtain the weighted spectrum signal. The calculation logic of the weighted spectrum signal is: is the weighted spectrum signal;

[0080] Perform inverse short-time Fourier transform on the weighted spectrum signal to obtain the denoised impedance time-frequency signal after suppressing motion artifacts. The calculation logic of the denoised impedance time-frequency signal is: Z bio (f k ,t real ) is the denoised impedance time-frequency signal;

[0081] Specifically, the time-varying transfer function is used to characterize the response relationship between human tissue impedance and motion state at different frequencies and times. The time-varying transfer function correction method constructed with the assistance of IMU information is used to suppress motion artifact errors caused by user micro-movements and body posture deviations. IMU (t real ) is the comprehensive strength of the three-axis acceleration and angular velocity from the inertial measurement unit, which is used to characterize the user at time t real The motion state of IMU(t real ) through the three-axis acceleration a x (t real ), a y (t real ) and a z (t real ) and the three-axis angular velocity w x (t real ), w y (t real ) and w z (t real ) is composed of: IMU(t real )=α·||a(t real )||+β·||ω(t real )||, where α and β are weighting coefficients, α and β are used to adjust the influence ratio of translation and rotation, with a value range of [0,1], and the sum of α and β is 1; a spectrum suppression mask is constructed to suppress the frequency band in the signal that is more affected by motion interference, M(f k ,t real ) value is closer to 1, indicating that the signal at that frequency point is cleaner. k ,t real) value is closer to 0, indicating that there is greater interference at this frequency point and that it needs to be suppressed. Short-time Fourier transform is a common tool for converting signals from the time domain to the frequency domain. It analyzes the local characteristics of the signal in time and frequency. Through short-time Fourier transform, the original impedance signal is mapped to the time-frequency plane in the frequency domain, making the time-frequency characteristics of the signal clearer. Using the spectrum suppression mask M(f k ,t real ) weights the time-frequency signal to suppress the frequency components affected by motion artifacts. The goal of this operation is to enhance the physiological characteristics of the signal by reducing the influence of motion; the weighted spectral signal is subjected to inverse short-time Fourier transform to convert it from the frequency domain back to the time domain and recover the denoised impedance signal. This process eliminates the influence of motion artifacts on the signal and obtains a cleaner physiological impedance signal. The denoised impedance time-frequency signal Z bio (f k ,t real ) retains the information about physiological tissue in the signal, while motion artifacts and noise are effectively removed; by combining short-time Fourier transform and spectral suppression mask, artifacts caused by motion are effectively removed, thereby obtaining a more stable and real physiological signal and improving the accuracy of impedance measurement.

[0082] The present invention is further configured such that the construction logic of the Cole impedance model is: i∈{m,f},Z i (f) is the impedance of the i-th type of organization at frequency f, R ∞,i is the high-frequency limit impedance of the i-th type of organization, R 0,i is the low-frequency limit impedance of the i-th type of tissue, τ i is the relaxation time constant of the i-th tissue, φ i is the phase delay factor of the i-th tissue, j is the imaginary unit, i is the index variable, m is muscle, and f is fat;

[0083] The denoised impedance time-frequency signal is reorganized into a four-dimensional impedance tensor data structure according to the time series, tissue type, and Cole impedance model parameter dimensions to form a unified impedance parameter tensor input format. The specific dimensional expression of the four-dimensional impedance tensor is: Z∈C K×2×4×T , K is the number of frequency points, 2 is the number of tissue types, 4 is the number of Cole impedance model parameters, and T is the number of time points. The index expression of each element in the four-dimensional impedance tensor is: Z[f,o,p,t], f is the corresponding frequency, o is the tissue, p is the Cole impedance model parameter, and t is the time point index. Tissue types include m and f, and Cole impedance model parameters include R0, R ∞ , τ and φ, where R0 is the low-frequency resistance, R ∞ is the high-frequency resistance, τ is the relaxation time constant, and φ is the phase delay factor;

[0084] Tucker decomposition is used to reduce the dimension of the four-dimensional impedance tensor and extract its features. The specific decomposition logic is: Z≈G×1U freq ×2U tissue ×3U param ×4U time ,in, is the core tensor, U tissue ∈R 2×2 、U param ∈R 4×2 and is a factor matrix;

[0085] Project the core tensor G along the tissue parameter dimension and factor matrix to the physical parameter space to obtain the tissue parameter tensor and extract the characteristic parameters of muscle and fat tissue, that is, execute Get the organizational parameter tensor P∈R 4×2×T , 4 is the number of Cole impedance model parameters, 2 is the number of tissue categories, and T is the number of time points; the specific index expression of the tissue parameter tensor is: P[p,i,t], where p∈{R0,R ∞ ,τ,φ} is the parameter dimension of Cole impedance model, i∈{m,f} is the tissue type, and t is the time point index;

[0086] The muscle activity index and fat metabolic rate are calculated based on the time gradient changes of tissue parameter tensors;

[0087] Specifically, the Cole impedance model is used to describe the electrical impedance characteristics of different tissue types at different frequencies. When the frequency increases, the charge in the tissue cannot respond to the external electric field in a timely manner, resulting in a decrease in impedance. Tissue types include muscle and fat. These two types of tissue are significantly distinguishable in bioimpedance spectrum analysis and are highly correlated with the assessment of body composition parameters such as body fat percentage, muscle mass, and total body water. Therefore, they are selected as key analysis objects; τ i It is used to describe the polarization response rate of charge in tissues, indicating the speed at which tissues transition from low frequency to high frequency. The smaller the value, the faster the response. i ∈[10 -4 ,10 -2 ]s;φ iIt is used to describe the frequency distribution characteristics of the tissue polarization process and reflect the breadth of the distribution of various relaxation mechanisms in the tissue. The smaller the value, the more concentrated the relaxation behavior, and the larger the value, the wider the heterogeneity. The value range is [0.4, 1]. In order to achieve multi-tissue, multi-frequency, multi-parameter, and time-continuous electrical impedance feature expression, the denoised impedance time-frequency signal is structured and reorganized to construct a unified four-dimensional impedance tensor data structure. The four-dimensional impedance tensor takes the time series as the main axis and integrates the frequency, tissue type, and Cole model parameter dimensions to form a complete and compact feature input format. The specific format is: Z∈C K×2×4×T ; Tucker decomposition is performed on the four-dimensional impedance tensor to achieve parameter dimensionality reduction and tissue feature extraction. G is the core tensor, which retains the main feature information and plays a dimensionality reduction role in the tensor structure. U freq ∈R K×k1 is the frequency factor matrix, U tissue ∈R 2×2 is the tissue type factor matrix, U param ∈R 4×2 is the Cole model parameter factor matrix, is the time factor matrix; k1 and k4 represent the number of retained principal components in the frequency dimension and time dimension respectively, and their values ​​are automatically determined by the cumulative explained variance, for example, the minimum dimension required when the cumulative contribution rate reaches 95% energy, to ensure the integrity and compactness of information retention after dimensionality reduction; the organizational parameter tensor P∈R 4×2×T It is used to express the changes of the four types of Cole impedance model parameters of two types of tissues at each time point T. The two types of tissues include muscle and fat. The four types of Cole impedance model parameters include R0, R ∞ ,τ,φ, and the tissue parameter tensor is used for the subsequent calculation of muscle activity index and fat metabolic rate.

[0088] The present invention is further configured such that the calculation logic of the muscle activity index is: MAI(t) is the muscle activity index, Δτ m is the relaxation time difference of muscle tissue, is the spatial gradient of the muscle tissue phase angle, Δt is the time step, and the muscle tissue relaxation time difference Δτ m The calculation logic is: Δτ m =τ m (t)-τ m (t-1), τ m (t) is the relaxation time constant of the muscle tissue at the current time point t, τ m (t-1) is the relaxation time constant of the muscle tissue at the previous time point t-1, and the spatial gradient of the muscle tissue phase angle The calculation logic is: φm (t) is the phase angle of the muscle tissue at time point t, φ m (t-1) is the phase angle of the muscle tissue at the previous time point t-1, φ m (t+1) is the phase angle of the muscle tissue at the next time point t+1;

[0089] The calculation logic of fat metabolic rate is: FMR(t) is fat metabolic rate, R 0,f (t) is the low-frequency resistance of fat tissue, R ∞,f (t) is the high-frequency resistance of fat tissue, R 0,m (t) is the low-frequency resistance of muscle tissue, Δφ f is the time change rate of the phase angle of fat tissue, the time change rate of the phase angle of fat tissue Δφ f The calculation logic is: Δφ f =φ f (t)-φ f (t-Δt), φ f (t) is the phase angle of fat tissue at the current time point t, φ f (t-Δt) is the phase angle of adipose tissue at the previous time point t-Δt;

[0090] The muscle activity index, fat metabolic rate, tissue relaxation time constant and tissue phase angle are fused as the dynamic tissue parameter matrix output. The expression of the dynamic tissue parameter matrix is: T(t) is the dynamic tissue parameter matrix, where

[0091] Specifically, the muscle activity index is calculated based on the dynamic change rate of the muscle impedance parameter, which directly reflects the excitation and contraction process of the muscle. During activity, the electrical properties of the muscle cell membrane, electrolyte flow, and changes in muscle fiber structure will cause fluctuations in the impedance parameter, thereby increasing MAI. When MAI increases, it means that the muscle is actively contracting and the electrophysiological activity is enhanced. When MAI decreases, it means that the muscle is in a state of rest or slight activity; Δτ m Indicates the rate of change of the capacitive response of muscle tissue, a larger value indicates a rapid change process; It is used to measure the intensity and direction of changes in the electrical polarization behavior of muscle tissue in a time series; the fat metabolic rate can capture the dynamic changes in the structure and electrophysiological response of adipose tissue, reflect the degree of metabolic activity, and quantify the metabolic activity and electrophysiological change rate of adipose tissue; Δφ fIt represents the time difference of the phase angle of fat tissue and reflects the rate of change of electrical characteristics. The dynamic tissue parameter matrix T(t) is used to perform structured extraction and dynamic expression of key parameters with strong physiological significance implied in the impedance spectrum. As a high-level representation feature of the electrophysiological state of tissue, the construction of the dynamic tissue parameter matrix T(t) can intuitively reflect tissue activity, metabolic efficiency, response speed and structural stability, so that the impedance data has a clear physiological meaning and improves the physiological interpretability of the impedance data.

[0092] The present invention is further configured to perform multimodal fusion of dynamic tissue parameters and user static features to construct a physiological feature vector, and the specific steps include:

[0093] The dynamic organizational parameter matrix and the user's static features are concatenated to form a joint vector. The feature concatenation logic is as follows: is the joint vector, u is the user's static feature, age is age, sex is gender, height is height, weight is weight, and d is the user's static feature dimension;

[0094] Construct a multi-layer perceptron network and combine the vector As input, output is a physiological feature vector, the expression form of the physiological feature vector is: F = [f1, f2, f3, f4, f5, f6], F is the physiological feature vector, f1 is the muscle metabolic state, f2 is the fat distribution coefficient, f3 is the hydration balance index, f4 is the skin conductivity, f5 is the respiratory variability index, and f6 is the heart rate variability index;

[0095] Specifically, the purpose of multimodal fusion of dynamic tissue parameters and user static features is to combine structured tissue physiological information with unstructured user background data to improve the ability to characterize individual health characteristics; the order of feature splicing is based on the following physical and physiological basis: dynamic tissue parameters represent the physiological state of the user at a specific time point and can reflect the health status of muscle and fat tissue; static features are the basic physiological information of the user, which helps the model understand the user's body composition and physiological characteristics; by splicing feature vectors from different sources, the joint vector formed can provide comprehensive physiological information for the model and enhance the model's ability to predict user health; after standardization or normalization preprocessing of dynamic tissue parameters and user static features, the joint vector is input into a multi-layer perceptron network, including several fully connected layers and activation functions, for modeling nonlinear mapping relationships, and finally outputting a four-dimensional physiological feature vector. The physiological feature vector is a digital representation of the user's individual physiological state. Each component has a clear physiological meaning and application value, which will not be elaborated here due to existing technology. The muscle metabolic state reflects the health status of the muscle tissue and can provide an assessment of exercise intensity and recovery ability. The fat distribution coefficient measures the distribution of fat in the body. The hydration balance index is used to monitor the body's water status. Skin conductivity is related to sweat gland activity and autonomic nervous system regulation, reflecting the body's stress state. The respiratory variability index is used to reflect the time-varying fluctuation characteristics of respiratory frequency, depth, and rhythm. Poor respiratory function or abnormal respiratory variability leads to reduced body metabolic efficiency and affects fat metabolism. The heart rate variability index is used to reflect the individual's stress response and recovery ability. By constructing the physiological feature vector as described above, dynamic tissue changes and static individual characteristics are jointly modeled, making the characteristics closer to individual differences in people, rather than being based on the average value of a certain dimension.

[0096] The present invention is further configured such that the construction logic of the metabolic impedance coupled differential equation is: is the rate of change of the body composition vector y(t) with respect to time t, f(t,y) is the function of the body composition vector y(t) and time t, F(t)∈R 6 is the physiological feature vector at time t, A∈R 3×6 is the coupling weight matrix, B∈R 3×4 is the sensitivity matrix, W∈R 6×6 is the weight mapping matrix, T(t)∈R 4 is the dynamic organizational parameter at time t;

[0097] The fourth-order Runge-Kutta numerical solution is used to solve the metabolic impedance coupled differential equation. The calculation logic of the solution is: Among them, y t ∈R 3is the estimated body composition vector at time t, corresponding to body fat percentage, muscle mass and total body water respectively, Δt is the time step, k1, k2, k3 and k4 are the estimated values ​​of the intermediate step length;

[0098] Upon detection When it is greater than the preset threshold θ, the sensitivity matrix is ​​updated and adjusted, and the update logic is: γ is the attenuation coefficient;

[0099] Specifically, the body composition vector y(t) includes body fat percentage, muscle mass, and body water, which represent the components of the body. These values ​​change over time, reflecting changes in human health and physiological status. They are real body composition data, and the body composition estimation vector y t It is the estimated value of body composition obtained by model calculation, which is an inferred value based on physiological characteristics and dynamic tissue parameters. The difference between the body composition estimation vector and the body composition vector is that one is the predicted value of the model and the other is the actual measurement value. This difference is to reflect the relationship between the model estimate and the real physiological data. In the model calculation, the ultimate goal is to use the predictive ability of the model to obtain a dynamic estimate of body composition; the coupling weight matrix A∈R 3×6 Used to control the influence of physiological eigenvectors on body composition changes, the sensitivity matrix B∈R 3×4 Used to control the influence of dynamic tissue parameters T(t) on body composition changes, the weight mapping matrix W∈R 6×6 Used to describe the relationship between physiological characteristics and body composition; the values ​​of matrices A, B and W are learned through an optimization algorithm. Specifically, these matrices are updated by minimizing the loss function, such as the mean square error, on the training data. Gaussian distribution is used for initialization in the initial stage to ensure that their values ​​are within a reasonable range to facilitate the learning process of the model. The optimization process uses a gradient descent algorithm to ensure that the influence of physiological characteristics and dynamic tissue parameters on body composition changes is reasonably adjusted; k1, k2, k3 and k4 are obtained by calculating different predicted values ​​of body composition changes to improve the accuracy of the results, and the existing technology is not described in detail; γ is used to determine the degree of influence of the change in dynamic tissue parameters on the update of the sensitivity matrix B, and the value range is [0,1]; by introducing the metabolic impedance coupled differential equation, the dynamic changes of components such as body fat, muscle mass and water over time can be simulated, which can provide strong support for personalized health management and exercise program design.

[0100] The present invention is further configured to: t The body composition vector is output after calibration. The specific steps include:

[0101] The body composition estimation vector is normalized by the Sigmoid mapping function. The processing logic is as follows: y *(t) is the normalized volume component vector, BF * is the original predicted value of body fat percentage, MM * is the raw predicted value of muscle mass, TBW * is the original predicted value of total body water;

[0102] The normalized output volume component vector y * (t) is further linearly mapped to a range with physiological significance: BF final is the body fat percentage and MM after mapping final is the mapped muscle mass, TBW final is the total body water after mapping, BF min and BF max They are the physiological upper and lower limits of body fat percentage, MM min and MM max They are the physiological upper and lower limits of muscle mass, TBW min and TBW max These are the physiological upper and lower limits of total body water, respectively;

[0103] Specifically, the purpose of the body composition vector calibration process is to convert the predicted body composition estimates into actual body composition values ​​with physiological significance and explanatory power; the Sigmoid function compresses any real number to [0,1], making the output suitable for further conversion into physiological proportions; the normalized value is mapped to the physiologically valid interval of each body component through linear interpolation; the abstract estimated value output by the model is converted into a physiological indicator that the user can understand, which helps to generate personalized health reports.

[0104] The present invention is further configured to input the mapped body composition vector into a preset health threshold discrimination module, analyze the degree of deviation between the body fat percentage, muscle mass, and total body water parameters and the corresponding preset physiological acceptable ranges, and then perform health status classification and output secondary health status labels, wherein the labels include normal state, marginal state, and abnormal state;

[0105] Triggering graded warning responses based on health status tags: When the judgment result is a marginal state, a first-level warning is triggered, driving the device to vibrate to alert the user; when the judgment result is an abnormal state, a second-level warning is triggered, and the abnormal information is synchronously sent to the associated medical service terminal;

[0106] Specifically, the above processing describes a mechanism for automatically judging and warning health status based on body composition assessment results, with the aim of converting the test results into user-friendly graded feedback and response strategies; the normal state refers to the measurement values ​​of each body composition parameter such as body fat percentage, muscle mass and total body water falling within the preset normal physiological range, indicating that the user is in a healthy state; the marginal state refers to the measurement value of any body composition parameter, although it has not exceeded the abnormal threshold, but has deviated from the normal range and is within the preset allowable deviation interval, which may reflect potential physiological fluctuations or changes in health trends; the abnormal state refers to the measurement value of any body composition parameter exceeding the corresponding abnormal threshold range, prompting the user that there may be a higher health risk; the system adopts different levels of response according to the current label to timely remind the user when the physical signs deviate.

[0107] Example 2

[0108] See also Figure 2 , the exemplary smart wearable health monitoring system based on bioimpedance analysis includes:

[0109] Multi-frequency impedance acquisition module: used to obtain multi-frequency human body electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously acquire the original impedance signal in the time domain;

[0110] Electrode quality assessment module: used to build a dual-channel impedance model based on the original impedance spectrum to model skin electrode contact artifacts and achieve quantitative assessment of electrode quality;

[0111] De-noising impedance time-frequency signal acquisition module: used to introduce inertial measurement unit information to construct a time-varying transfer function, suppress motion artifacts in the frequency domain, and obtain a de-noising impedance time-frequency signal;

[0112] Tissue impedance modeling and tensor feature extraction module: used to establish the Cole impedance model, extract the impedance characteristic parameters of muscle and fat tissues through tensor decomposition method, and construct a dynamic tissue parameter matrix;

[0113] Physiological feature vector generation module: used to fuse dynamic tissue parameters with user static features and generate physiological feature vectors through a multi-layer perceptron;

[0114] Metabolic impedance coupling modeling module: used to construct metabolic impedance coupling differential equations based on physiological eigenvectors to obtain body composition estimation vectors, and use hyperbolic tangent activation functions to achieve dynamic tissue parameter weight allocation and sensitivity adjustment;

[0115] Body composition calibration and grading evaluation module: used to calibrate the body composition estimation vector to output the body composition vector, and perform grading evaluation on the calibrated body composition vector.

[0116] It should be noted that the smart wearable health monitoring system based on bioimpedance analysis provided in the above embodiment and the smart wearable health monitoring method based on bioimpedance analysis provided in the above embodiment belong to the same concept, wherein the specific manner in which each module and unit performs the operation has been described in detail in the method embodiment and will not be repeated here. In actual application, the smart wearable health monitoring system based on bioimpedance analysis provided in the above embodiment can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0117] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. 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 computer-readable storage medium. 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 a wired (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0118] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0119] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0120] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0121] Those skilled in the art will appreciate that the units and algorithm steps of each example 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel 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.

[0122] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0124] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0125] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0126] If the functions are implemented in the form of software functional units 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 the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0127] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A smart wearable health monitoring method based on bioimpedance analysis, characterized in that: include: Acquire multi-frequency human electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously collect the original impedance signal in the time domain; A dual-channel impedance model is constructed based on the original impedance spectrum to dynamically model the skin electrode contact artifact, thereby achieving a quantitative assessment of the electrode contact quality. The inertial measurement unit information is introduced to construct a time-varying transfer function, and the motion artifacts are suppressed in the frequency domain to obtain the denoised impedance time-frequency signal. The Cole impedance model was established, and the impedance characteristic parameters of muscle and fat tissues were extracted by tensor decomposition method to construct a dynamic tissue parameter matrix. The dynamic tissue parameters are fused with the user's static features and input to generate physiological feature vectors through a multi-layer perceptron. Based on the physiological eigenvector, a metabolic impedance coupled differential equation is constructed to obtain the body composition estimation vector, and the hyperbolic tangent activation function is used to achieve dynamic tissue parameter weight distribution and sensitivity adjustment. The body composition estimation vector is calibrated to output a body composition vector, and the calibrated body composition vector is graded and evaluated.

2. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 1, characterized in that: The original impedance spectrum is the representation of the original impedance signal in the frequency domain. The specific expression of the original impedance spectrum is: Z raw (f k ), where f k is the kth frequency point; The original impedance signal refers to the electrical impedance data measured at a specific time point and a specific frequency. The specific expression of the original impedance signal is: Z raw (f k ,t real ), where t real is the sampling time, f k is the kth frequency point; Based on the original impedance spectrum, a dual-channel model is constructed to extract the pure biological impedance spectrum related to the intrinsic characteristics of physiological tissue. The construction logic of the dual-channel model is as follows: is the original impedance spectrum from measurement channel A, is the original impedance spectrum from measurement channel B, Z b (f k ) is the intrinsic impedance component of biological tissue, and is the coupling impedance caused by the skin electrode contact interface; Perform differential operation on the signals of channels A and B to eliminate the bioimpedance term Z b (f k ), extract the contact impedance differential spectrum. The calculation logic for extracting the contact impedance differential spectrum is: ΔZ c (f k ) is the contact impedance difference spectrum, is the contact impedance spectrum of channel A, is the contact impedance spectrum of channel B, f k is the kth frequency point, κ is the skin dielectric loss factor, θ skin is the skin polarization angle; When ΔZ c (f k ) value is within the preset threshold range, indicating that the electrode contact quality is good and there is no abnormality in the signal measurement; When ΔZ c (f k ) value is higher or lower than the preset threshold range, and when κ exceeds the preset threshold, it indicates that the electrode contact quality is poor, resulting in measurement error, and an early warning signal is issued to inform the user.

3. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 1, characterized in that: The construction logic of the time-varying transfer function is: H(f k ,t real ) is the time-varying transfer function, Z raw (f k ,t real ) is the original impedance signal, IMU (t real ) is the dynamic response strength signal from the inertial measurement unit, τ motion is the motion delay response time, f k is the kth frequency point, t real is the sampling time, j is the imaginary unit; Based on the time-varying transfer function H(f k ,t real )Calculate the modulus of the transfer time-varying function |H(f k ,t real )|, according to the transfer time-varying function modulus|H(f k ,t real )|Build a spectrum suppression mask. The construction logic of the spectrum suppression mask is: M(f k ,t real )=1-|H(f k ,t real )|,M(f k ,t real ) is the spectrum suppression mask; Perform short-time Fourier transform on the original impedance signal, map the original impedance signal to the frequency domain, and combine it with the spectrum suppression mask to obtain the weighted spectrum signal. The calculation logic of the weighted spectrum signal is: is the weighted spectrum signal; Perform inverse short-time Fourier transform on the weighted spectrum signal to obtain the denoised impedance time-frequency signal after suppressing motion artifacts. The calculation logic of the denoised impedance time-frequency signal is: Z bio (f k ,t real ) is the denoised impedance time-frequency signal.

4. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 1, characterized in that: The construction logic of Cole impedance model is: Z i (f) is the impedance of the i-th type of organization at frequency f, R ∞,i is the high-frequency limit impedance of the i-th type of organization, R 0,i is the low-frequency limit impedance of the i-th type of tissue, τ i is the relaxation time constant of the i-th tissue, φ i is the phase delay factor of the i-th tissue, j is the imaginary unit, i is the index variable, m is muscle, and f is fat; The denoised impedance time-frequency signal is reorganized into a four-dimensional impedance tensor data structure according to the time series, tissue type, and Cole impedance model parameter dimensions to form a unified impedance parameter tensor input format. The specific dimensional expression of the four-dimensional impedance tensor is: Z∈C K×2×4×T , K is the number of frequency points, 2 is the number of tissue types, 4 is the number of Cole impedance model parameters, and T is the number of time points. The index expression of each element in the four-dimensional impedance tensor is: Z[f,o,p,t], f is the corresponding frequency, o is the tissue, p is the Cole impedance model parameter, and t is the time point index. Tissue types include m and f, and Cole impedance model parameters include R0, R ∞ , τ and φ, where R0 is the low-frequency resistance, R ∞ is the high-frequency resistance, τ is the relaxation time constant, and φ is the phase delay factor; Tucker decomposition is used to reduce the dimension of the four-dimensional impedance tensor and extract its features. The specific decomposition logic is: Z≈G×1U freq ×2U tissue ×3U param ×4U time ,in, is the core tensor, U tissue ∈R 2×2 、U param ∈R 4×2 and is a factor matrix; Project the core tensor G along the tissue parameter dimension and factor matrix to the physical parameter space to obtain the tissue parameter tensor and extract the characteristic parameters of muscle and fat tissue, that is, execute Get the organizational parameter tensor P∈R 4×2×T , 4 is the number of Cole impedance model parameters, 2 is the number of tissue categories, and T is the number of time points; the specific index expression of the tissue parameter tensor is: P[p,i,t], where p∈{R0,R ∞ ,τ,φ} is the parameter dimension of Cole impedance model, i∈{m,f} is the tissue type, and t is the time point index; Muscle activity index and fat metabolic rate were calculated based on the temporal gradient changes of tissue parameter tensors.

5. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 4, characterized in that: The calculation logic of muscle activity index is: MAI(t) is the muscle activity index, Δτ m is the relaxation time difference of muscle tissue, is the spatial gradient of the muscle tissue phase angle, Δt is the time step, and the muscle tissue relaxation time difference Δτ m The calculation logic is: Δτ m =τ m (t)-τ m (t-1), τ m (t) is the relaxation time constant of the muscle tissue at the current time point t, τ m (t-1) is the relaxation time constant of the muscle tissue at the previous time point t-1, and the spatial gradient of the muscle tissue phase angle The calculation logic is: φ m (t) is the phase angle of the muscle tissue at time point t, φ m (t-1) is the phase angle of the muscle tissue at the previous time point t-1, φ m (t+1) is the phase angle of the muscle tissue at the next time point t+1; The calculation logic of fat metabolic rate is: FMR(t) is fat metabolic rate, R 0,f (t) is the low-frequency resistance of fat tissue, R ∞,f (t) is the high-frequency resistance of fat tissue, R 0,m (t) is the low-frequency resistance of muscle tissue, Δφ f is the time change rate of the phase angle of fat tissue, the time change rate of the phase angle of fat tissue Δφ f The calculation logic is: Δφ f =φ f (t)-φ f (t-Δt), φ f (t) is the phase angle of fat tissue at the current time point t, φ f (t-Δt) is the phase angle of adipose tissue at the previous time point t-Δt; The muscle activity index, fat metabolic rate, tissue relaxation time constant and tissue phase angle are fused as the dynamic tissue parameter matrix output. The expression of the dynamic tissue parameter matrix is: T(t) is the dynamic tissue parameter matrix, where 6. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 1, characterized in that: Perform multimodal fusion of dynamic tissue parameters and user static features to construct physiological feature vectors. The specific steps include: The dynamic organizational parameter matrix and the user's static features are concatenated to form a joint vector. The feature concatenation logic is as follows: is the joint vector, u is the user's static feature, age is age, sex is gender, height is height, weight is weight, and d is the user's static feature dimension; Construct a multi-layer perceptron network and combine the vector As input, a physiological feature vector is output, and the expression of the physiological feature vector is: F = [f1, f2, f3, f4, f5, f6], where F is the physiological feature vector, f1 is the muscle metabolic state, f2 is the fat distribution coefficient, f3 is the hydration balance index, f4 is the skin conductivity, f5 is the respiratory variability index, and f6 is the heart rate variability index.

7. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 1, characterized in that: The construction logic of the metabolic impedance coupled differential equation is: is the rate of change of the body composition vector y(t) with respect to time t, f(t,y) is the function of the body composition vector y(t) and time t, F(t)∈R 6 is the physiological feature vector at time t, A∈R 3×6 is the coupling weight matrix, B∈R 3×4 is the sensitivity matrix, W∈R 6×6 is the weight mapping matrix, T(t)∈R 4 is the dynamic organizational parameter at time t; The fourth-order Runge-Kutta numerical solution is used to solve the metabolic impedance coupled differential equation. The calculation logic of the solution is: Among them, y t ∈R 3 is the estimated body composition vector at time t, corresponding to body fat percentage, muscle mass and total body water respectively, Δt is the time step, k1, k2, k3 and k4 are the estimated values ​​of the intermediate step length; Upon detection When it is greater than the preset threshold θ, the sensitivity matrix is ​​updated and adjusted, and the update logic is: γ is the attenuation coefficient.

8. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 7, characterized in that: The body composition estimation vector y t The body composition vector is output after calibration. The specific steps include: The body composition estimation vector is normalized by the Sigmoid mapping function. The processing logic is as follows: y * (t) is the normalized volume component vector, BF * is the original predicted value of body fat percentage, MM * is the raw predicted value of muscle mass, TBW * is the original predicted value of total body water; The normalized output volume component vector y * (t) is further linearly mapped to a range with physiological significance: BF final is the body fat percentage and MM after mapping final is the mapped muscle mass, TBW final is the total body water after mapping, BF min and BF max They are the physiological upper and lower limits of body fat percentage, MM min and MM max They are the physiological upper and lower limits of muscle mass, TBW min and TBW max They are the physiological upper and lower limits of total body water, respectively.

9. The method for smart wearable health monitoring based on bioimpedance analysis according to claim 8, characterized in that: The mapped body composition vector is input into a preset health threshold discrimination module to analyze the degree of deviation between the body fat percentage, muscle mass, and total body water parameters and the corresponding preset physiological acceptable ranges, and then perform health status classification and output secondary health status labels, including normal state, marginal state, and abnormal state; Triggering graded warning responses based on health status tags. When the result is a marginal state, a level 1 warning is triggered, driving the device to vibrate to alert the user; When the judgment result is an abnormal state, a secondary warning is triggered and the abnormal information is synchronously sent to the associated medical service terminal.

10. A smart wearable health monitoring system based on bioimpedance analysis, used to implement the smart wearable health monitoring method based on bioimpedance analysis according to any one of claims 1 to 9, characterized in that: include: Multi-frequency impedance acquisition module: used to obtain multi-frequency human body electrical response data through multi-frequency excitation signals, construct the original impedance spectrum of biological tissue in the frequency domain, and synchronously acquire the original impedance signal in the time domain; Electrode quality assessment module: used to build a dual-channel impedance model based on the original impedance spectrum to model skin electrode contact artifacts and achieve quantitative assessment of electrode quality; De-noising impedance time-frequency signal acquisition module: used to introduce inertial measurement unit information to construct a time-varying transfer function, suppress motion artifacts in the frequency domain, and obtain a de-noising impedance time-frequency signal; Tissue impedance modeling and tensor feature extraction module: used to establish the Cole impedance model, extract the impedance characteristic parameters of muscle and fat tissues through tensor decomposition method, and construct a dynamic tissue parameter matrix; Physiological feature vector generation module: used to fuse dynamic tissue parameters with user static features and generate physiological feature vectors through a multi-layer perceptron; Metabolic impedance coupling modeling module: used to construct metabolic impedance coupling differential equations based on physiological eigenvectors to obtain body composition estimation vectors, and use hyperbolic tangent activation functions to achieve dynamic tissue parameter weight allocation and sensitivity adjustment; Body composition calibration and grading evaluation module: used to calibrate the body composition estimation vector to output the body composition vector, and perform grading evaluation on the calibrated body composition vector.

Images