Health condition monitoring method based on smart bracelet

Abstract

The invention belongs to the technical field of signal processing, and particularly relates to a health condition monitoring method based on a smart bracelet, and the method comprises the steps: 1, regarding a physiological signal collected by each sensor on the smart bracelet in a set time interval as a combination of a plurality of different frequency components, and containing band-pass noise; 2, evolving each time domain signal in a high-order tensor field to obtain a wavelet coefficient; 3, calculating the fractal dimension and singular spectrum of each sensor under each mother wavelet based on the wavelet coefficient; obtaining a chaos state between the sensors; and 4, introducing a continuously evolved global health threshold value by integrally summarizing the collaborative state and the chaotic state of the sensor, and ensuring the numerical value to be stable. According to the method, the problems of insufficient multi-sensor signal fusion, limited anti-noise capability, lack of dynamics and individuation and the like in the prior art are solved, and the accuracy, robustness and real-time performance of health monitoring are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and in particular relates to a health status monitoring method based on a smart bracelet. Background Art

[0002] With the rapid popularization of wearable devices, smart bracelets have become an important technical carrier in the field of health monitoring. These devices are widely used in daily health management, sports tracking and disease early warning due to their lightness, real-time and high efficiency. In the existing technology, smart bracelets collect users' physiological signals through built-in multiple sensors (such as heart rate sensors, blood oxygen sensors, skin conductance sensors, etc.), and use signal processing algorithms to analyze the user's health status. However, although these technologies have achieved the monitoring and evaluation of certain key health indicators, under the actual needs of complex environments and multi-dimensional signal processing, the existing technologies still face many shortcomings. Current smart bracelets usually rely on independent sensor modules to capture a single type of physiological signal, such as: Heart rate monitoring: Photoplethysmography (PPG) is used to calculate heart rate in real time by detecting changes in blood volume in the blood. Blood oxygen detection: The oxygen saturation in the blood is measured by the difference in the absorption of infrared light and red light. Skin conductance: The level of sympathetic nerve activity is reflected by changes in skin resistance, which is often used for stress assessment. Motion data monitoring: The number of steps, speed and movement pattern of the user are captured by accelerometers and gyroscopes.

[0003] These methods perform well in monitoring in a single dimension, but most of them are based on a single parameter and lack the ability to conduct comprehensive analysis of multi-sensor data. In addition, the data processing algorithms in the prior art are usually based on relatively basic signal analysis methods, such as low-pass filtering, Fourier transform, etc., which cannot fully reveal the multi-scale dynamic characteristics and complex coupling relationships of physiological signals. The prior art usually analyzes the signals of a single sensor independently, while ignoring the potential coupling relationship between multiple sensors. For example, heart rate and skin conductance signals may be jointly regulated by the sympathetic nervous system, and when the two change synergistically, they may reveal the user's stress state or emotional fluctuations. However, the current smart bracelet lacks linkage analysis of these multi-dimensional data, resulting in a one-sided assessment of the user's overall health status. Since smart bracelets operate in complex scenarios (such as strenuous exercise or changes in ambient light), the signals collected by the sensors are often mixed with a large amount of random noise and environmental interference. For example, the heart rate sensor may have measurement errors due to changes in the tightness of the bracelet or the state of movement. The prior art usually relies on simple filters in signal processing, and these methods are prone to weakening the core characteristics of the signal or losing key mutation information. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to provide a health status monitoring method based on a smart bracelet. The present invention overcomes the problems of insufficient multi-sensor signal fusion, limited noise resistance, lack of dynamics and personalization in the prior art, and significantly improves the accuracy, robustness and real-time performance of health monitoring.

[0005] The technical solution adopted by the present invention is as follows: A health status monitoring method based on a smart bracelet, the method comprising: Step 1: The physiological signals collected by each sensor on the smart bracelet within a set time interval are regarded as a combination of several different frequency components, including bandpass noise, so that each physiological signal is represented by a unified time domain signal; Step 2: Evolve each time domain signal in a high-order tensor field. During the evolution process, use the Laplace operator to eliminate sharp or high-frequency noise to ensure local smoothness. At the same time, limit over-smoothing through the difference term with the time domain signal to obtain a smooth signal. Use multi-mother wavelet joint transform to perform multi-scale decomposition on the smooth signal to obtain wavelet coefficients. Step 3: Calculate the fractal dimension and singular spectrum of each sensor under each mother wavelet based on the wavelet coefficients; use the fractal dimension and singular spectrum as coupling conditions to discretely iterate the cooperative state between sensors; add chaotic mapping to the cooperative state, use fractal difference to correct the chaotic mapping, and obtain the chaotic state between sensors; Step 4: By summarizing the overall collaborative state and chaotic state of the sensor, a continuously evolving global health threshold is introduced and the numerical stability is ensured; according to the relative position of the health threshold and the average state of the sensor, it is divided into multiple levels and the final health assessment result is output; the average state of the sensor is defined as the weighted average of the chaotic state and collaborative state of the sensor.

[0006] Furthermore, the smart bracelet includes sensors, each sensor in the time interval The physiological signal is obtained by continuous sampling, and the time domain signal corresponding to the physiological signal is ;in, is the sampling time, and its value range is ; is the sensor serial number index; the sensor includes at least: a blood oxygen sensor, a skin resistance sensor, a temperature sensor, a heart rate sensor, a blood pressure sensor, and a respiratory rate sensor.

[0007] Furthermore, the time domain signal It is expressed by the following formula: ; in, is the number of frequency components; is an integer subscript index; For the frequency components in the sensor The sine amplitude on reflects the intensity of the sine component of the frequency component in the time domain signal; For the frequency components in the sensor The cosine amplitude on reflects the strength of the cosine component of the frequency component in the time domain signal; For the The frequency value of the frequency component; For the frequency components in the sensor The sinusoidal phase offset of For the frequency components in the sensor The cosine phase offset of Indicates the frequency range Integrate the noise signal within; is the lower limit of the frequency range; is the upper limit of the frequency range; represents the bandpass noise, describing the The sensor is sampled at The frequency is The signal value generated by the continuous noise component.

[0008] Furthermore, let the high-order tensor field in step 2 be , when the time domain signal evolves in the high-order tensor field, the corresponding evolution equation is: ; in, is an integer subscript index, and ; is the difference operator; is the Laplace operator; is the evolution time, and the sampling time phase distinction; It is a high-order tensor field with a dimension of at least two dimensions and a range of real numbers; is a smooth signal.

[0009] Furthermore, in step 2, the smooth signal is decomposed into multiple scales using a multi-mother wavelet joint transform to obtain wavelet coefficients using the following formula: ; in, Indicated in scale , translation , mother wavelet ,sensor The wavelet coefficients under ; is the scale, used to stretch or compress the wavelet function; is the translation amount on the time axis; is the mother wavelet number, there are wavelet functions of different types or shapes; Indicates the time interval Perform integration; For the The mother wavelet function is: .

[0010] Furthermore, in step 2 and step 3, the fractal dimension of each sensor under each mother wavelet is calculated based on the wavelet coefficients by the following formula: ; in, For the sensor , mother wavelet In the broad index The fractal dimension of It is a generalized index and can take positive, negative or zero values ​​except 1. The singular spectrum is calculated by the following formula: ; in, for The corresponding fractal dimension.

[0011] Furthermore, in step 3, the fractal dimension and singular spectrum are used as coupling conditions to discretize the cooperative state between the iterative sensors through the following formula: ; in, is the discrete iteration step When the sensor The corresponding collaborative state; is the local convergence coefficient set, controlling Towards target value The speed of convergence; For sensor Target value of is the discrete iteration step When the sensor The corresponding collaborative state.

[0012] Furthermore, the chaotic map is added to the cooperative state through the following formula, and the chaotic map is corrected by using the fractal difference to obtain the chaotic state between the sensors: ; in, For discrete time steps When the sensor The chaotic state; For sensor The corresponding Logistic map growth rate; is a typical nonlinear term of the Logistic mapping; For discrete time steps When the sensor The chaotic state; The power exponent of the penalty term is set, which determines the sensitivity of the difference term; is the set fractal compensation coefficient; For a specific order and a specific mother wavelet The fractal dimension of and All are set values; For a specific order and mother wavelet The singular spectrum difference under .

[0013] Furthermore, through the following formula, in step 4, the following partial differential equation is used to introduce a global health threshold of continuous evolution by summarizing the overall coordination state and chaotic state of the sensor, and ensure numerical stability: ; ; in, For sensor The historical mean of As the evolution time Changing global health thresholds; is the set cumulative gain coefficient; , is the integral variable; For The cooperative state calculated at this discrete time step; For The chaotic state calculated at this discrete time step; is the adjustment index; is the set threshold smoothness coefficient; : The threshold value in evolution time The second-order derivative diffusion term on is used to prevent excessive oscillation; is the initial threshold.

[0014] Furthermore, the final health assessment result is output through the following formula: ; in, The upper boundary of the setting; is the lower boundary of the setting; is a discrete time step Health assessment results at the time of Discrete time step The global health threshold at ; is the discrete iteration step When the sensor The corresponding collaborative state; is the discrete iteration step When the sensor The corresponding chaotic state.

[0015] By adopting the above technical scheme, the present invention produces the following beneficial effects: the present invention incorporates the data of multiple sensors (such as heart rate, blood oxygen, skin conductance, etc.) on the smart bracelet into a unified framework for processing by introducing a modeling method of a high-order tensor field. This method effectively solves the limitations of independent analysis of a single sensor in the prior art and realizes dynamic coupling and collaborative analysis between multi-sensor signals. The fusion analysis of multiple sensors can reveal complex physiological patterns. For example, the synchronous fluctuations of heart rate and blood oxygen may point to rapid recovery after exercise, while the correlation between skin conductance and acceleration signals may indicate the user's stress or emotional state. Through the linkage analysis of multi-sensor signals, the present invention not only improves the integrity of health assessment, but also reduces health judgment errors caused by errors or interference of a single sensor. For example, when the heart rate sensor data is interfered by motion artifacts, the system can refer to the data of other sensors (such as acceleration signals) for correction, thereby significantly improving the accuracy and robustness of the monitoring results. In complex environments (such as strenuous exercise, changes in light, etc.), the signals collected by the smart bracelet inevitably contain a lot of noise and interference. The present invention introduces partial differential equations and smoothing mechanisms through the evolution model of high-order tensor fields, suppresses high-frequency noise with the Laplace operator, and maintains the key features of the signal through the difference constraint term. This method effectively avoids the excessive smoothing of the signal or the loss of important details that may be caused by traditional filters. In addition, in the process of multi-scale decomposition of the signal, the present invention decomposes the different frequency components of the signal into a multi-level structure through the joint transformation of multiple mother wavelets. This decomposition can not only distinguish the main frequency components and random interference in the signal, but also capture the mixed characteristics of short-term mutations and long-term trends, making health assessment more sensitive and comprehensive. For example, when detecting heart rate variability (HRV), the present invention can extract low-frequency and high-frequency components at the same time, thereby more accurately reflecting the user's sympathetic and parasympathetic nerve activity levels. The present invention deeply explores the multi-scale and nonlinear characteristics of physiological signals through advanced feature indicators such as fractal dimension and singular spectrum. The fractal dimension can reflect the complexity and self-similarity of the signal, and the singular spectrum can describe the singularity distribution of the signal in different local areas. The introduction of these features makes up for the shortcomings of traditional linear analysis methods, making the system more sensitive in detecting complex dynamic patterns (such as arrhythmias, apnea, etc.). In addition, by introducing chaotic mapping, the present invention combines the nonlinear characteristics of physiological signals with synergistic states and fractal features, further improving the system's ability to capture abnormal signals. For example, when the user's heart rate signal fluctuates abnormally and the fractal complexity of the blood oxygen signal increases significantly, the exponential amplification characteristics of the chaotic state can quickly mark this abnormal state, thereby providing a reliable basis for health warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A method flow chart of a health status monitoring method based on a smart bracelet provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0018] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other alternative features that are equivalent or have similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0019] Example 1: Reference Figure 1 , a health status monitoring method based on a smart bracelet, the method comprising: Step 1: The physiological signals collected by each sensor on the smart bracelet within a set time interval are regarded as a combination of several different frequency components, including bandpass noise, so that each physiological signal is represented by a unified time domain signal; The signals collected by the sensor can essentially be regarded as the superposition of multiple frequency components. Taking the heart rate signal as an example, its main frequency component is usually located in the low frequency band, which is manifested as regular heartbeat fluctuations. In the case of movement or emotional changes, higher frequency fluctuations or abnormal interference may also be superimposed. Similarly, the skin conductance signal reflects sympathetic nerve activity, usually containing a slowly changing baseline signal and a faster changing peak feature. In addition, the signal captured by the acceleration sensor may contain both low-frequency body swing components and high-frequency jitter components. The common feature of all these signals is that they contain both relatively regular components dominated by physiological activities and highly random noise or interference. Based on this, the principle of step 1 is to uniformly model these multi-band signals as a combination of several harmonic components, while taking into account the bandpass noise components of continuous frequency bands. The core idea of ​​this modeling is to express the composition structure of the signal in mathematical form, reflecting the main physiological information and random factors contained in the signal in a more intuitive way. The signal collected by each sensor can be regarded as the superposition of several sine and cosine functions, each function represents a specific frequency component and is defined by amplitude and phase. On this basis, in order to further consider the irregular noise interference in the signal, a bandpass noise term is introduced in the modeling, which reflects the distribution of noise in a certain frequency range through integration in the frequency domain. This representation method can not only capture the regularity of the signal, but also accurately describe its irregularity.

[0020] Step 2: Evolve each time domain signal in a high-order tensor field. During the evolution process, use the Laplace operator to eliminate sharp or high-frequency noise to ensure local smoothness. At the same time, limit over-smoothing through the difference term with the time domain signal to obtain a smooth signal. Use multi-mother wavelet joint transform to perform multi-scale decomposition on the smooth signal to obtain wavelet coefficients. The high-order tensor field is a representation of multidimensional signals. It integrates the multi-sensor signals on the smart bracelet into a unified mathematical framework, so that the correlation between time, sensor position and signal amplitude can be effectively modeled. In this tensor field, each point represents a combination of a moment, a sensor and a specific signal value. This form is not only convenient for processing the spatial correlation between signals, but also provides convenience for describing the evolution of sensor signals in the time dimension. In step 2, the denoising and smoothing of the signal is completed by the evolution of the tensor field. This process is essentially an iterative optimization process, with the help of the Laplace operator to achieve local smoothing of the signal. The Laplace operator is a classic diffusion tool. By calculating the mean change of the signal value in the local neighborhood in the tensor field, it can effectively eliminate sharp or high-frequency noise components and make the signal smooth. This diffusion method can simulate the physical diffusion process, smooth the sharp irregular signal distribution, and retain the global structural characteristics in the signal. However, simple smoothing may cause the signal to diffuse excessively, thereby losing key health information in the original signal. In order to avoid this problem, a difference term restriction mechanism is introduced in the process of tensor field evolution. The difference term constrains the signal smoothing process within a certain range by measuring the deviation between the smoothed signal and the original signal, so that the smoothed signal can remove high-frequency noise and be as close to the main mode of the original signal as possible. This process is particularly important in smart bracelet applications, because the characteristics of physiological signals such as heart rate and blood oxygen are often manifested as low-frequency periodic fluctuations. Excessive smoothing will erase these key characteristic patterns, making health assessment lose its basis.

[0021] After the tensor field evolution is completed, the smooth state of the signal is preserved, and then the multi-mother wavelet joint transform stage is entered. The goal of this stage is to decompose the signal at multiple scales to extract rich information in both time and frequency dimensions. Wavelet transform is a powerful mathematical tool that can decompose signals into different scales while retaining high-resolution characteristics in the time and frequency domains. Traditional wavelet analysis is usually based on a single mother wavelet function, such as Daubechies or Haar wavelets, but in health monitoring scenarios, a single mother wavelet may not be able to fully capture the diverse features in the signal. For example, a heart rate signal may contain smooth main fluctuations or may be mixed with faster local changes; a skin conductance signal may have sharp peaks and slow background drifts. To this end, the concept of multi-mother wavelet joint transform is introduced in step 2, using multiple mother wavelet functions to decompose the signal, and each mother wavelet extracts specific features to obtain more comprehensive multi-scale information. In the multi-mother wavelet transform process, each mother wavelet function will convolve the signal for different scale parameters and time shift parameters to generate corresponding wavelet coefficients. These coefficients not only reflect the local change characteristics of the signal in different time periods, but also reveal the strength distribution of different frequency components. For example, small scales correspond to high-frequency components and can capture rapid changes in the signal, while large scales correspond to low-frequency components and can reflect the global trend of the signal. By comprehensively analyzing the decomposition results of multiple mother wavelets, a multi-dimensional wavelet coefficient matrix can be obtained, in which each dimension contains the change characteristics of the signal within a specific time and frequency range.

[0022] Step 3: Calculate the fractal dimension and singular spectrum of each sensor under each mother wavelet based on the wavelet coefficients; use the fractal dimension and singular spectrum as coupling conditions to discretely iterate the cooperative state between sensors; add chaotic mapping to the cooperative state, use fractal difference to correct the chaotic mapping, and obtain the chaotic state between sensors; Step 4: By summarizing the overall collaborative state and chaotic state of the sensor, a continuously evolving global health threshold is introduced and the numerical stability is ensured; according to the relative position of the health threshold and the average state of the sensor, it is divided into multiple levels and the final health assessment result is output; the average state of the sensor is defined as the weighted average of the chaotic state and collaborative state of the sensor.

[0023] First, the signal decomposition results obtained by multi-mother wavelet transform provide a basis for the calculation of fractal dimension and singular spectrum. Fractal dimension is an important indicator to describe signal complexity, which can reflect the irregularity and self-similarity characteristics of the signal at different scales. For example, heart rate variability usually has fractal characteristics, which is reflected in the correlation between local fluctuations and global trends in time series, while the peaks and background drifts in skin conductance signals are manifested as the complexity of multi-scale distribution. By calculating the scale coefficients obtained by wavelet decomposition, the generalized fractal dimension of the signal can be estimated, and then the characteristic complexity of the signal at different time scales can be quantitatively described. At the same time, the introduction of singular spectrum further enriches the characterization of signal distribution characteristics. It reveals the sparse and dense distribution of signals in different local areas by observing the behavior of fractal dimension changing with scale. This multi-fractal analysis method enables the health signal of each sensor to be mapped into a high-dimensional feature space containing fractal dimension and singular spectrum, thereby providing multi-level information for subsequent dynamic analysis. Based on the feature extraction of fractal dimension and singular spectrum, this step further uses these complex indicators as coupling conditions between sensors to construct a discrete iterative collaborative state model. The concept of collaborative state originates from the interaction between multiple sensors, which describes the consistency and difference between different sensor signals in dynamic changes. For example, when the fractal dimension changes of heart rate signal and skin conductance signal show a strong positive correlation, it can be inferred that these signals may be regulated by a common physiological mechanism (such as sympathetic nervous system activity); and when the singular spectrum characteristics of acceleration signal deviate significantly from other signals, it may mean that some abnormal movement interferes with the overall physiological state. By modeling these collaborative relationships, the potential correlation between sensor signals can be revealed. The iterative process of collaborative state is usually based on the dynamic adjustment mechanism of feature differences. When the fractal feature differences of two sensors are small, their states tend to be consistent; when the difference is large, the collaborative model will exert a greater correction force on this difference to reflect the potential physiological changes.

[0024] On the basis of collaborative state modeling, this step also introduces chaos mapping to enhance the system's dynamic response ability to complex signals. Chaotic mapping is introduced to simulate the common nonlinear and sensitive characteristics in sensor signals. For example, the heart rate signal may show rapid changes in nonlinear complexity under stress, and this change is often not accurately described by a simple linear model. By embedding chaotic dynamics in the evolution of collaborative states, these tiny dynamic changes can be captured more sensitively. Chaotic mapping also allows the system to make adaptive adjustments to abnormal deviations in fractal dimension or singular spectrum. For example, when the fractal dimension of a sensor is significantly lower than the normal level, the chaotic mapping can amplify the weight of this signal, so that the system can identify potential health risks more quickly. Another important role of chaotic mapping is to enhance the overall complexity and sensitivity of the system. In the chaotic model, the state of the system is very sensitive to the initial conditions, and this characteristic is used to detect small differences in sensor signals. By introducing the difference correction term of fractal dimension in the chaotic mapping, the health signal characteristics of multiple sensors can be better integrated while maintaining the overall complexity of the system. Ultimately, the chaotic state between sensors not only retains the multi-scale characteristics revealed in fractal analysis, but also captures the nonlinear changes of health signals through dynamic modeling, providing higher flexibility for multi-sensor collaborative analysis.

[0025] Embodiment 2: The smart bracelet includes sensors, each sensor in the time interval The physiological signal is obtained by continuous sampling, and the time domain signal corresponding to the physiological signal is ;in, is the sampling time, and its value range is ; is the sensor serial number index; the sensor includes at least: a blood oxygen sensor, a skin resistance sensor, a temperature sensor, a heart rate sensor, a blood pressure sensor, and a respiratory rate sensor.

[0026] Specifically, the smart bracelet first relies on its blood oxygen sensor to monitor the oxygen saturation in the blood ( ), which is an important parameter reflecting the oxygen supply capacity of blood, especially when the user is in high-intensity exercise, sleeping state or high-altitude environment. By continuously sampling and recording blood oxygen signals, the dynamic changes of blood oxygen levels can be captured, possible hypoxia can be detected in time, and early warning support can be provided for the health of respiratory and circulatory systems. At the same time, the skin resistance sensor reflects the activity of the user's sympathetic nervous system by monitoring the changes in skin conductance. This signal is often highly dynamic and can reveal fluctuations in emotional states such as stress and anxiety, which is of great significance for mental health assessment. In addition, the temperature sensor indirectly infers the temperature change of the user by detecting the temperature of the user's skin surface. When the user's body temperature rises or falls abnormally, this sensor can provide early warning of diseases or other health risks at an early stage. The heart rate sensor plays a key role in the health monitoring method of smart bracelets. It records the user's heart rate signal in real time through photoplethysmography (PPG). The dynamic fluctuation of the heart rate signal reflects the working state of the cardiovascular system, and plays an important role in exercise monitoring, rest state assessment, and heart health risk prediction. Combined with the data provided by the blood pressure sensor, the heart rate signal can be further used to assess the overall health of the cardiovascular system. Blood pressure sensors provide systolic and diastolic blood pressure data by measuring pulse wave velocity or vibration-based measurement methods, playing an irreplaceable role in hypertension monitoring and cardiovascular disease early warning. On the other hand, respiratory rate sensors calculate the user's respiratory rate by detecting chest expansion, blood oxygen fluctuations or heart rate variability (HRV). This indicator is important for monitoring sleep apnea syndrome, exercise status adjustment, and respiratory diseases.

[0027] Smart bracelet in time interval All sensors are sampled continuously to record every moment These signal data are expressed in a unified mathematical framework as ,in represents the sampling time, Represents the sensor serial number. Such a unified modeling method can not only describe the multi-sensor signals in a time series, but also provide a basis for multi-dimensional signal analysis. In the process of mathematical modeling, the signal of each sensor is regarded as a function that changes continuously on the time axis. The smart bracelet uses high-frequency sampling technology to record the subtle fluctuations of the signal. This method not only retains the high-precision characteristics of the signal, but also can reveal the regularity of dynamic changes in subsequent analysis. A significant advantage of this signal modeling method is that it realizes the standardization of cross-sensor data, which provides the possibility for subsequent multi-sensor joint analysis. For example, the correlation between blood oxygen signals and heart rate signals can be analyzed by synchronously comparing the time series after unified modeling, thereby revealing the potential impact of hypoxia on heart rate variability. Similarly, the drastic fluctuations in skin resistance signals may have a synergistic relationship with short-term changes in heart rate or blood pressure. Through this framework, the interaction between multi-sensor data can be deeply explored. This multi-dimensional analysis method is difficult to achieve in the traditional single sensor monitoring method. Unified modeling enables the smart bracelet to describe the user's physiological state more comprehensively and accurately. In addition, the continuous sampling method in Example 2 ensures the temporal continuity of the signal and the ability to record dynamic changes. Unlike intermittent measurements, continuous sampling can capture subtle changes in signals, such as fluctuations in heart rate during different stages of exercise, the downward trend in blood oxygen levels during sleep, or the slow changes in skin temperature with the external environment. This data recording method provides smart bracelets with higher resolution, thereby capturing more potential problems in health assessments. and sensor serial number Through bivariate modeling, the system can synchronously process data from different sensors at different time points, laying a data foundation for subsequent high-order tensor modeling and feature extraction.

[0028] Example 3: Time Domain Signal It is expressed by the following formula: ; in, is the number of frequency components; is an integer subscript index; For the frequency components in the sensor The sine amplitude on reflects the intensity of the sine component of the frequency component in the time domain signal; For the frequency components in the sensor The cosine amplitude on reflects the strength of the cosine component of the frequency component in the time domain signal; For the The frequency value of the frequency component; For the frequency components in the sensor The sinusoidal phase offset of For the frequency components in the sensor The cosine phase offset of Indicates the frequency range Integrate the noise signal within; is the lower limit of the frequency range; is the upper limit of the frequency range; represents the bandpass noise, describing the The sensor is sampled at The frequency is The signal value generated by the continuous noise component.

[0029] Specifically, in the health status monitoring method based on the smart bracelet of the present invention, the formula proposed in Example 3 is calculated by Several discrete frequency components and a bandpass noise integral term are introduced simultaneously, which cleverly combines the periodicity and randomness of physiological signals, so as to more flexibly capture the multiple changing characteristics of the user's physiological state. In the discrete frequency part, this method uses several sine and cosine functions to express the periodic oscillations presented by typical life rhythms such as heart rate and breathing. Each frequency component is controlled by two sets of key parameters: amplitude and phase. The amplitude determines the contribution strength of the periodic component in the overall signal, and the phase is used to characterize its starting position and local fluctuation characteristics on the time axis. and The combination of can mathematically construct a variety of waveform superpositions in different forms to simulate the rich dynamic changes of blood oxygen, skin resistance, temperature, heart rate, blood pressure and respiratory rate sensors under the influence of various external environments and personal physiques. When the signal collected by the sensor shows a strong periodicity, the discrete part can prominently reflect the frequency range and amplitude of the main component. When there are multiple main frequency components that are close or coupled to each other in the signal, different The corresponding sine and cosine terms will represent the complementarity and superposition of these components, making the fitting of complex waveforms more accurate. The integral description of can express the random interference hidden in high-frequency jitter or low-frequency drift, thereby incorporating random fluctuations caused by daily activities, changes in ambient light, and skin sweat secretion into a unified mathematical framework. Continuously distributed in the frequency domain, the use of integral form can be used to describe a wide range of frequency band disturbances in an integrated manner, which can better reflect the random impact of noise on the overall signal compared to simply relying on discrete frequency terms. In the actual operation of the smart bracelet, each sensor is often affected by both physiological and non-physiological factors. If only a fixed periodic component is considered, short-term bursts or irregular changes may be missed. However, through bandpass noise, the interference information with a more dispersed frequency distribution can be taken into account, so that the model can be balanced in terms of characterization accuracy and adaptability. It is worth noting that the discrete frequency component and the bandpass noise are not separated. Together, they constitute a signal description system with the dual characteristics of "rule + randomness", which can provide more complete information in subsequent signal denoising, multi-scale decomposition and fractal analysis. Specifically, when the subsequent algorithm extracts wavelet coefficients or calculates fractal dimensions, the regular fluctuations from the discrete part will make the main physiological rhythm easier to focus on, while the bandpass noise provides the possibility of revealing subtle fluctuations and abnormal events. Once the system detects that the energy of a specific frequency deviates significantly from the normal state, the formula can be used to trace back to whether it is caused by discrete frequency gain or noise band surge, thereby more clearly identifying the source of the abnormal component in the time dimension and frequency dimension. At the same time, the amplitude and phase parameters in the formula can be personalized for different sensors. For example, the blood oxygen sensor and the heart rate sensor may share similar low-frequency harmonics, and may also experience higher-frequency random disturbances under the influence of environmental noise. With this holistic modeling method that is compatible with discrete harmonics and bandpass noise, the smart bracelet can more accurately grasp the specific patterns contained in each physiological signal and unify multi-source data into an analysis framework. In this way, whether it is the use of high-order tensor partial differential equations for smoothing in subsequent steps or the use of multi-mother wavelet transforms for frequency band division, a more comprehensive description of the initial signal can be obtained based on this formula, avoiding the limitations of simple filtering or single frequency fitting.

[0030] Example 4: Assume that the high-order tensor field in step 2 is , when the time domain signal evolves in the high-order tensor field, the corresponding evolution equation is: ; in, is an integer subscript index, and ; is the difference operator; is the Laplace operator; is the evolution time, and the sampling time phase distinction; It is a high-order tensor field with a dimension of at least two dimensions and a range of real numbers; is a smooth signal.

[0031] Specifically, in the smart bracelet health monitoring method described in the present invention, the reason why the evolution equation proposed in Example 4 can effectively smooth the time domain signals collected by multiple sensors is that the core is to use the high-order tensor field to smooth the signal in the evolution time. The algorithm performs iterative updates in the dimension and uses the dual diffusion effect of the differential operator and the Laplace operator to suppress random noise, while maintaining the faithful preservation of physiological characteristics through the difference term between the original signal. Specifically, when the signal is uniformly mapped to This includes at least the sampling time " and "Sensor Index "After the two-dimensional field is formed, the conditions for multi-dimensional coupling smoothness are created: because in this tensor structure, each can simultaneously sense the states of adjacent sampling points and other sensors, thereby performing differential operations (denoted as ) to obtain local change information, and through the Laplace operator (denoted as ) to achieve diffusion and equilibrium in a larger range. When the “ " is no longer equal to zero, it means that the system has made some changes to the current signal: on the one hand, The represented differencing action will smooth out high frequency noise as local extremes; on the other hand, This provides a basis for the information coupling between different sensors, so that if a sensor is subject to instantaneous interference and has a spike, the smooth state of the adjacent sensor can also "pull it back" or compensate for it. At the same time, if only relying on this diffusion process, the signal may be constantly pushed to the "average value" and lose its original physiological characteristics. For this reason, the equation also deliberately retains The difference limit term of , thereby penalizing over-smoothing, At each step of evolution, the original signal "Pull" back to avoid over-smoothing out key periodic or slow-changing features. It is worth noting that the setting of coefficients in the equation (such as the weight distribution of diffusion and penalty terms) is not arbitrary, but is tuned according to the different physiological indicators monitored by the smart bracelet: for example, when monitoring heart rate, more attention may be paid to removing high-frequency jitter; when monitoring skin conductance, it is more necessary to retain peak fluctuations to analyze stress response. The increasing Gradually converges, and the obtained It can be regarded as a smooth signal with greatly reduced noise in each frequency band and the main physiological rhythm completely preserved. In this way, when the system enters the subsequent steps (including multi-mother wavelet decomposition, fractal dimension calculation and health threshold determination), it can perform in-depth analysis based on this relatively pure signal, thereby effectively reducing the probability of misjudgment caused by external interference. Furthermore, in the context of multi-sensor coupling, this evolution equation can uniformly process the time domain data of all sensors, allowing multiple signals such as blood oxygen, heart rate, and skin resistance to exchange information with each other in the same iterative process, thereby allowing the health monitoring of smart bracelets to achieve higher accuracy and reliability. For example, when the heart rate signal and the blood oxygen signal show obvious abnormalities at the same time, the system can capture this global deviation in the denoising and smoothing stage, and determine the possible physiological cause of the abnormality in subsequent analysis. It can be seen that the formula of Example 4 is not limited to one-time noise filtering at the principle level, but relies on a whole set of evolutionary mechanisms such as "diffusion-difference-coupling" to undergo multiple tiny iterations in the multi-dimensional tensor field to approximate more realistic and smoother physiological data. It can fully suppress random interference and maintain sensitivity to the original signal characteristics on a global scale, providing a powerful signal preprocessing method for health status monitoring based on smart bracelets.

[0032] Embodiment 5: In step 2, the smooth signal is decomposed into multiple scales using a multi-mother wavelet joint transform to obtain wavelet coefficients using the following formula: ; in, Indicated in scale , translation , mother wavelet ,sensor The wavelet coefficients under ; is the scale, used to stretch or compress the wavelet function; is the translation amount on the time axis; is the mother wavelet number, there are wavelet functions of different types or shapes; Indicates the time interval Perform integration; For the The mother wavelet function is: .

[0033] Specifically, in the health status monitoring method for smart bracelets of the present invention, the core idea of ​​Example 5 is to use the joint transformation of multiple mother wavelets to further project the smooth signal obtained in the previous step into the multi-dimensional space of different scales and time translations, so as to extract richer local and overall features. Specifically, when the smooth signal processed by the high-order tensor field evolution equation is obtained After that, it can be regarded as a relatively stable time series data with most of the noise interference removed. Integrate over each set of "scales "Translation ” and “mother wavelet number "By performing wavelet convolution operation at Such a set of multi-dimensional wavelet coefficients. The reason why the concept of multiple mother wavelets is emphasized is that a single type of wavelet function can often only highlight the local features of a certain specific form. For example, the Mexican hat wavelet is more suitable for detecting spikes or mutations, the Morlet wavelet is more advantageous in analyzing continuously modulated signals, and the Daubechies wavelet has a good balance between time domain and frequency domain smoothness. If the data collected by the smart bracelet only uses a single mother wavelet, information is often missed or specific patterns cannot be accurately captured due to the diversity and complexity of physiological signals. By By expanding the value, that is, selecting wavelet functions of various shapes, the same smooth signal can be deconstructed at different "observation angles", thereby fully revealing the multi-level features hidden in signals such as heart rate, blood oxygen, skin resistance and acceleration.

[0034] Combining some formulas given in Example 5, it can be seen that when calculating the wavelet coefficients, it is necessary to follow The idea is to perform integral operations, where This normalization factor can ensure that the same mother wavelet maintains energy consistency at different stretching scales. Since the signals of smart bracelets often have both slow-changing periodic behaviors and fast and violent short-term fluctuations, the scale The smaller it is, the more it can focus on detecting high-frequency or short-term mutations; The larger the value, the more it tends to show global trends and low-frequency features. When convolution is performed at each scale, a series of wavelet coefficients can be obtained. The larger the value of these coefficients, the higher the similarity between the signal and the wavelet function used at this scale and time position. In this way, if there is an obvious peak in a certain time domain interval, such as a rapid surge in skin resistance due to emotion or exercise, a large wavelet coefficient will be obtained at the corresponding high-frequency scale; conversely, if the heart rate fluctuates periodically at different times of the day, a relatively significant coefficient peak will be shown at a larger scale.

[0035] It is worth noting that Example 5 also provides the wavelet function In a similar way to the Mexican hat shape, it contains a The correction term and a The Gaussian attenuation term is multiplied by a normalization coefficient . This form makes the mother wavelet have the second-order derivative Gaussian distribution characteristics of "protrusion in the center and depression on both sides" in the time domain, so it is more sensitive to sudden changes, peak-valley structures and partial symmetry changes. For the application of smart bracelets, this mother wavelet has obvious advantages in detecting short-term heart rate surges or instantaneous peaks in skin resistance; and if functions such as Morlet or Daubechies are added to the framework of multiple mother wavelets, irregular jitter, smooth main waves and a wider global situation can be taken into account. Through this joint transformation of multiple mother wavelets, a four-dimensional data structure will eventually be formed, that is, ,in Indicates the sensor number, represents the type of mother wavelet, Indicates scale, Indicates time translation. This multidimensional coefficient can play a key role in subsequent fractal analysis and chaos mapping: on the one hand, it provides high-resolution time-frequency information for calculating fractal dimension and energy distribution; on the other hand, the comparison of different mother wavelet coefficients can also reveal the complex pattern of multi-sensor coordinated changes. For example, if at the same time At a certain scale, different sensors If there are coefficient peaks on both mother wavelets, it may mean that the user has undergone intense exercise or emotional fluctuations during this period, causing the heart rate and skin resistance to increase synchronously; if a certain mother wavelet coefficient is significantly prominent while other mother wavelet coefficients are low, it means that the shape of the physiological signal is closer to a certain specific waveform. Through this complementary analysis, we can effectively break through the limitations of a single mother wavelet, so that the smart bracelet can still capture diverse subtle features when facing complex physiological information.

[0036] Embodiment 6: In step 2 and step 3, the fractal dimension of each sensor under each mother wavelet is calculated based on the wavelet coefficients by the following formula: ; in, For the sensor , mother wavelet In the broad index The fractal dimension of It is a generalized index and can take positive, negative or zero values ​​except 1. The singular spectrum is calculated by the following formula: ; in, for The corresponding fractal dimension.

[0037] Specifically, when multiple mother wavelet transform is used to obtain After that, we can superimpose different The energy distribution of the values ​​is quantified to quantify the overall energy distribution of each sensor under different wavelet functions and different scales, so as to finally extract and This is a set of key indicators that characterize the local irregularity and multi-fractal structure of the signal. Specifically, when the formula mentions In fact, the wavelet coefficients of each time translation are Then, we examine how this accumulation varies with the scale. The logarithmic rate of change during contraction. and The ratio between When the limit is , which is used to measure the complexity of the signal in different energy intervals or local distributions. If a sensor signal exhibits strong multi-scale characteristics under a mother wavelet, then in different The accumulated amount under the value will be significantly different, making Follow If the signal has a large change range, it often means that the signal has a high degree of irregularity in the local structure, that is, a large number of multi-scale fluctuations or mutations are hidden in the time domain. On the contrary, if the signal is relatively smooth and lacks multi-fractal characteristics, it will be The energy distribution under the value is not much different, resulting in The change curve is relatively gentle. It is worth noting that It can take not only positive values, but also negative values ​​or zero, thereby amplifying or suppressing the sensitivity to high or low amplitude coefficients, which allows researchers to conduct targeted multifractal analysis on different physiological signal characteristics (such as sudden changes in heart rate or small fluctuations in skin conductance). The calculation of right Perform partial derivative operations to reveal the distribution of signals at different singularities. Since fractal analysis often focuses on the local characteristics of signals corresponding to certain key points (such as the minimum singularity and the maximum global dimension), As an approximate expansion of the singular spectrum, it can help to understand the behavior of the signal under extreme energy distribution, such as when How to amplify the response to peak energy when the maximum value is taken. How to highlight the capture of weak components when taking the minimum. What is more practical is that smart bracelets often have to deal with a variety of different forms of physiological data in practice, such as blood oxygen, blood pressure, skin conductance or acceleration, etc. Each type of data may show significant local changes or multi-scale coupling phenomena in different time periods. After decomposition by multiple mother wavelets, the wavelet coefficients generated by each sensor and each mother wavelet at each scale can be used as input, and with the help of the fractal dimension and singular spectrum formula in this embodiment, it can help evaluate whether a physiological signal presents a relatively single scaling behavior, or whether there are obvious multiple scaling characteristics in both local peaks and slow-changing intervals. If the latter is true, it means that the user's physiological indicator has a richer dynamic change, which may mean a more sensitive physiological regulation or a potential stress response, and thus can be used as a high-risk or special concern object in subsequent health monitoring. In addition, in different Repeated calculation under the condition of (mother wavelet type) and , and can verify from multiple angles whether the multifractal properties of the signal are stable and consistent: if the fractal dimension and singular spectrum have similar changing trends under different mother wavelets, it means that the physiological signal does have relatively stable multi-scale complexity; otherwise, it may indicate that its multifractal characteristics are more caused by specific waveforms or specific frequency bands, which need to be further identified at the physical or physiological level. Furthermore, these fractal indicators can not only be used as a single static health assessment, but also track the user's status changes over time: once the fractal dimension curve of a sensor is found to rise sharply in a short period of time or the singular spectrum Abnormal fluctuations may indicate that the physiological state is developing in an abnormal direction, providing a basis for the health monitoring system to issue an early warning. At the same time, since the method of the present invention has smoothed, denoised and decomposed the signal by multiple mother wavelets in the previous steps, the fractal dimension and singular spectrum index finally calculated are not easily disturbed by short-term random noise, but more reflect the real complexity and local singularity carried by the signal itself. Through such fractal analysis, the multi-sensor signal is connected from the initial raw acquisition to the final identification of the health state, so that the smart bracelet can not only capture the performance of physiological signals at the smooth or statistical level, but also understand its deep laws at the nonlinear and multi-scale levels. This is of great significance for the detection of long-term chronic diseases, short-term abnormal events and personalized health management, because the fractal dimension and singular spectrum can essentially measure the abnormal details that are difficult to perceive by conventional statistical methods, and provide a higher-dimensional auxiliary decision-making basis for comprehensive health assessment.

[0038] Embodiment 7: In step 3, the fractal dimension and singular spectrum are used as coupling conditions to discretely iterate the cooperative state between sensors through the following formula: ; in, is the discrete iteration step When the sensor The corresponding collaborative state; is the local convergence coefficient set, controlling Towards target value The speed of convergence; For sensor Target value of is the discrete iteration step When the sensor The corresponding collaborative state.

[0039] Specifically, in the health monitoring method based on a smart bracelet described in the present invention, Example 7 incorporates the previously obtained fractal dimension and singular spectrum index into the evolution of the "cooperative state" between sensors through a discrete iterative formula, thereby realizing the dynamic fusion of each sensor signal in the multi-dimensional feature space. Specifically, with the extraction of wavelet coefficients, fractal dimensions and singular spectra in the previous steps, the system has mastered the multi-scale complexity information presented by different sensors under each type of mother wavelet. In this step, by introducing the so-called The smart bracelet can discretely iterate a quantitative index that can reflect the mutual influence and coupling relationship between sensors, and further judge the health mode or abnormal warning signal of the overall physiological signal. are assigned a current step size The collaborative state , it depends not only on a value called "target value" The reference value will also be affected by the additional impact of fractal feature differences. The target value here is often set based on the ideal or expected level of a physiological indicator of the sensor. For example, a heart rate sensor may have a medically referenced steady-state heart rate range, a skin resistance sensor may have a baseline level, and a blood oxygen sensor may have an optimal target that is adapted to comprehensive conditions such as altitude and exercise intensity. In this way, when the synergistic state With expected goals When the difference is large, the convergence coefficient in the iterative formula This causes the sensor's coordinated state to gradually move toward the target value, ensuring that the system can self-adjust to a known standard or base level.

[0040] However, if only If the sensor is only “close” to a target value, then this iteration does not fully reflect the shaping of the overall physiological signal by the mutual influence between sensors. With Qiyi Spectrum , an exponential coupling term is constructed to reflect the differences between different sensors at the multifractal level. The fractal dimension of another sensor When comparing the singular spectra of , if the two values ​​are close, it means that the two sensors show high consistency in some multi-scale features; if the difference is large, it means that there is a clear disagreement or complementary relationship between their descriptions of the same physiological signal. Based on this idea, a formula of the form , the negative sign and square operation in the exponent can make the function value approach zero as the numerical difference increases; when the numerical difference is small, the function value is higher, thus contributing a more obvious positive thrust in the collaborative state update. In other words, when each sensor updates its collaborative state, it will simultaneously refer to the fractal dimension of other sensors under the condition of multiple mother wavelets, and convert this "multiple fractal differences" into a coupling energy, which is accumulated to its own iteration result. If a sensor is similar to many other sensors in some fractal indicators, it will get more positive coupling "addition", making its own state converge faster to the overall consistent direction; if it deviates greatly from most other sensors in fractal indicators, it will not get much coupling energy in this iteration, so the collaborative state will be more "isolated" in terms of value, leaving room for further adjustment in subsequent iterations.

[0041] The physiological meaning of this coupling term is that different sensors often measure physiological information of different dimensions, such as blood oxygen, heart rate, skin conductance and acceleration. Although the physical principles of these signals are different, from the perspective of multi-scale analysis and fractals, they may also present similar multi-fractal structures due to the influence of certain physiological mechanisms or external environments. If two sensors both detect high-frequency fluctuations in a certain period of time and the fractal dimension is large, it means that the complexity of the two sensors increases synchronously during this period, indicating that the user may have experienced some strong stimulation or rapid movement changes. Through the exponential coupling term in the formula, this synchronous high complexity is quantified as a consistent improvement in the collaborative state of both parties, making it easier for the system to perceive that this is a real event rather than noise or failure of a single sensor. If in the next iteration, more sensors also show similar high-fractal features, it will continue to push the system to converge towards a clearer global signal pattern. From an algorithmic point of view, the discrete iterative formula The structure of implies a typical iterative update mechanism: each discrete step After the end, the sensor Assign a new coroutine state . After the system has iterated several times, It may gradually tend to a stable value, providing input for the next stage of health assessment or anomaly detection. Correspondingly, if some sensors cannot reach a high degree of consistency with other sensors at the fractal level for a long time, its collaborative state may be inconsistent with the target value. Or if the synergistic values ​​of other sensors differ greatly, this difference will be recorded and amplified in subsequent iterations in a timely manner, providing the health monitoring system with possible abnormal or diverse signal clues.

[0042] At the implementation level, This local convergence coefficient is one of the key hyperparameters in this example. It controls the convergence of the collaborative state to the target value. The speed of convergence - if If it is too large, it will be pulled closer quickly in a short time. and , which may result in the abandonment of some coupling information between sensors; if If it is too small, the system will need too many iterations to reach a stable state, thus affecting efficiency. Therefore, in different application scenarios, the sensitivity and real-time requirements of the monitored physiological indicators can be adjusted. For example, for real-time signals such as heart rate, the value can be increased appropriately to allow the system to capture abnormalities more quickly; for physiological signals that change slowly, such as blood pressure or body temperature, the value can be reduced appropriately. To avoid overcorrection caused by noise.

[0043] As for the target value The settings can also be dynamically adjusted in combination with statistical or medical standards. For example, in applications with significant circadian rhythms (sleep monitoring, nighttime blood oxygen monitoring, etc.), different settings may be used during the day and at night. to express the corresponding ideal interval; and when switching between exercise states, a set of benchmarks can be automatically replaced to prevent the user's heart rate or blood oxygen level during strenuous exercise from being too different from the target value in the quiet state, resulting in meaningless deviations in the collaborative state. Therefore, this discrete iteration of collaborative states is not only a mechanical numerical method, but can also be combined with personalized and contextualized health monitoring strategies to allow smart bracelets to interpret and analyze sensor data in more realistic conditions in different time periods or activity modes. In summary, the formula of Example 7 provides the present invention with a unified framework that incorporates advanced features such as fractal dimensions and singular spectra into a multi-sensor coupling mechanism. By discretely iterating the collaborative states between sensors, the system can finely identify similarities and differences between sensors in a high-dimensional feature space. Whether it is the slight difference in the numerical values ​​of the fractal dimension and the singular spectrum, or the target value Deviations from actual measured values ​​can be comprehensively reflected in this iterative process. As the number of iterations continues to advance, the system gradually reveals the overall health status or abnormal signs, which can not only correct single-point sensor abnormalities, but also respond sensitively to real multi-sensor changes. Through this coupling and coordination mechanism, the smart bracelet is no longer just a simple curve analysis of single-dimensional physiological parameters, but can go deep into the potential correlations between multiple scales, multiple angles, and multiple indicators, helping users or medical experts to more accurately capture the complex dynamic laws in health status.

[0044] Embodiment 8: By using the following formula, a chaotic map is added to the collaborative state, and the chaotic map is corrected using fractal difference to obtain the chaotic state between sensors: ; in, For discrete time steps When the sensor The chaotic state; For sensor The corresponding Logistic map growth rate; is a typical nonlinear term of the Logistic mapping; For discrete time steps When the sensor The chaotic state; The power exponent of the penalty term is set, which determines the sensitivity of the difference term; is the set fractal compensation coefficient; For a specific order and a specific mother wavelet The fractal dimension of and All are set values; For a specific order and mother wavelet The singular spectrum difference under .

[0045] Specifically, the chaotic state variable At discrete time steps The sensor reflects The current dynamic level of the logistic map is updated in the form of a typical logistic map, and multiple correction terms are embedded to adapt to the complexity of physiological signals in multi-sensor scenarios. This typical nonlinear operation is extremely sensitive to the initial value and can amplify the slight difference in the system, thus maintaining a high level of alertness to abnormal or unexpected events. However, if it is limited to the pure Logistic process, the chaotic behavior of the system will be difficult to combine with the fractal difference and personalized threshold in the multi-sensor coupling system of the smart bracelet. Therefore, in this embodiment, firstly add This penalty term is used to characterize the negative feedback of the gap between the coordinated state and the target value on the chaotic system. Its physical meaning is that when the coordinated state of a sensor deviates seriously from the ideal level, it means that the physiological signal of the sensor may have extreme or abnormal changes, and the system needs to suppress this deviation immediately to prevent it from being infinitely amplified in the chaotic process, thereby maintaining the robustness of the overall health monitoring. At the same time, the power exponent of the penalty term The sensitivity of this deviation is controlled. The larger the value, the more emphasis is placed on suppressing larger gaps. If the value is too small, the reaction to small fluctuations in the difference will be stronger. In addition to this negative correction term, the formula also includes This fractal compensation term is used to reflect the fractal dimension and singular spectrum at a specific order. and mother wavelet The positive impact of the difference under the chaotic system. When the difference between the fractal dimension and the singular spectrum value is larger, it often means that the signal is more significant in multi-scale and irregularity at this time, which may correspond to sudden movement, emotional fluctuations or sympathetic nerve activity; the system can "convert" this potential multi-scale complexity into positive compensation for chaotic variables by performing logarithmic operations on the difference. Since the chaotic system has the characteristic of exponentially amplifying small disturbances, when the fractal difference is large, this positive correction helps to highlight the nonlinear characteristics of the signal in the chaotic state, allowing the smart bracelet to more quickly notice the significant physiological changes that may be contained here during subsequent monitoring and judgment. For example, when the user's blood oxygen and heart rate increase rapidly during exercise, and the difference between the fractal dimensions of the two is significant, the fractal compensation term will significantly raise the chaotic state of the corresponding sensor, thereby providing a strong prompt signal for the overall health assessment module. If in the next iteration, the cooperative state of the sensor also shows a tendency to continue to deviate from the target value, then the chaotic state will stabilize or continue to climb under the interaction of the penalty term and the compensation term, to show the amplification and attention to this complex physiological state. For smart bracelets, such a design can strike a balance between maintaining sensitivity to minor anomalies and quickly locating large deviations. If there is no penalty term, the logarithmic compensation term may cause the system to blindly expand the chaotic effect when fractal differences are detected, resulting in unnecessary excessive alarms; if the fractal compensation term is not set, then the system will find it difficult to distinguish between ordinary small-scale disturbances and truly multi-scale characteristic signal changes, thereby losing the high sensitivity of the chaotic system. By organically combining fractal differences, cooperative state deviations and the Logistic model, Example 8 implements a chaotic state generation strategy that combines nonlinear amplification capabilities, noise tolerance and multi-dimensional physiological characteristic complementarity. The algorithm generates a chaotic state at each discrete step. Update each sensor When the user is exercising, feeling emotionally unstable, or experiencing sudden changes in stress, the judgment is made based on the real-time physiological signal characteristics, so that the user can quickly get more targeted prompts or alarms from the smart bracelet. On a deeper level, this chaotic mapping can also record the differential evolution of different sensors in terms of multi-fractal characteristics: if a sensor always maintains a fractal dimension or singular spectrum that is not synchronized with most sensors, it will also be reflected in the continuous iterations. The system can identify the potential abnormal evolution trend based on the numerical trend of the patient's health status, and then make further comprehensive judgments based on the synergistic state or global health threshold. This can not only help doctors or professional users capture complex physiological dynamics, but also provide more interpretable and sophisticated data support for individual multi-indicator self-monitoring.

[0046] Embodiment 9: Through the following formula, in step 4, the following partial differential equation is used to introduce a global health threshold of continuous evolution by summarizing the overall coordination state and chaotic state of the sensor, and ensure numerical stability: ; ; in, For sensor The historical mean of As the evolution time Changing global health thresholds; is the set cumulative gain coefficient; , is the integral variable; For The cooperative state calculated at this discrete time step; For The chaotic state calculated at this discrete time step; is the adjustment index; is the set threshold smoothness coefficient; : The threshold value in evolution time The second-order derivative diffusion term on is used to prevent excessive oscillation; is the initial threshold.

[0047] Specifically, in the health monitoring method based on the smart bracelet described in the present invention, the partial differential equation given in Example 9 provides a multi-dimensional coupling mechanism for realizing the continuous evolution of the global health threshold, so that the system can dynamically integrate high-dimensional information from the collaborative state and the chaotic state on the time axis, and at the same time A second-order derivative diffusion term is applied to avoid numerical divergence and excessive oscillation. The threshold function set here is It represents a benchmark or reference for the overall health level at a given moment, and it continuously obtains pull or inhibition signals from the latest "cooperative-chaotic" state of each sensor in order to adjust the threshold in real time in the context of changing physiological, environmental and individual differences. , the system will be at discrete time points (in Acting as an integral variable) obtains the sum of its cooperative state and chaotic state, and then adds it to the current threshold If the sensor average state is significantly higher than the threshold, Generate positive thrust to gradually raise it; conversely, when the status of most sensors is lower than the threshold, it will produce a negative pull on the threshold to adjust it down to match the overall actual level. This weighted form can be used according to different The value amplifies or weakens the attention to different discrete time steps. Larger, indicating that the system is more inclined to larger The key weighting of the interval is equivalent to focusing on the data in certain key periods; if If the value is too small, all intervals will have relatively balanced weights. This term is a typical diffusion or suppression partial differential operation, that is, when the threshold fluctuates violently on the evolution time axis, the second-order derivative term will have a smoothing or damping effect on it, so It will not cause large fluctuations or divergence due to a small amount of extreme data. The size of can make the threshold function strike a balance between sensitivity and stability. If is larger, it tends to implement more smoothing at the evolution time level to prevent the threshold from experiencing extreme jumps; if If the threshold is too small, it can more sensitively follow the changes in the sensor state without being suppressed too much. The above mechanism ensures that the system can correct the global threshold in a relatively flexible way when facing multi-source physiological data with high noise, so as to cope with complex situations such as individual short-term abnormalities, changes in motion scenes or interference from external conditions. As for the initial conditions of the equation, Give a feasible starting point, usually using the historical average of each sensor as a reference, so that the system can have a general understanding of the actual health indicator level at the initial moment. Such an initialization strategy can be set in combination with the user's physiological data over the past few days or longer periods of time. If the user is in a quiet state for a long time, the average values ​​of sensors such as heart rate, blood oxygen, and skin conductance are relatively stable, then the initial threshold will tend to a range that matches this state; if the user already has a higher or lower physical fitness and health benchmark, the system can also make adaptive corrections based on the historical average. Then, as time evolves Each iteration is based on the current sensor's collaborative state. and chaotic state Update the comprehensive mean of , and smoothed under the diffusion mechanism provided by the Laplace operator. This means that in the entire health monitoring process, the threshold is not fixed, but will respond flexibly to the slow or sudden changes in multiple physiological signal conditions: if most sensors show a significant improvement in status over a period of time, the system will quickly raise the threshold to avoid misjudging the normal high-level status as abnormal; conversely, if some sensors experience temporary fluctuations due to stress reactions or emergencies, while most other sensors remain stable, then in The threshold value will not fluctuate excessively under the pull of the sensor, and short-term anomalies can be regarded as local fluctuations and gradually digested during the evolution process. It is worth emphasizing that such a threshold evolution strategy can combine the cooperative state and chaotic state characteristics in the previous embodiment to obtain sensitivity to the multi-scale and nonlinear characteristics of the signal: when the fractal dimension and singular spectrum indicate that the sensor presents complex dynamics, the system may be able to be in a shorter time. This will lead to a larger increase (or decrease); when the sensor state returns to a stable state, the threshold will be gradually dragged to the new steady-state range by the diffusion term. This dynamic and continuous adjustment process is very necessary in the wearable device scenario, because users' daily activities are rich and varied, and their physical indicators often fluctuate dramatically in a fast-paced environment. If a fixed threshold is always used to determine health, it is easy to cause misjudgment or delayed alarm. By introducing evolution time By incorporating the correlation characteristics of multiple sensors of the smart bracelet into the equation, the system can maintain a good balance between noise and real signals, and distinguish short-term anomalies from true long-term trends. It can be seen that the partial differential equation of Example 9 constructs a high-dimensional dynamic control center for the smart bracelet at the principle level: at the front end, the collaborative state and the chaotic state respectively gather information from multi-source signals such as fractal dimension, singular spectrum and sensor target value; at the back end, this information is integrated into the integral term of the threshold evolution equation, and then the adaptive adjustment of the threshold is achieved through second-order derivative smoothing, thereby providing a reliable and flexible reference benchmark for subsequent health status judgment. By as well as By adjusting hyperparameters such as , the system can optimally match user data in multiple application modes: for example, focusing on fast sensitivity in short-term exercise monitoring, focusing on smooth tracking and noise robustness in chronic disease management, and avoiding frequent threshold jumps that interfere with analysis in nighttime sleep monitoring. Ultimately, this evolving threshold will be fully combined with the individual physiological characteristics of the user, drawing a more accurate and dynamic "health dividing line" for the overall assessment of health status by the smart bracelet, which can not only adapt to changes in user status in real time, but also track the trends of physical fitness and health levels over a longer period, allowing the health monitoring system to achieve a unity of refinement, personalization and high fault tolerance.

[0048] Example 10: Output the final health assessment result through the following formula: ; in, The upper boundary of the setting; is the lower boundary of the setting; is a discrete time step Health assessment results at the time of Discrete time step The global health threshold at ; is the discrete iteration step When the sensor The corresponding collaborative state; is the discrete iteration step When the sensor The corresponding chaotic state.

[0049] Specifically, in discrete time steps The system has already provided each sensor Iteratively calculates the coordination state and chaotic state . The collaborative state focuses on capturing the coupling relationship between multiple sensors under complex features such as fractal dimension and singular spectrum, and takes a certain target value or benchmark level as a reference; the chaotic state further introduces the concept of nonlinear mapping, and through the dynamic adjustment of fractal differences and collaborative deviations, it mines the subtle correlation between short-term fluctuations and long-term trends of signals. Adding the two and taking the average value on all sensors will give an instant indicator representing the user's global health measure; at the same time, through the partial differential equations introduced in the previous embodiment, the system evolves in time The global health threshold is also maintained in the dimension When these two quantities are compared, it is possible to assess whether the user's overall physiological state at the moment is better than, close to, or worse than the threshold. If it is far beyond the preset boundary or far below the preset bottom line, it can be distinguished into a more detailed health level.

[0050] In actual judgment, the formula defines four branches. The first case is "extremely healthy", which means that the sum of the cooperative state and chaotic state of all the user's current sensors is much higher than the system's expectation of the overall health level. It is an upper boundary threshold, which is used to distinguish between users in normal health and "peak zone" performance that exceeds the health benchmark. This situation often corresponds to the user having a strong physiological function state at the moment, such as steady breathing, ideal blood oxygen content, appropriate heart rate variability, etc.; it is also possible that after the exercise, the body's good stress response is stimulated and the state is high. By identifying this interval that clearly exceeds the threshold, the system can not only motivate users to continue to maintain their current healthy habits, but also give positive feedback on the health score.

[0051] The second case is "good", which indicates that the user's global health metric is and Between, that is, in a relatively normal and stable range: This segment covers most of the daily status: the user's physical indicators maintain a considerable synchronization with the global threshold; although there is no obvious over-performance, it will not fall out of the healthy range. This usually means that the user is in a stable and regular physiological range, such as normal blood oxygen, slightly increased heart rate or normal skin conductance signal, and there is no significant abnormal data. This interval judgment can provide users with a "daily good" safety signal, letting them know that they are not on the edge of risk and do not need to intervene deliberately.

[0052] The third situation is "sub-health", when the global health metric falls below When the value is below 0.00, it means that the overall physiological level observed by the sensor is significantly lower than the expected value. Possible reasons include: the user is tired for a long time, under great pressure, lack of sleep, etc.; it may also be an early pathological sign, such as a slight increase in body temperature but not yet reaching the fever standard, or a decrease in heart rate variability, slightly low blood oxygen, etc. By setting To capture this lower limit gap, the system can promptly issue a "sub-health" reminder to users, allowing them to be more vigilant and pay more attention to potential problems. If this deviation continues to intensify in subsequent iterations, it may also trigger further health intervention or early warning processes.

[0053] The last type is "abnormal". If numerical collapse or non-convergence occurs in the threshold evolution equation, the system will directly give an "abnormal" judgment. Numerical collapse or non-convergence means that In the process of evolution or in the calculation of collaborative or chaotic states, problems such as extreme oscillations, overflows, or null values ​​may occur. These problems may be hardware failures of sensors, data loss, or extreme abnormalities of user physiological indicators. If the system cannot even maintain its own reference threshold, it must jump out of the normal evaluation process and perform emergency processing or reminders to prompt users or equipment maintenance personnel to check whether there are major abnormalities in the hardware, software, or user health itself. Such a "protective" branch can prevent numerical problems in signal processing from being mistaken for extreme abnormalities or not being detected, and also provides users with a "last resort" risk assessment strategy.

[0054] From the perspective of algorithm principle, this four-level judgment formula largely reflects the direct comparison mode between the dynamic threshold and the average state of multiple sensors, subtracting the instantaneous mixed value of many physiological indicators from the current threshold, and setting upper and lower limits to further subdivide the health level. and After designing the operations of multiple differential coupling, fractal compensation and nonlinear amplification, the final judgment only needs to observe the resultant force and The distance between them not only simplifies the complexity of the judgment link itself, but also fully retains the system's information aggregation on multi-dimensional data. From the perspective of practical applications, this segmented judgment is usually combined with the UI design or alarm mechanism on the bracelet - for example, when the result is "extremely healthy", the mobile phone or bracelet interface can display green or blue and make incentive prompts; if it is "good", it can display more neutral colors and conventional prompts; if it is "sub-healthy", it will display yellow or orange, prompting users to pay attention to rest or supplement nutrition; if it is "abnormal", it will trigger a red warning, reminding users to conduct more in-depth examinations or even contact medical institutions. In addition, and These two boundaries can be flexibly set during deployment based on the user's personal basic data and actual needs. It may be set relatively high to prevent slightly higher physiological indicators from being considered "extremely healthy" and appearing frequently; for the elderly or people with certain chronic diseases, it is hoped that Don’t make it too small, so as to maintain a certain sensitivity to its relatively fragile signs. Of course, these boundary parameters can also be adjusted adaptively based on historical data and user feedback after the system has been running for a period of time. It can also be compared with physical time or discrete sampling time A certain mapping relationship is established to allow the system to periodically generate a health level determination result within the existing iteration cycle, helping users obtain real-time health trends.

[0055] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A health status monitoring method based on a smart bracelet, characterized in that: The method comprises: Step 1: The physiological signals collected by each sensor on the smart bracelet within a set time interval are regarded as a combination of several different frequency components, including bandpass noise, so that each physiological signal is represented by a unified time domain signal; Step 2: Evolve each time domain signal in a high-order tensor field. During the evolution process, use the Laplace operator to eliminate sharp or high-frequency noise to ensure local smoothness. At the same time, limit over-smoothing through the difference term with the time domain signal to obtain a smooth signal. Use multi-mother wavelet joint transform to perform multi-scale decomposition on the smooth signal to obtain wavelet coefficients. Step 3: Calculate the fractal dimension and singular spectrum of each sensor under each mother wavelet based on the wavelet coefficients; use the fractal dimension and singular spectrum as coupling conditions to discretely iterate the cooperative state between sensors; add chaotic mapping to the cooperative state, use fractal difference to correct the chaotic mapping, and obtain the chaotic state between sensors; Step 4: By summarizing the overall collaborative state and chaotic state of the sensor, a continuously evolving global health threshold is introduced and the numerical stability is ensured; according to the relative position of the health threshold and the average state of the sensor, it is divided into multiple levels and the final health assessment result is output; the average state of the sensor is defined as the weighted average of the chaotic state and collaborative state of the sensor.

2. The health status monitoring method based on the smart bracelet according to claim 1, characterized in that: The smart bracelet includes sensors, each sensor in the time interval The physiological signal is obtained by continuous sampling, and the time domain signal corresponding to the physiological signal is ;in, is the sampling time, and its value range is ; is the sensor serial number index; the sensor includes at least: a blood oxygen sensor, a skin resistance sensor, a temperature sensor, a heart rate sensor, a blood pressure sensor, and a respiratory rate sensor.

3. The health status monitoring method based on the smart bracelet according to claim 2, characterized in that: Time domain signal It is expressed by the following formula: ； in, is the number of frequency components; is an integer subscript index; For the frequency components in the sensor The sine amplitude on reflects the intensity of the sine component of the frequency component in the time domain signal; For the frequency components in the sensor The cosine amplitude on reflects the strength of the cosine component of the frequency component in the time domain signal; For the The frequency value of the frequency component; For the frequency components in the sensor The sinusoidal phase offset of For the frequency components in the sensor The cosine phase offset of Indicates the frequency range Integrate the noise signal within is the lower limit of the frequency range; is the upper limit of the frequency range; represents the bandpass noise, describing the The sensor is sampled at The frequency is The signal value generated by the continuous noise component.

4. The health status monitoring method based on the smart bracelet according to claim 3 is characterized in that: Assume that the high-order tensor field in step 2 is , when the time domain signal evolves in the high-order tensor field, the corresponding evolution equation is: ； in, is an integer subscript index, and ; is the difference operator; is the Laplace operator; is the evolution time, and the sampling time phase distinction; It is a high-order tensor field with a dimension of at least two dimensions and a range of real numbers; is a smooth signal.

5. The health status monitoring method based on the smart bracelet according to claim 4, characterized in that: In step 2, the smooth signal is decomposed into multiple scales using multi-mother wavelet joint transform to obtain wavelet coefficients using the following formula: ； in, Indicated in scale , translation , mother wavelet ,sensor The wavelet coefficients under ; is the scale, used to stretch or compress the wavelet function; is the translation amount on the time axis; is the mother wavelet number, there are wavelet functions of different types or shapes; Indicates the time interval Perform integration; For the The mother wavelet function is: 。 6. The health status monitoring method based on the smart bracelet according to claim 5, characterized in that: In step 2 and step 3, the fractal dimension of each sensor under each mother wavelet is calculated based on the wavelet coefficients by the following formula: ； in, For the sensor , mother wavelet In the broad index The fractal dimension of It is a generalized index and can take positive, negative or zero values ​​except 1. The singular spectrum is calculated by the following formula: ； in, for The corresponding fractal dimension.

7. The health status monitoring method based on the smart bracelet according to claim 6, characterized in that: In step 3, the fractal dimension and singular spectrum are used as coupling conditions to discretize the cooperative state between iterative sensors through the following formula: ； in, is the discrete iteration step When the sensor The corresponding collaborative state; is the local convergence coefficient set, controlling Towards target value The speed of convergence; For sensor Target value of is the discrete iteration step When the sensor The corresponding collaborative state.

8. The health status monitoring method based on the smart bracelet according to claim 7, characterized in that: By adding the chaotic map to the collaborative state and using the fractal difference to correct the chaotic map, the chaotic state between sensors is obtained: ； in, For discrete time steps When the sensor The chaotic state; For sensor The corresponding Logistic map growth rate; is a typical nonlinear term of the Logistic mapping; For discrete time steps When the sensor The chaotic state; The power exponent of the penalty term is set, which determines the sensitivity of the difference term; is the set fractal compensation coefficient; For a specific order and a specific mother wavelet The fractal dimension of and All are set values; For a specific order and mother wavelet The singular spectrum difference under .

9. The health status monitoring method based on the smart bracelet according to claim 8, characterized in that: Through the following formula, in step 4, the following partial differential equation is used to introduce a global health threshold of continuous evolution by summarizing the overall coordination state and chaotic state of the sensor and ensure numerical stability: ； ； in, For sensor The historical mean of As the evolution time Changing global health thresholds; is the set cumulative gain coefficient; , is the integral variable; For The cooperative state calculated at this discrete time step; For The chaotic state calculated at this discrete time step; is the adjustment index; is the set threshold smoothness coefficient; : The threshold value in evolution time The second-order derivative diffusion term on is used to prevent excessive oscillation; is the initial threshold.

10. The health status monitoring method based on the smart bracelet according to claim 9, characterized in that: The final health assessment result is output through the following formula: ； in, The upper boundary of the setting; The lower boundary of the setting; is a discrete time step Health assessment results at the time of Discrete time step The global health threshold at ; is the discrete iteration step When the sensor The corresponding collaborative state; is the discrete iteration step When the sensor The corresponding chaotic state.

Images