Traditional Chinese medicine pulse diagnosis data analysis method and system

Abstract

The invention belongs to the technical field of traditional Chinese medicine pulse diagnosis, and discloses a traditional Chinese medicine pulse diagnosis data analysis method and system, physiological noise influence factors are obtained by analyzing historical data, physiological noise in real-time pulse condition data is inhibited by using the factors, and a pulse diagnosis result is obtained on the basis of pulse condition characteristics of the pulse condition data after physiological noise inhibition. The method is used for evaluating the health risk type and the corresponding risk degree, physiological interference can be effectively coped with, and the pulse condition monitoring accuracy of sub-health people is remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of traditional Chinese medicine pulse diagnosis, and more specifically, to a method and system for analyzing traditional Chinese medicine pulse diagnosis data. Background Art

[0002] As a sub-health conditioning device, the RST-AI Wave Health Instrument's core technology lies in wave theory and frequency resonance technology in physics. The device aims to achieve health conditioning effects by emitting low-frequency sound waves and scanning the human body using the principle of frequency resonance. In order to more accurately treat users' sub-health symptoms, in-depth analysis of the user's pulse data is particularly important. Through pulse data analysis, the user's potential health risk types can be effectively identified, thereby providing users with more targeted health conditioning solutions. Therefore, the combined application of the RST-AI Wave Health Instrument and Traditional Chinese Medicine pulse diagnosis equipment has important clinical significance and application value.

[0003] With the rapid development of miniaturized and intelligent pulse diagnosis technology, home pulse monitoring has gradually become possible, providing solid technical support for the home use of the RST-AI pulse health instrument. However, pulse monitoring for sub-healthy people still faces many challenges. The pulse characteristics of sub-healthy people are often subtle and changeable, and are easily affected by daily physiological activities. For example, common physiological factors such as exercise, eating habits, and mood swings can cause physiological changes in the pulse. These physiological changes often mask or interfere with potential health risk signals in the pulse, making pulse monitoring and analysis in sub-healthy states particularly complex and difficult.

[0004] Currently, existing pulse diagnosis data analysis methods are designed to focus on identifying typical pulse characteristics in disease states, with the main purpose of assisting disease diagnosis by identifying specific pulse patterns. However, when these methods are applied to the pulse data analysis of sub-healthy people, their effectiveness is significantly challenged. Sub-healthy pulses usually have small fluctuation amplitudes, large individual differences, and strong physiological interference factors, which makes it difficult for traditional analysis methods to effectively distinguish normal physiological fluctuations from abnormal health risk signals. This lack of differentiation ability directly leads to a reduction in the accuracy and reliability of monitoring results, and thus cannot meet the actual needs of sub-healthy people for home health management.

[0005] In order to effectively improve the effectiveness of home pulse monitoring for sub-healthy people, it is urgent to develop a new pulse data analysis method. This method should be able to more accurately identify subtle pulse changes in sub-healthy people, while minimizing the interference caused by fluctuations in physiological status. More importantly, this new analysis method should not rely too much on the recognition of complex pulse patterns, but should suppress physiological interference from the source, thereby more clearly highlighting the pulse characteristics that truly reflect the health status. In this way, the user's potential health risk type can be more accurately judged, providing more effective health management methods for sub-healthy people. Therefore, it is particularly urgent to develop an innovative technical solution that can effectively deal with physiological interference and significantly improve the accuracy of pulse monitoring for sub-healthy people.

[0006] In view of the above problems, the existing technology is in urgent need of improvement. Summary of the Invention

[0007] The purpose of this application is to provide a method and system for analyzing traditional Chinese medicine pulse diagnosis data, which can effectively deal with physiological interference and significantly improve the accuracy of pulse monitoring in sub-healthy people.

[0008] In a first aspect, the present application provides a method for analyzing pulse diagnosis data in traditional Chinese medicine, the method comprising the following steps: A1. Obtain the user's historical monitoring data to analyze the correlation between changes in each physiological state parameter and each pulse data change, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; historical monitoring data includes historical pulse data and historical physiological state parameters collected simultaneously; pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and physiological state parameters include activity level, heart rate and respiratory rate; A2. Synchronously collect the user's real-time physiological status parameters and real-time pulse data; A3. According to the physiological noise factor, calculate the theoretical impact of each real-time physiological state parameter on each real-time pulse data; A4. Subtract the corresponding theoretical impact value from each real-time pulse data to obtain the pulse data after physiological noise suppression; A5. Based on the pulse data after physiological noise suppression, real-time pulse characteristics are extracted; the real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength changes, real-time pulse position offset degree and real-time irregular pulse degree; A6. Based on the extracted real-time pulse characteristics and combined with the user's historical pulse characteristics, the health risk type and corresponding risk level are assessed; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree, and historical pulse arrhythmia degree.

[0009] Preferably, step A1 includes: A101. The historical pulse data and historical physiological state parameters are preprocessed to obtain the preprocessed historical pulse data and preprocessed historical physiological state parameters; the preprocessing includes removing outliers, filling missing values ​​and data smoothing; A102. Calculate the mutual information value between each physiological state parameter and each pulse condition parameter using the mutual information method to obtain a preliminary physiological noise impact factor; A103. Use the t-test to evaluate whether each physiological state parameter has a significant effect on each pulse condition parameter, and classify the combination of each physiological state parameter and each pulse condition parameter into a significant effect combination and a non-significant effect combination. A104. Set the preliminary physiological noise impact factors corresponding to the non-significant impact combinations to zero, and keep the preliminary physiological noise impact factors corresponding to the significant impact combinations to obtain the final physiological noise impact factors.

[0010] Preferably, step A102 includes: Discretization processing is performed on various physiological state parameters and various pulse condition parameters based on the equal frequency binning method to obtain discrete physiological state parameters and discrete pulse condition parameters; According to the discretized physiological state parameters and the discretized pulse parameters, the mutual information calculation formula is used to calculate the mutual information value between each physiological state parameter and each pulse parameter, and the preliminary physiological noise impact factor is obtained.

[0011] Preferably, step A3 includes: A301. Based on the user's historical physiological state parameters and historical pulse data, calculate the user's individual physiological state parameter baseline value; A302. Perform difference calculation on the real-time physiological state parameter and the individual physiological state parameter baseline value to obtain the standardized real-time physiological state parameter; A303. For each item of real-time pulse data, calculate the product of each standardized real-time physiological state parameter and the corresponding physiological noise impact factor to obtain the theoretical impact value of each real-time physiological state parameter on this real-time pulse data.

[0012] Preferably, step A302 includes: Using the user's historical physiological state parameters, the partial correlation analysis method is used to analyze and quantify the mutual influence relationship between various physiological state parameters, and the partial correlation coefficient matrix between physiological parameters is obtained; constructing a physiological parameter correction matrix according to the partial correlation coefficient matrix between the physiological parameters; Subtract the individual physiological state parameter baseline value from the real-time physiological state parameter to obtain the preliminarily standardized real-time physiological state parameter; The preliminarily standardized real-time physiological state parameters are corrected according to the physiological parameter correction matrix to obtain final standardized real-time physiological state parameters.

[0013] Preferably, step A5 includes: A501 obtains the user's historical pulse cycle duration based on historical pulse data to determine the time window length, and divides the pulse data after physiological noise suppression into multiple pulse cycle data segments based on the time window length; A502. Calculate the average pulse rate of the pulse rate signal of each pulse cycle data segment to obtain a pulse rate sequence; A503. Calculate the difference between adjacent pulse rate values ​​in the pulse rate sequence to obtain a pulse rate difference sequence, and use the Bessel correction formula to calculate the standard deviation of the pulse rate difference sequence as the real-time pulse rate variability; A504. After performing wavelet transform on the pulse force signal of each pulse cycle data segment, the peak value of the pulse force signal is extracted to obtain a pulse force peak sequence, and the standard deviation of the pulse force peak sequence is calculated as the real-time pulse force intensity change; A505. Integrate the pulse position signal of each pulse cycle data segment and divide it by the pulse cycle duration to obtain the average value of the pulse position signal, form a pulse position average value sequence, and calculate the standard deviation of the pulse position average value sequence as the real-time pulse position deviation degree; A506. Detect the peak position of the pulse force signal of each pulse cycle data segment, determine the real-time pulse cycle length, obtain the real-time pulse cycle length sequence, and calculate the standard deviation of the real-time pulse cycle length sequence as the real-time arrhythmia degree.

[0014] Preferably, step A501 includes: Perform autocorrelation analysis on the user's historical pulse data to obtain an autocorrelation coefficient sequence, and determine the time delay corresponding to the first local maximum point of the autocorrelation coefficient sequence as the user's historical pulse cycle duration; Scaling the historical pulse cycle duration according to the predicted proportional coefficient to obtain the time window length; According to the length of the time window, the pulse data after physiological noise suppression is divided into multiple pulse cycle data segments using the sliding window method, and the step length of the sliding window is 1 / 2 of the time window length.

[0015] Preferably, step A6 includes: A601. Obtain a health risk assessment matrix; the rows of the health risk assessment matrix represent the health risk type, the columns represent the pulse characteristics, and the matrix elements represent the weights of different health risk types on different pulse characteristics; A602. Based on the extracted real-time pulse characteristics and the user's historical pulse characteristics, the user's comprehensive pulse feature vector is calculated. The elements of the comprehensive pulse feature vector include comprehensive pulse rate variability, comprehensive pulse strength changes, comprehensive pulse position offset degree and comprehensive arrhythmia degree; A603. Multiplying the comprehensive pulse characteristic vector with the health risk assessment matrix to obtain a health risk assessment vector; the elements of the health risk assessment vector represent risk scores for different health risk types; A604. Normalize the health risk assessment vector to obtain a normalized health risk assessment vector, determine the health risk type corresponding to the element with the largest value in the normalized health risk assessment vector as the user's health risk type, and determine the value of the element as the user's health risk level.

[0016] Preferably, step A602 includes: A clustering algorithm is used to divide the user's historical pulse characteristics into multiple pulse categories, and the cluster center of each pulse category is calculated to obtain a cluster center set of the user's historical pulse characteristics; Calculate the Euclidean distance between the extracted real-time pulse feature and each cluster center in the set of historical pulse feature cluster centers, and select the cluster center with the smallest Euclidean distance as the user's representative historical pulse feature; The extracted real-time pulse features are weightedly fused with representative historical pulse features to obtain the user's comprehensive pulse feature vector.

[0017] In a second aspect, the present application provides a TCM pulse diagnosis data analysis system, the system comprising a monitoring device and a pulse diagnosis and analysis device, the monitoring device and the pulse diagnosis and analysis device being communicatively connected; The monitoring device includes a pulse diagnosis sensor and a physiological state sensor, wherein the pulse diagnosis sensor is used to collect the user's pulse data, and the physiological state sensor is used to synchronously collect the user's physiological state parameters; the pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and the physiological state parameters include activity level, heart rate and respiratory rate; The pulse diagnosis and analysis device comprises: A receiving module is used for receiving the real-time physiological state parameters and real-time pulse data collected by the monitoring device in real time; The impact factor evaluation module is used to obtain the user's historical monitoring data, analyze the correlation between the changes in each physiological state parameter and the changes in each pulse data, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; the historical monitoring data includes the historical pulse data and historical physiological state parameters collected synchronously; An impact degree evaluation module is used to calculate the theoretical impact value of each real-time physiological state parameter on each real-time pulse data according to the physiological noise impact factor; The noise suppression module is used to subtract the corresponding theoretical impact value from each real-time pulse data to obtain the pulse data after physiological noise suppression; A feature analysis module is used to extract real-time pulse characteristics based on the pulse data after physiological noise suppression; the real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength change, real-time pulse position deviation degree and real-time pulse arrhythmia degree; The risk assessment module is used to assess the health risk type and corresponding risk level based on the extracted real-time pulse characteristics combined with the user's historical pulse characteristics; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree and historical pulse arrhythmia degree.

[0018] Beneficial effects: The present application provides a method and system for analyzing TCM pulse diagnosis data, which obtains the physiological noise influencing factor by analyzing historical data, and uses the factor to suppress the physiological noise in real-time pulse data. The pulse characteristics of the pulse data after physiological noise suppression are then used to evaluate the health risk type and the corresponding risk level. This method can effectively deal with physiological interference and significantly improve the accuracy of pulse monitoring for sub-healthy people. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of the TCM pulse diagnosis data analysis method provided in an embodiment of the present application.

[0020] Figure 2 This is a schematic diagram of the structure of the Traditional Chinese Medicine pulse diagnosis data analysis system provided in an embodiment of the present application.

[0021] Explanation of reference numerals: 1. Monitoring device; 101. Pulse diagnosis sensor; 102. Physiological status sensor; 2. Pulse diagnosis analysis device; 201. Receiving module; 202. Impact factor evaluation module; 203. Impact degree evaluation module; 204. Noise suppression module; 205. Feature analysis module; 206. Risk assessment module. DETAILED DESCRIPTION

[0022] The technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0023] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0024] refer to Figure 1 , this application proposes a method for analyzing pulse diagnosis data in traditional Chinese medicine, which includes the following steps: A1. Obtain the user's historical monitoring data to analyze the correlation between changes in each physiological state parameter and each pulse data change, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; historical monitoring data includes historical pulse data and historical physiological state parameters collected simultaneously; pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and physiological state parameters include activity level, heart rate and respiratory rate; A2. Synchronously collect the user's real-time physiological status parameters and real-time pulse data; A3. Calculate the theoretical impact of each real-time physiological state parameter on each real-time pulse data based on the physiological noise factor; A4. Subtract the corresponding theoretical impact value from each real-time pulse data to obtain the pulse data after physiological noise suppression; A5. Based on the pulse data after physiological noise suppression, real-time pulse characteristics are extracted; real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength changes, real-time pulse position deviation degree and real-time irregular pulse degree; A6. Based on the extracted real-time pulse characteristics and combined with the user's historical pulse characteristics, the health risk type and corresponding risk level are assessed; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree, and historical pulse arrhythmia degree.

[0025] In step A1, historical monitoring data can be read from a database. The historical monitoring data can include pulse data and physiological state parameters of the user under different physiological states over a period of time. For example, the pulse, heart rate, respiratory rate, and other data of the user under different states such as resting state, exercise state, and after eating can be recorded. By analyzing these historical data, a statistical relationship model between changes in physiological state parameters and changes in pulse data can be established, and the degree of influence of physiological noise on pulse can be quantified to obtain a physiological noise impact factor.

[0026] Wherein, in step A2, can adopt pulse sensor and physiological state sensor synchronous data acquisition to realize, pulse sensor can be non-contact pulse sensor or contact pressure sensor, and physiological state sensor can be heart rate sensor, respiratory belt or motion sensor etc. Synchronous acquisition guarantees the consistency of real-time pulse data and physiological state parameters in time, and provides data basis for subsequent noise suppression and feature extraction.

[0027] Wherein, among the step A3, can adopt the method that real-time physiological state parameter and corresponding physiological noise influence factor are multiplied each other to realize, for example, if heart rate is 0.2 to the physiological noise influence factor of pulse rate, and real-time heart rate is 70 beats / minute, then heart rate is 70*0.2=14 beats / minute to the theoretical influence value of pulse rate.Through the calculation theory influence value, can predict the noise interference degree that current physiological state produces pulse condition.

[0028] Wherein, in step A4, can adopt subtraction operation to realize, for example, if real-time pulse rate data is 80 times / minute, and the theoretical influence value of heart rate on pulse rate is 14 times / minute, then the pulse rate data after physiological noise suppression is 80-14=66 times / minute; If multiple real-time physiological state parameters all have theoretical influence values ​​to pulse rate, then need to deduct all these theoretical influence values ​​with real-time pulse rate data. By deducting the theoretical influence values, can effectively suppress the physiological noise component in the pulse data, retain purer pulse signal.

[0029] Wherein, among the step A5, can adopt signal processing and feature extraction algorithm to realize, for example, pulse rate variability can obtain by calculating the standard deviation of pulse cycle sequence, the pulse strength variation can obtain by the amplitude variation of analyzing pulse waveform, the pulse position offset degree can obtain by the baseline drift of analyzing pulse waveform, and the arrhythmia degree can obtain by detecting the irregularity of pulse cycle.The real-time pulse characteristics of extraction can objectively quantify the dynamic change and the characteristic information of pulse.

[0030] Among them, in step A6, machine learning and pattern recognition algorithms can be used to implement it. For example, a health risk assessment model can be established, which uses real-time pulse characteristics and historical pulse characteristics as input and outputs the health risk type and risk level. The historical pulse characteristics can provide the user's baseline pulse level and long-term change trend. Combined with the real-time characteristics, health risk assessment can be made more comprehensive and individualized.

[0031] Specifically, this application provides a Traditional Chinese Medicine (TCM) pulse diagnosis data analysis method designed to address the issue of pulse data in sub-healthy individuals being susceptible to physiological noise interference, thereby enabling more accurate health risk assessments. First, in step A1, the system learns the user's historical monitoring data, analyzes the correlation between physiological state parameters (such as activity level, heart rate, and respiratory rate) and pulse data (such as pulse rate, pulse force, pulse position, and pulse rhythm), quantifies the impact of physiological noise on the pulse, and obtains a physiological noise impact factor. The key to this step lies in personalized learning, which can capture the specific impact of an individual's physiological state on the pulse. Next, in step A2, the system synchronously collects the user's real-time physiological state parameters and real-time pulse data, providing real-time data for subsequent noise suppression. Step A3 uses the physiological noise impact factor obtained in step A1 to predict the theoretical impact value of each real-time physiological state parameter on each item of real-time pulse data, thereby estimating the degree of noise interference caused by the current physiological state. The core step A4 subtracts the theoretical impact value from the real-time pulse data to achieve physiological noise suppression, attempting to remove components in the pulse data caused by physiological state fluctuations and retain a purer pulse signal. On the basis of noise suppression, step A5 extracts the real-time pulse characteristics in the pulse data after physiological noise suppression, including objective indicators such as real-time pulse rate variability, real-time pulse strength change, real-time pulse position deviation degree and real-time arrhythmia degree, which fully reflect the dynamic changes and characteristic information of pulse. Finally, step A6 combines the real-time pulse characteristics with the user's historical pulse characteristics to carry out health risk type and risk degree assessment. Introducing historical pulse characteristics can provide the user's baseline pulse level and long-term change trend, making health risk assessment more comprehensive and individualized. Through the above steps, the application has constructed a complete set of traditional Chinese medicine pulse diagnosis data analysis method, effectively reduces the interference of physiological noise to pulse analysis, and promotes the accuracy and reliability of health risk assessment for sub-healthy people.

[0032] Through the above technical solution, the present application can effectively reduce the interference of physiological noise on pulse data analysis, so that the pulse data of sub-healthy people can more accurately reflect their health status, thereby realizing accurate identification of health risk types of sub-healthy people and effective assessment of risk levels.

[0033] Preferably, step A1 includes: A101. Preprocess the historical pulse data and historical physiological status parameters to obtain preprocessed historical pulse data and preprocessed historical physiological status parameters; preprocessing includes removing outliers, filling missing values, and data smoothing; A102. Calculate the mutual information value between each physiological state parameter and each pulse condition parameter using the mutual information method to obtain a preliminary physiological noise impact factor; A103. Use the t-test to evaluate whether each physiological state parameter has a significant effect on each pulse condition parameter, and classify the combination of each physiological state parameter and each pulse condition parameter into a significant effect combination and a non-significant effect combination. A104. Set the preliminary physiological noise impact factors corresponding to the non-significant impact combinations to zero, and keep the preliminary physiological noise impact factors corresponding to the significant impact combinations to obtain the final physiological noise impact factors.

[0034] Among them, in step A101, the preprocessing operation can include a variety of specific methods. For example, the removal of outliers can use the box plot method or the Z-score standardization method to identify and eliminate data points that are beyond a reasonable range; filling missing values ​​can use mean filling, median filling or linear interpolation methods to fill in data gaps; data smoothing can use moving average filtering, Savitzky-Golay filtering or wavelet denoising techniques to reduce data noise and highlight data trends.

[0035] Among them, step A102 is intended to preliminarily quantify the degree of influence of the physiological state parameters on the pulse parameters. The mutual information method, as an information theory method, can effectively capture the nonlinear correlation between variables. During specific implementation, the physiological state parameters and pulse parameters can be discretized first, for example, by using equal-width binning or equal-frequency binning methods to convert continuous variables into discrete variables, and then based on the discretized data, the mutual information value is calculated using the mutual information calculation formula. The numerical value of the mutual information value directly reflects the degree of correlation between the physiological state parameters and the pulse parameters. The larger the numerical value, the higher the correlation.

[0036] Wherein, step A103 is used to distinguish the significant influence of physiological state parameters on pulse parameters and non-significant influence. As a statistical hypothesis testing method, t-test can evaluate whether there is a significant difference between the mean values ​​of two groups of samples. In this solution, t-test is used to evaluate whether there is a significant change in the mean value of pulse parameters under different physiological state parameter conditions. By setting the significance level (e.g., p<0.05), it can be judged whether the influence of physiological state parameters on pulse parameters has statistical significance (the specific process is prior art and will not be described in detail here). Parameter combinations are divided into significant influence combinations and non-significant influence combinations, which provide a basis for the screening of subsequent noise influencing factors. Wherein, each significant influence combination and non-significant influence combination all include a physiological state parameter and a pulse data. The physiological state parameters in the same significant influence combination have a significant influence on the pulse data, and the physiological state parameters in the same non-significant influence combination do not have a significant influence on the pulse data.

[0037] Wherein, among the step A104, by the preliminary physiological noise influence factor that non-significant influence combination corresponds is set to zero, can effectively get rid of the interference of the unobvious physiological state parameter to the pulse parameter influence, retain the physiological noise influence factor that really has significant impact.This processing mode can promote the accuracy and the reliability of the physiological noise influence factor, for follow-up noise suppression provides more accurate parameter.Step A101 is carried out data pre-processing, and step A102 is carried out mutual information calculation, and step A103 is carried out significance test, and step A104 is carried out influence factor screening, by above-mentioned link, from coarse to fine, from the outside to the inside, improved the accuracy and the reliability of the physiological noise influence factor assessment, for follow-up realization accurate pulse data analysis has laid the foundation.

[0038] Specifically, this scheme is in order to obtain the physiological noise influence factor more accurately, thereby promotes the accuracy of follow-up pulse data analysis.First, by step A101, historical pulse data and historical physiological state parameters are pre-processed, this comprises removing abnormal values, filling missing values ​​and data smoothing, the purpose is to improve data quality, reduce the interference of data own problems to subsequent analysis, for follow-up accurately calculates the physiological noise influence factor and lays the foundation.Then, in step A102, adopt mutual information method to calculate the mutual information value between each physiological state parameter and each pulse parameter, mutual information can effectively measure the nonlinear correlation between variables, thereby preliminarily quantifies the potential influence degree of the physiological state parameter to the pulse parameter, obtains preliminary physiological noise influence factor.Further, in order to distinguish the significant influence of the physiological state parameter to the pulse parameter and the non-significant influence, in step A103, introduce the t test.By t test, assess whether each physiological state parameter has the significance of statistical significance to the influence of each pulse parameter, and parameter combination is divided into the significant influence combination and the non-significant influence combination. At last, in step A104, to the non-significantly affected combination, the preliminary physiological noise influence factor of its correspondence is set to zero, and for the significantly affected combination, then retain its preliminary physiological noise influence factor as final physiological noise influence factor.By setting the influence factor of the non-significantly affected combination to zero, can effectively get rid of those interferences of the unobvious physiological state parameters to the pulse parameter influence, thereby make the final physiological noise influence factor that obtains can more accurately reflect the real physiological noise with significant impact, and then promote accuracy and the validity of follow-up physiological noise suppression.Step A1 has realized the optimization of physiological noise influence factor assessment through links such as data pre-processing, mutual information calculation, significance test and influence factor screening.

[0039] Through the above technical solution, the present application can more accurately obtain the physiological noise impact factor, improving the accuracy of subsequent pulse data analysis. Through data preprocessing, the quality of input data is guaranteed; through the mutual information method, the nonlinear correlation between physiological state parameters and pulse parameters is effectively quantified; through the t-test, significant and non-significant effects are distinguished, eliminating the interference of non-significant physiological noise; through the influence factor screening, a more accurate physiological noise impact factor is obtained, laying the foundation for subsequent physiological noise suppression.

[0040] Preferably, step A102 may include: Discretization processing is performed on various physiological state parameters and various pulse condition parameters based on the equal frequency binning method to obtain discrete physiological state parameters and discrete pulse condition parameters; According to the discretized physiological state parameters and the discretized pulse parameters, the mutual information calculation formula is used to calculate the mutual information value between each physiological state parameter and each pulse parameter, and the preliminary physiological noise impact factor is obtained.

[0041] The equal-frequency binning method divides a numerical variable into multiple bins, ensuring that the data frequency within each bin is roughly equal. This can be achieved using the following method: First, determine the number of bins required, for example, 5 or 10. Then, sort the historical data for physiological status parameters and pulse parameters. Then, based on the number of bins and the total amount of data, calculate the number of data points that should be included in each bin. Finally, assign the data points to the bins in order of sorting, ensuring that the data frequency within each bin is roughly equal. Using the equal-frequency binning method, continuous physiological status parameters and pulse parameters can be converted into discrete variables for subsequent mutual information calculation. This ensures that the amount of data in each bin is roughly the same, avoiding information bias caused by uneven data distribution and improving the representativeness of the discretized data. Compared to directly calculating mutual information using continuous data, using discretized data significantly reduces the computational effort, improves computational efficiency, and meets real-time requirements.

[0042] Wherein, discretized physiological state parameter and discretized pulse condition parameter refer to the parameters obtained after being processed by the equal frequency binning method. Physiological state parameter such as activity level, heart rate and respiratory rate, and pulse condition parameter such as pulse rate, pulse force, pulse position and pulse rhythm, after being processed by the equal frequency binning method, are all converted into discrete numerical value or category. For example, the heart rate parameter is originally a continuous heartbeat value, but after binning, it can be divided into several discrete intervals, such as "low heart rate", "medium heart rate", "high heart rate" etc. Pulse condition parameter also carries out similar processing, thereby continuous pulse condition data is converted into discrete pulse condition categories.

[0043] The mutual information formula is used to calculate the mutual dependence between two random variables. Specifically, this can be achieved using the Shannon mutual information formula, which is expressed as: I(X;Y) = ∑∑p(x,y)log(p(x,y) / (p(x)p(y))), where X and Y represent the discretized physiological state parameters and the discretized pulse parameters, respectively. p(x,y) is the joint probability distribution of X and Y, p(x) and p(y) are the marginal probability distributions of X and Y, respectively. x represents a specific value of the discretized physiological state parameter, and y represents a specific value of the discretized pulse parameter. In actual calculations, the corresponding probability values ​​can be estimated by statistically analyzing the joint frequency distribution and marginal frequency distribution of the discretized physiological state parameters and pulse parameters, and then calculating the mutual information value. A larger mutual information value indicates a higher correlation between the physiological state parameters and the pulse parameters, that is, a larger physiological noise impact factor.

[0044] Wherein, preliminary physiological noise influence factor refers to the quantitative value of the degree of influence of the physiological state parameters on the pulse parameters calculated by the mutual information method. This factor is "preliminary" because it will be revised through the t-test and significance assessment of steps A103 and A104 to obtain the final physiological noise influence factor. The preliminary physiological noise influence factor reflects the potential influence of the physiological state parameters on the pulse parameters before statistical significance is considered.

[0045] Specifically, to enhance the effectiveness of the pulse data analysis method for home pulse monitoring in sub-healthy individuals, step A102 is proposed to more accurately assess the impact of physiological noise on pulse data. In step A102, first, the historical physiological state parameters and historical pulse data are discretized using the equal-frequency binning method. The advantage of this processing method is that it can convert continuous variables into discrete variables, simplifying the complexity of the subsequent mutual information calculation. At the same time, by ensuring that the frequency of each bin data is roughly equal, the impact of the uneven distribution of the original data can be reduced, reducing the interference of outliers and noise on the mutual information calculation results, and improving the robustness of the calculation results. Then, based on the discretization process, the mutual information calculation formula is used to calculate the mutual information value between each physiological state parameter and each pulse parameter. The mutual information value can effectively measure the nonlinear correlation between the physiological state parameters and the pulse parameters, more comprehensively reflecting the degree of influence of physiological noise on the pulse. Through step A102, a preliminary physiological noise impact factor can be obtained, laying the foundation for effectively suppressing physiological noise from the pulse data in the future, thereby more accurately extracting pulse characteristics reflecting the health status of the human body.

[0046] Through the above technical solutions, the present application can more accurately and reliably assess the degree of influence of physiological noise on pulse data. By introducing the equal frequency binning method before the mutual information calculation, the interference of uneven data distribution and outliers on the mutual information calculation results is reduced, and the accuracy of the physiological noise impact factor assessment is improved. Thus, for the subsequent more effective suppression of physiological noise in pulse data, a strong guarantee is provided for improving the accuracy and effectiveness of pulse diagnosis data analysis.

[0047] In some embodiments, step A3 comprises: A301. Based on the user's historical physiological state parameters and historical pulse data, calculate the user's individual physiological state parameter baseline value; A302. Perform difference calculation on the real-time physiological state parameter and the individual physiological state parameter baseline value to obtain the standardized real-time physiological state parameter; A303. For each item of real-time pulse data, calculate the product of each standardized real-time physiological state parameter and the corresponding physiological noise impact factor to obtain the theoretical impact value of each real-time physiological state parameter on this real-time pulse data.

[0048] In step A301, the individual physiological state parameter baseline value can be calculated as the average value of the user's historical physiological state parameters. For example, for the heart rate parameter, the average value of the user's heart rate monitoring data over the past week can be taken as the heart rate baseline value.

[0049] In step A302, the normalization process can be achieved by subtracting the corresponding baseline value from the real-time physiological state parameter, so as to obtain the offset of the physiological parameter relative to the individual's normal level.

[0050] Wherein, among the step A303, the calculating of theoretical influence value is realized by multiplication operation, and promptly the real-time physiological state parameter after the standardization and the predetermined physiological noise influence factor are multiplied each other, obtain the concrete influence numerical value of each physiological state parameter to each pulse data.The physiological noise influence factor has characterized the influence degree of the physiological state parameter variation to the pulse data, and this factor has been determined in previous step.Through above-mentioned steps, theoretical influence value can more accurately reflect the individualized physiological noise level, for follow-up noise suppression provides more accurate basis.

[0051] Specifically, in the pulse diagnosis data analysis process, in order to more accurately assess the impact of physiological noise on pulse data, first step A301 is performed to calculate the individual physiological state parameter baseline value. For example, the user's historical activity volume, heart rate and respiratory rate data for the past month can be collected, and the average value of each physiological parameter is calculated respectively, and these average values ​​are used as the user's individual activity volume baseline value, heart rate baseline value and respiratory rate baseline value. Thus, the individualized normal reference level of the physiological parameter is determined. Then, in step A302, when the user's physiological state parameters are monitored in real time, for example, if the user's current heart rate is measured to be 75 beats / minute and the heart rate baseline value is 70 beats / minute, then by performing a difference operation, the real-time heart rate parameter obtained after standardization is 5 beats / minute, indicating that the real-time heart rate has increased by 5 beats / minute relative to the individual baseline. This standardization process effectively eliminates the differences in the absolute values ​​of the physiological parameters between individuals and highlights the relative changes of the physiological parameters. At last, in step A303, to real-time pulse data, for example real-time pulse rate data, with the standardized real-time physiological state parameter that obtains in the step A302, as standardized real-time heart rate parameter 5 beats / minute, multiplied each other with the physiological noise influence factor that obtains in advance.Suppose that heart rate is 0.2 to the physiological noise influence factor of pulse rate, then calculating heart rate is 5*0.2=1 to the theoretical influence value of real-time pulse rate data.This numerical value has represented that current heart rate can cause pulse rate data to produce numerical value be 1 fluctuation with respect to the variation of baseline in theory.In this way, can quantize the theoretical influence value of each real-time physiological state parameter to each real-time pulse data, for follow-up deducts these theoretical influence values ​​from real-time pulse data, realizes that physiological noise suppresses and gets ready.

[0052] In some specific embodiments, among the step A301, individual physiological state parameter baseline value can also be calculated based on the historical physiological state parameter and the historical pulse data of the user by statistical methods such as weighted average, median or mode.For example, when calculating the heart rate baseline value, according to the quality of the historical pulse data, the historical heart rate data can be weighted, the heart rate data of the synchronous collection of high-quality pulse data are given higher weight, and the heart rate data of the synchronous collection of low-quality pulse data are given lower weight, then weighted average is carried out, and more reliable heart rate baseline value is obtained. As a preferred embodiment, among the step A302, the real-time physiological state parameter after the standardization can also be obtained by calculating the ratio of real-time physiological state parameter and individual physiological state parameter baseline value, or by methods such as Z-score standardization, to adapt to different data analysis needs.Among the step A303, the calculation of the theoretical influence value can also introduce nonlinear model, and for example, when the variation range of the physiological state parameter was larger, the physiological noise influence factor can be carried out adaptive adjustment according to the variation range, to more accurately describe the nonlinear influence of the physiological state parameter on the pulse data.Thus, the flexibility and the accuracy of the theoretical influence value calculation can be promoted.

[0053] Preferably, step A302 may include: Using the user's historical physiological state parameters, the partial correlation analysis method is used to analyze and quantify the mutual influence relationship between various physiological state parameters, and the partial correlation coefficient matrix between physiological parameters is obtained; According to the partial correlation coefficient matrix between physiological parameters, a physiological parameter correction matrix is ​​constructed; Subtract the individual physiological state parameter baseline value from the real-time physiological state parameter to obtain the preliminarily standardized real-time physiological state parameter; The real-time physiological state parameters after preliminary standardization are corrected according to the physiological parameter correction matrix to obtain the final normalized real-time physiological state parameters.

[0054] By collecting historical physiological status data from users, partial correlation analysis, a statistical method, is used to deeply explore and quantify the degree of mutual influence between various physiological status parameters. Partial correlation analysis focuses on examining the correlation between two variables after excluding the influence of one or more other variables. In this approach, partial correlation analysis is used to assess the net correlation between any two physiological parameters after excluding the influence of other physiological parameters. Through calculation, a partial correlation coefficient matrix between physiological parameters is obtained. Each element in this matrix is ​​a partial correlation coefficient. The numerical value represents the quantitative value of the mutual influence between the two physiological parameters, and the sign indicates the direction of influence. Specifically, partial correlation analysis can be implemented using statistical analysis software such as SPSS, R, and Python, or relevant function libraries in programming languages, such as the `statsmodels` library in Python or the `pcor` function in R. Historical physiological status parameters can be obtained from users' daily monitoring data, physical examination reports, or medical records.

[0055] The purpose of constructing the correction matrix is ​​to be able to adjust the preliminary standardization results according to the mutual influence relationship between the parameters in the subsequent standardization process to improve the accuracy of the standardization results. There are many ways to construct the correction matrix. For example, the partial correlation coefficient matrix can be directly used as the correction matrix, or the partial correlation coefficient matrix can be further processed by normalization, threshold processing, or function transformation to obtain the final correction matrix. As a preferred method, the absolute value of the partial correlation coefficient matrix can be normalized to ensure that the numerical range of the correction matrix is ​​between [0,1] to avoid over-correction.

[0056] Subtracting the individual physiological state parameter baseline value from the real-time physiological state parameter to obtain the preliminarily normalized real-time physiological state parameter refers to performing a basic normalization operation before obtaining the physiological parameter correction matrix. This step is consistent with the basic normalization method described in the previous scheme, that is, for each real-time physiological state parameter collected, the pre-calculated individual physiological state parameter baseline value of that parameter is subtracted to eliminate the impact of individual baseline differences on the real-time physiological state parameter, thereby obtaining the preliminarily normalized data.

[0057] Among them, according to the physiological parameter correction matrix, the real-time physiological state parameters after preliminary standardization are corrected to obtain the final standardized real-time physiological state parameters, which means that after the completion of the preliminary standardization, in order to further improve the accuracy of the standardization results, the physiological parameter correction matrix is ​​introduced to perform a secondary correction on the preliminary standardization results. The correction process is specifically: using the constructed physiological parameter correction matrix, combined with the real-time physiological state parameters after preliminary standardization, matrix operations or weighted calculations are performed to achieve the correction of the preliminary standardization results. Through correction, the standardization deviation caused by the mutual influence between physiological parameters can be weakened, and more accurate final standardized real-time physiological state parameters can be obtained. For example, the real-time physiological state parameter vector after preliminary standardization can be multiplied by the physiological parameter correction matrix to obtain a corrected physiological state parameter vector as the final standardization result.

[0058] Specifically, this solution aims to provide a more accurate method for normalizing real-time physiological state parameters. When normalizing real-time physiological state parameters, the solution first subtracts the individual's baseline physiological state parameter value from the real-time physiological state parameter to complete preliminary normalization, preliminarily eliminating the influence of individual physiological baseline differences. However, the key improvement of this solution lies in its recognition that physiological state parameters do not exist in isolation but rather interact with each other in complex ways. To quantify these interactions, the solution utilizes the user's historical physiological state parameters and employs partial correlation analysis to calculate a partial correlation coefficient matrix between physiological parameters. This matrix effectively reflects the degree and direction of the net correlation between any two physiological parameters, excluding the influence of other parameters. Subsequently, based on this partial correlation coefficient matrix, a physiological parameter correction matrix is ​​constructed for subsequent correction steps. After obtaining the preliminary normalization results, the solution does not directly use them as the final result. Instead, it further corrects the preliminary normalized data using the physiological parameter correction matrix. This correction process accounts for the interactions between physiological parameters and effectively corrects for standardization deviations arising from these interactions. Through this secondary correction mechanism, the resulting standardized real-time physiological state parameters can more accurately reflect the user's true physiological state, providing a higher-quality data foundation for subsequent pulse data analysis. This refined standardization process can improve the accuracy and reliability of subsequent pulse feature extraction and health risk assessment.

[0059] Through the above technical solution, the present application can more accurately standardize the real-time physiological status parameters, fully consider the mutual influence relationship between various physiological status parameters, effectively eliminate the standardization deviation caused by the mutual influence between parameters, thereby obtaining more accurate physiological status parameter standardization results, and providing more reliable basic data for subsequent pulse data analysis, thereby improving the accuracy and effectiveness of traditional Chinese medicine pulse diagnosis data analysis.

[0060] In some embodiments, step A5 comprises: A501 obtains the user's historical pulse cycle duration based on historical pulse data to determine the time window length, and divides the pulse data after physiological noise suppression into multiple pulse cycle data segments based on the time window length; A502. Calculate the average pulse rate of the pulse rate signal of each pulse cycle data segment to obtain a pulse rate sequence; A503. Calculate the difference between adjacent pulse rate values ​​in the pulse rate sequence to obtain a pulse rate difference sequence, and use the Bessel correction formula to calculate the standard deviation of the pulse rate difference sequence as the real-time pulse rate variability; A504. After performing wavelet transform on the pulse force signal of each pulse cycle data segment, the peak value of the pulse force signal is extracted to obtain a pulse force peak sequence, and the standard deviation of the pulse force peak sequence is calculated as the real-time pulse force intensity change; A505. Integrate the pulse position signal of each pulse cycle data segment and divide it by the pulse cycle duration to obtain the average value of the pulse position signal, form a pulse position average value sequence, and calculate the standard deviation of the pulse position average value sequence as the real-time pulse position deviation degree; A506. Detect the peak position of the pulse force signal of each pulse cycle data segment, determine the real-time pulse cycle length, obtain the real-time pulse cycle length sequence, and calculate the standard deviation of the real-time pulse cycle length sequence as the real-time arrhythmia degree.

[0061] Wherein, among the step A501, can adopt the autocorrelation analysis method, user's historical pulse data are carried out autocorrelation analysis, obtain autocorrelation coefficient sequence, determine the time delay corresponding to first local maximum point of autocorrelation coefficient sequence, as user's historical pulse cycle duration; Time window length refers to the time interval length that is used to segment pulse data, and the pulse cycle data segment refers to the pulse data fragment that obtains according to the time window length segmentation. Time window length can adopt fixed value, realize according to methods such as historical pulse cycle duration self-adaptation. The pulse data segmentation can adopt methods such as isometric segmentation, sliding window segmentation to realize.

[0062] In step A502, the pulse rate sequence is a sequence consisting of the average pulse rate values ​​of each pulse cycle data segment. The average pulse rate can be calculated by using methods such as arithmetic mean and weighted mean.

[0063] In step A503, real-time pulse rate variability refers to the degree of pulse rate fluctuation over a short period of time, reflecting the activity of the cardiac autonomic nervous system. The pulse rate difference sequence can be obtained by directly subtracting adjacent pulse rate values ​​or calculating the percentage change between adjacent pulse rate values. The Bessel correction formula can use different standard deviation calculation formulas, such as the unbiased sample standard deviation formula.

[0064] In step A504, real-time pulse strength changes refer to the degree of change in pulse force over a short period of time, reflecting myocardial contractility and vascular elasticity. Wavelet transforms can use different wavelet basis functions and wavelet decomposition levels, such as Daubechies wavelets and Symlets wavelets. The peak value of the pulse force signal can be extracted using methods such as local maximum detection and threshold detection.

[0065] In step A505, the real-time pulse position deviation refers to the degree of pulse position deviation over a short period of time, reflecting the fullness and tension of the blood vessels. The pulse position signal can be integrated using numerical integration methods such as trapezoidal integration or Simpson integration. The pulse cycle duration can be the real-time pulse cycle duration determined in step A506 or the historical pulse cycle duration obtained in step A501.

[0066] In step A506, the real-time arrhythmia level refers to the degree of irregularity of the pulse rhythm over a short period of time, reflecting the stability of the heart rhythm. The peak position of the pulse force signal can be determined using a peak detection algorithm, such as a wave crest detection algorithm. The real-time pulse cycle duration sequence can be obtained using methods such as the time interval between adjacent peak positions or the heartbeat interval.

[0067] Specifically, in order to effectively extract real-time pulse features that can be used for health risk assessment, the present application solution first performs step A501 in step A5 to determine a suitable time window length and segment the pulse data after physiological noise suppression into multiple pulse cycle data segments. This can decompose the continuous pulse data into smaller analysis units, so that subsequent feature extraction can be performed for each pulse cycle, thereby improving the accuracy and pertinence of the analysis. Determining the time window length is the key to data segmentation. If the time window length is too long, it may result in multiple pulse cycles being included in a time window, making it difficult to accurately analyze the features of each pulse cycle; if the time window length is too short, it may not be able to fully include the data of a pulse cycle, which will also affect the accuracy of the analysis. Therefore, in step A501, the historical pulse cycle duration is obtained through the user's historical pulse data, and the time window length is determined based on this. This can ensure that the time window length is compatible with the user's pulse cycle, thereby improving the rationality of data segmentation. After completing pulse data segmentation, steps A502 through A506 design specific extraction methods for four key real-time pulse characteristics: real-time pulse rate variability, real-time pulse strength variation, real-time pulse position deviation, and real-time pulse arrhythmia. To extract real-time pulse rate variability, step A502 calculates the average pulse rate for each pulse cycle data segment to obtain a pulse rate sequence. Step A503 then calculates the standard deviation of the pulse rate difference sequence to quantify the degree of pulse rate fluctuation, accurately reflecting the pulse rate variability. The application of the Bessel correction formula helps improve the accuracy of standard deviation estimation. To extract real-time pulse strength variation, step A504 uses wavelet transform to extract the peak value of the pulse signal, obtaining a pulse peak sequence, and calculating the standard deviation of this sequence. Wavelet transform effectively extracts local features of the pulse signal, and the peak value represents the strength of the pulse signal. Quantifying changes in pulse strength using the standard deviation of the peak sequence allows for sensitive capture of pulse strength fluctuations. To extract the real-time degree of pulse position deviation, step A505 integrates the pulse position signal and divides it by the pulse cycle duration to obtain a pulse position average value sequence, and then calculates the standard deviation of this sequence. Integrating the pulse position signal and dividing it by the pulse cycle duration to obtain the average value effectively reduces the interference of instantaneous fluctuations in the pulse position signal and more stably reflects the average level of the pulse position within a cycle. Quantifying the degree of pulse position deviation using the standard deviation of the average value sequence can effectively reflect the overall pulse position deviation. To extract the real-time degree of pulse arrhythmia, step A506 determines the real-time pulse cycle duration based on the peak position of the pulse force signal, obtains a real-time pulse cycle duration sequence, and calculates its standard deviation. Determining the pulse cycle duration directly based on the peak position of the pulse force signal can accurately reflect the actual pulse interval. Quantifying the degree of pulse arrhythmia using the standard deviation of the cycle duration sequence can effectively reflect the rhythmic changes of the pulse rhythm.Steps A501 to A506 constitute a complete real-time pulse feature extraction scheme. Targeting the four key pulse characteristics of pulse rate variability, pulse strength change, pulse position deviation degree, and arrhythmia degree, refined extraction methods are designed to provide comprehensive and accurate pulse feature parameters for subsequent health risk assessment.

[0068] Through the above technical solution, the present application can effectively extract real-time pulse characteristics that reflect the pulse status of the human body, and provide more comprehensive and accurate pulse characteristic parameters for subsequent health risk assessment. It solves the problem that the existing technology cannot effectively extract the real-time pulse characteristics of sub-healthy people, resulting in inaccurate pulse monitoring and analysis results in sub-healthy states, thereby meeting the health management needs of sub-healthy people.

[0069] Preferably, step A501 may include: Perform autocorrelation analysis on the user's historical pulse data to obtain an autocorrelation coefficient sequence, and determine the time delay corresponding to the first local maximum point of the autocorrelation coefficient sequence as the user's historical pulse cycle duration; The historical pulse cycle duration is scaled according to the predicted proportional coefficient to obtain the time window length; According to the length of the time window, the pulse data after physiological noise suppression is divided into multiple pulse cycle data segments using the sliding window method, and the step length of the sliding window is 1 / 2 of the time window length.

[0070] Among them, the user's historical pulse data is subjected to autocorrelation analysis in order to identify periodic components in the signal. Autocorrelation analysis can effectively reveal the repetitive pattern of the signal by calculating the correlation between the signal and itself under different time delays. The time delay corresponding to the first local maximum point of the autocorrelation coefficient sequence is determined as the estimated value of the historical pulse cycle duration. The introduction of the prediction scale coefficient is to be able to adjust the time window length as needed in practical applications. By scaling the historical pulse cycle duration, a more suitable time window length can be obtained to adapt to the needs of different users' physiological rhythms or monitoring scenarios. The sliding window method is adopted to segment the pulse data after physiological noise suppression. The window size is determined by the time window length, and the step size is set to half of the window length. This setting ensures the continuity of data segmentation and allows overlap between adjacent data segments to avoid omission of pulse cycle information.

[0071] Specifically, to more accurately segment the pulse cycle data, an autocorrelation analysis is first performed to calculate a sequence of autocorrelation coefficients. Within the sequence of autocorrelation coefficients, the first local maximum point is searched for. The time delay corresponding to this maximum point is the estimated historical pulse cycle duration. Subsequently, the historical pulse cycle duration is multiplied by a prediction scale factor to obtain the final time window length. The prediction scale factor can be pre-set based on empirical data or experimental results. For example, if the prediction scale factor is set to 1.1, the time window length is amplified by 10% based on the historical pulse cycle duration. After obtaining the time window length, a sliding window method is used to segment the pulse data after physiological noise suppression. The sliding window size is set to the previously calculated time window length, and the sliding step size is set to half the time window length. For example, if the time window length is 0.8 seconds, the sliding step size is 0.4 seconds. Using the sliding window, the pulse data after physiological noise suppression is segmented into a series of continuous and overlapping pulse cycle data segments.

[0072] In some embodiments, step A6 includes: A601. Obtain a health risk assessment matrix; the rows of the health risk assessment matrix represent the health risk type, the columns represent the pulse characteristics, and the matrix elements represent the weights of different health risk types on different pulse characteristics; A602. Based on the extracted real-time pulse characteristics and the user's historical pulse characteristics, the user's comprehensive pulse feature vector is calculated. The elements of the comprehensive pulse feature vector include comprehensive pulse rate variability, comprehensive pulse strength changes, comprehensive pulse position offset degree and comprehensive arrhythmia degree; A603. Multiplying the comprehensive pulse characteristic vector with the health risk assessment matrix to obtain a health risk assessment vector; the elements of the health risk assessment vector represent risk scores for different health risk types; A604. Normalize the health risk assessment vector to obtain a normalized health risk assessment vector, determine the health risk type corresponding to the element with the largest value in the normalized health risk assessment vector as the user's health risk type, and determine the value of the element as the user's health risk level.

[0073] Among them, in step A601, the health risk assessment matrix can be pre-set to represent the degree of association between different health risk types and different pulse characteristics. Each row of the matrix represents a specific health risk type, such as cardiovascular risk, liver function risk, etc.; each column represents a pulse characteristic, such as pulse rate variability, pulse strength change, etc. Each element in the matrix is ​​a weight value, and the numerical value reflects the importance of the corresponding health risk type in the specific pulse characteristic. The weight value can be determined based on medical knowledge, clinical experience or big data statistical analysis results.

[0074] Wherein, among the step A602, the computational process of comprehensive pulse characteristic vector, at first, adopts clustering algorithm to analyze user's historical pulse characteristics, historical pulse characteristics are divided into multiple pulse categories, and calculate the cluster center of each pulse category (each cluster center is the four-dimensional data point that comprises pulse rate variability, pulse strength variation, pulse position deviation degree and arrhythmia degree), obtain user's historical pulse characteristics cluster center set thus.Then, calculate the Euclidean distance of each cluster center in the real-time pulse characteristics that extracts and the historical pulse characteristics cluster center set, select the cluster center with minimum Euclidean distance as the user's representative historical pulse characteristics.Finally, the real-time pulse characteristics that extracts and representative historical pulse characteristics are carried out weighted fusion, obtain user's comprehensive pulse characteristics vector.Weighted fusion can adopt preset weight ratio, for example real-time pulse characteristics weight is 0.7, and representative historical pulse characteristics weight is 0.3.

[0075] In step A603, the comprehensive pulse characteristic vector is multiplied by the health risk assessment matrix. This matrix multiplication maps the comprehensive pulse characteristic vector to a health risk assessment vector. Each element of the health risk assessment vector represents a risk score for a corresponding health risk type, and the numerical value reflects the likelihood that the user has that health risk.

[0076] In step A604, the health risk assessment vector is normalized to bring the risk scores for different health risk types into the same range, for example, between 0 and 1, for easier comparison and judgment. After normalization, the health risk type corresponding to the element with the largest value in the health risk assessment vector is determined as the user's health risk type, and the value of that element is determined as the user's health risk level.

[0077] Specifically, a health risk assessment matrix can be a pre-built matrix with four rows and four columns. The rows represent four health risk types: cardiovascular risk, liver function risk, kidney function risk, and spleen and stomach function risk; the columns represent four pulse characteristics: comprehensive pulse rate variability, comprehensive pulse strength variation, comprehensive pulse position deviation, and comprehensive pulse arrhythmia. The weights in the matrix are set based on expert experience. For example, cardiovascular risk is given a higher weight for comprehensive pulse rate variability, while the weight for comprehensive pulse position deviation is lower.

[0078] When calculating the comprehensive pulse characteristic vector, assume that the extracted real-time pulse characteristic vector is [0.8, 0.5, 0.2, 0.6], and the user's historical pulse characteristic cluster center set contains three cluster centers. Calculate the Euclidean distance between the real-time pulse characteristic vector and these three cluster centers respectively, and select the cluster center with the smallest Euclidean distance as the representative historical pulse characteristic. Assume that the representative historical pulse characteristic vector is [0.6, 0.4, 0.3, 0.5]. Then, the real-time pulse characteristic vector and the representative historical pulse characteristic vector are weighted fused with weights of 0.7 and 0.3 respectively, and the comprehensive pulse characteristic vector is [0.74, 0.47, 0.27, 0.57].

[0079] Multiply the comprehensive pulse characteristic vector by the health risk assessment matrix to obtain the health risk assessment vector. After normalizing the health risk assessment vector, assume that the normalized health risk assessment vector is [0.2, 0.8, 0.1, 0.5]. Therefore, the user's health risk type is determined to be liver function risk, and the health risk level is 0.8.

[0080] In some specific embodiments, the weight values ​​of the health risk assessment matrix can be adjusted according to the individual characteristics of the user. For example, for elderly users, the weights of cardiovascular risk and renal function risk can be appropriately increased; for young people, the weights of liver function risk and spleen and stomach function risk can be appropriately increased. The number of historical pulse feature clustering centers can be adjusted according to the amount of historical data of the user. The larger the amount of historical data, the more clustering centers can be appropriately increased to more finely portray the historical pulse characteristics of the user. The weight ratio of weighted fusion can also be adjusted according to the actual application scenario. For example, in a scenario where rapid response is required, the weight of real-time pulse features can be appropriately increased.

[0081] By structurally integrating pulse characteristics and health risk information through the health risk assessment matrix, and effectively fusing real-time and historical pulse data through the comprehensive pulse feature vector, accurate and personalized health risk assessment is achieved.

[0082] Preferably, step A602 may include: A clustering algorithm is used to divide the user's historical pulse characteristics into multiple pulse categories, and the cluster center of each pulse category is calculated to obtain a cluster center set of the user's historical pulse characteristics; Calculate the Euclidean distance between the extracted real-time pulse feature and each cluster center in the set of historical pulse feature cluster centers, and select the cluster center with the smallest Euclidean distance as the user's representative historical pulse feature; The extracted real-time pulse features are weightedly fused with representative historical pulse features to obtain the user's comprehensive pulse feature vector.

[0083] Wherein, clustering algorithm is used to divide the user's historical pulse condition characteristics into multiple pulse condition categories, for example, can adopt K-means clustering algorithm. In K-means clustering algorithm, first need to determine the cluster number K value, K value can be selected according to actual application scenario and data characteristics, or by methods such as elbow rule, silhouette coefficient auxiliary determination. Then, randomly select K historical pulse condition feature vectors as initial cluster centers. Next, calculate the distance between each historical pulse condition feature vector and each cluster center, and divide each historical pulse condition feature vector into the category where the nearest cluster center is located. Afterwards, recalculate the cluster center of each category, and the cluster center is updated to the mean vector of all historical pulse condition feature vectors in the category. Repeat the above-mentioned iterative process until the cluster center no longer changes significantly or reaches the preset number of iterations. Thus, the user's historical pulse condition characteristics are divided into multiple pulse condition categories, and the cluster center of each pulse condition category is calculated, thereby constructing the user's historical pulse condition feature cluster center set.

[0084] Further, after the Euclidean distance of each cluster center in the real-time pulse feature and historical pulse feature cluster center set of calculation extraction, select the cluster center with minimum Euclidean distance as the representative history pulse feature of the user, and the representative history pulse feature is closer to the pulse state of the current user.

[0085] As a kind of preferred embodiment, when the real-time pulse feature and representative history pulse feature of extracting are carried out weighted fusion, can adopt linear weighted fusion method.For example, the weight of setting real-time pulse feature is w1, the weight of representative history pulse feature is w2, and w1+w2=1.The computing formula of comprehensive pulse characteristic vector is: Comprehensive pulse characteristic vector = w1*real-time pulse characteristic+w2*representative historical pulse characteristic The weights w1 and w2 can be adjusted based on actual application requirements. For example, when the reliability of real-time pulse features is high, the weight of w1 can be appropriately increased; when the reference value of historical pulse features is high, the weight of w2 can be appropriately increased. Through weighted fusion, real-time pulse features and representative historical pulse features are effectively combined to obtain the user's comprehensive pulse feature vector.

[0086] Specifically, by adopting a clustering algorithm to analyze the user's historical pulse characteristics, several representative pulse category centers are extracted therefrom to reflect the historical distribution characteristics of the user's pulse. After obtaining the user's current real-time pulse characteristics, the distance between the real-time pulse characteristics and each historical pulse category center is calculated to find the historical pulse category center closest to the current pulse state and use it as the representative historical pulse characteristics. The representative historical pulse characteristics can be regarded as a summary and refinement of the user's historical pulse experience, which can reflect the individualized baseline and long-term variation trend of the user's pulse. Then, the real-time pulse characteristics are fused with the representative historical pulse characteristics, and the user's current pulse state and historical pulse characteristics are comprehensively considered to generate a comprehensive pulse feature vector. The comprehensive pulse feature vector not only contains the user's current pulse information, but also incorporates the individualized historical pulse experience, making the expression of the pulse characteristics more comprehensive and accurate. Thus, the individualization and accuracy of the pulse feature vector can be improved, providing more reliable data support for subsequent health risk assessment, thereby improving the reliability of health risk assessment.

[0087] refer to Figure 2 , the present application provides a traditional Chinese medicine pulse diagnosis data analysis system, the system includes a monitoring device 1 and a pulse diagnosis and analysis device 2, the monitoring device 1 and the pulse diagnosis and analysis device 2 are communicatively connected; The monitoring device 1 includes a pulse diagnosis sensor 101 and a physiological state sensor 102. The pulse diagnosis sensor 102 is used to collect the user's pulse data, and the physiological state sensor 102 is used to synchronously collect the user's physiological state parameters; the pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and the physiological state parameters include activity level, heart rate and respiratory rate; The pulse diagnosis and analysis device 2 includes: Receiving module 201, for receiving real-time physiological state parameters and real-time pulse data collected by monitoring device 1 in real time; The impact factor evaluation module 202 is used to obtain the user's historical monitoring data, analyze the correlation between the changes in each physiological state parameter and the changes in each pulse data, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; the historical monitoring data includes the historical pulse data and historical physiological state parameters collected simultaneously (the specific process is referred to step A1 above); The impact degree evaluation module 203 is used to calculate the theoretical impact value of each real-time physiological state parameter on each real-time pulse data according to the physiological noise impact factor (for the specific process, refer to step A3 above); Noise suppression module 204, for subtracting the corresponding theoretical impact value from each real-time pulse data to obtain pulse data after physiological noise suppression (for the specific process, refer to step A4 above); Feature analysis module 205 is used to extract real-time pulse characteristics based on the pulse data after physiological noise suppression; the real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength change, real-time pulse position deviation degree and real-time pulse arrhythmia degree (for details, refer to step A5 above); The risk assessment module 206 is used to assess the health risk type and corresponding risk level based on the extracted real-time pulse characteristics and the user's historical pulse characteristics; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree and historical pulse arrhythmia degree (for the specific process, refer to step A6 above).

[0088] The monitoring device 1 and the pulse analysis device 2 can be connected by wireless communication technology such as Bluetooth or Wi-Fi for data transmission, or by wired connection mode such as USB interface for data transmission to ensure the stability and real-time performance of data transmission. The pulse diagnosis sensor 101 can be a piezoelectric sensor, a photoelectric sensor or a micro-electromechanical system sensor, etc., for detecting the pulse waveform of the user's radial artery, thereby obtaining pulse data. The physiological state sensor 102 can include an acceleration sensor for detecting activity, a heart rate sensor such as a photoelectric volume pulse wave sensor or an electrocardiogram sensor for detecting heart rate, and a respiratory sensor such as a thermistor sensor or a pressure sensor for detecting respiratory rate, so as to fully reflect the user's physiological state.

[0089] In some specific embodiments, monitoring device 1 is designed as a wearable device, such as a smart bracelet or smart watch, which integrates a pulse diagnosis sensor 101 and a physiological state sensor 102, so that users can monitor pulse and physiological state at home or in daily life. The pulse diagnosis and analysis device 2 can be implemented as a smart phone App or a cloud server, which receives and processes the data uploaded by monitoring device 1, and performs noise suppression, feature extraction and risk assessment. Historical monitoring data and historical pulse characteristics are stored in a cloud server or a local database for impact factor evaluation and risk assessment.

[0090] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for analyzing pulse diagnosis data in traditional Chinese medicine, characterized in that: The method comprises the following steps: A1. Obtain the user's historical monitoring data to analyze the correlation between changes in each physiological state parameter and each pulse data change, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; historical monitoring data includes historical pulse data and historical physiological state parameters collected simultaneously; pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and physiological state parameters include activity level, heart rate and respiratory rate; A2. Synchronously collect the user's real-time physiological status parameters and real-time pulse data; A3. According to the physiological noise factor, calculate the theoretical impact of each real-time physiological state parameter on each real-time pulse data; A4. Subtract the corresponding theoretical impact value from each real-time pulse data to obtain the pulse data after physiological noise suppression; A5. Based on the pulse data after physiological noise suppression, real-time pulse characteristics are extracted; the real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength changes, real-time pulse position offset degree and real-time irregular pulse degree; A6. Based on the extracted real-time pulse characteristics and combined with the user's historical pulse characteristics, the health risk type and corresponding risk level are assessed; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree, and historical pulse arrhythmia degree.

2. A TCM pulse diagnosis data analysis method according to claim 1, characterized in that: Step A1 includes: A101. The historical pulse data and historical physiological state parameters are preprocessed to obtain the preprocessed historical pulse data and preprocessed historical physiological state parameters; the preprocessing includes removing outliers, filling missing values ​​and data smoothing; A102. Calculate the mutual information value between each physiological state parameter and each pulse condition parameter using the mutual information method to obtain a preliminary physiological noise impact factor; A103. Use t-test to evaluate whether each physiological state parameter has a significant effect on each pulse condition parameter, and classify the combination of each physiological state parameter and each pulse condition parameter into a significant effect combination and a non-significant effect combination. A104. Set the preliminary physiological noise impact factors corresponding to the non-significant impact combinations to zero, and keep the preliminary physiological noise impact factors corresponding to the significant impact combinations to obtain the final physiological noise impact factors.

3. A TCM pulse diagnosis data analysis method according to claim 2, characterized in that: Step A102 includes: Discretization processing is performed on various physiological state parameters and various pulse condition parameters based on the equal frequency binning method to obtain discrete physiological state parameters and discrete pulse condition parameters; According to the discretized physiological state parameters and the discretized pulse parameters, the mutual information calculation formula is used to calculate the mutual information value between each physiological state parameter and each pulse parameter, and the preliminary physiological noise impact factor is obtained.

4. A method for analyzing pulse diagnosis data in traditional Chinese medicine according to claim 1, characterized in that: Step A3 includes: A301. Based on the user's historical physiological state parameters and historical pulse data, calculate the user's individual physiological state parameter baseline value; A302. Perform difference calculation on the real-time physiological state parameter and the individual physiological state parameter baseline value to obtain the standardized real-time physiological state parameter; A303. For each item of real-time pulse data, calculate the product of each standardized real-time physiological state parameter and the corresponding physiological noise impact factor to obtain the theoretical impact value of each real-time physiological state parameter on this real-time pulse data.

5. A method for analyzing pulse diagnosis data in traditional Chinese medicine according to claim 4, characterized in that: Step A302 includes: Using the user's historical physiological state parameters, the partial correlation analysis method is used to analyze and quantify the mutual influence relationship between various physiological state parameters, and the partial correlation coefficient matrix between physiological parameters is obtained; constructing a physiological parameter correction matrix according to the partial correlation coefficient matrix between the physiological parameters; Subtract the individual physiological state parameter baseline value from the real-time physiological state parameter to obtain the preliminarily standardized real-time physiological state parameter; The preliminarily standardized real-time physiological state parameters are corrected according to the physiological parameter correction matrix to obtain final standardized real-time physiological state parameters.

6. A method for analyzing pulse diagnosis data in traditional Chinese medicine according to claim 1, characterized in that: Step A5 includes: A501 obtains the user's historical pulse cycle duration based on historical pulse data to determine the time window length, and divides the pulse data after physiological noise suppression into multiple pulse cycle data segments based on the time window length; A502. Calculate the average pulse rate of the pulse rate signal of each pulse cycle data segment to obtain a pulse rate sequence; A503. Calculate the difference between adjacent pulse rate values ​​in the pulse rate sequence to obtain a pulse rate difference sequence, and use the Bessel correction formula to calculate the standard deviation of the pulse rate difference sequence as the real-time pulse rate variability; A504. After performing wavelet transform on the pulse force signal of each pulse cycle data segment, the peak value of the pulse force signal is extracted to obtain a pulse force peak sequence, and the standard deviation of the pulse force peak sequence is calculated as the real-time pulse force intensity change; A505. Integrate the pulse position signal of each pulse cycle data segment and divide it by the pulse cycle duration to obtain the average value of the pulse position signal, form a pulse position average value sequence, and calculate the standard deviation of the pulse position average value sequence as the real-time pulse position deviation degree; A506. Detect the peak position of the pulse force signal of each pulse cycle data segment, determine the real-time pulse cycle length, obtain the real-time pulse cycle length sequence, and calculate the standard deviation of the real-time pulse cycle length sequence as the real-time arrhythmia degree.

7. A method for analyzing TCM pulse diagnosis data according to claim 6, characterized in that: Step A501 includes: Perform autocorrelation analysis on the user's historical pulse data to obtain an autocorrelation coefficient sequence, and determine the time delay corresponding to the first local maximum point of the autocorrelation coefficient sequence as the user's historical pulse cycle duration; Scaling the historical pulse cycle duration according to the predicted proportional coefficient to obtain the time window length; According to the length of the time window, the pulse data after physiological noise suppression is divided into multiple pulse cycle data segments using the sliding window method, and the step length of the sliding window is 1 / 2 of the time window length.

8. A method for analyzing TCM pulse diagnosis data according to claim 1, characterized in that: Step A6 includes: A601. Obtain a health risk assessment matrix; the rows of the health risk assessment matrix represent the health risk type, the columns represent the pulse characteristics, and the matrix elements represent the weights of different health risk types on different pulse characteristics; A602. Based on the extracted real-time pulse characteristics and the user's historical pulse characteristics, the user's comprehensive pulse feature vector is calculated. The elements of the comprehensive pulse feature vector include comprehensive pulse rate variability, comprehensive pulse strength changes, comprehensive pulse position offset degree and comprehensive arrhythmia degree; A603. Multiplying the comprehensive pulse characteristic vector with the health risk assessment matrix to obtain a health risk assessment vector; the elements of the health risk assessment vector represent risk scores for different health risk types; A604. Normalize the health risk assessment vector to obtain a normalized health risk assessment vector, determine the health risk type corresponding to the element with the largest value in the normalized health risk assessment vector as the user's health risk type, and determine the value of the element as the user's health risk level.

9. A method for analyzing TCM pulse diagnosis data according to claim 8, characterized in that: Step A602 includes: A clustering algorithm is used to divide the user's historical pulse characteristics into multiple pulse categories, and the cluster center of each pulse category is calculated to obtain a cluster center set of the user's historical pulse characteristics; Calculate the Euclidean distance between the extracted real-time pulse feature and each cluster center in the set of historical pulse feature cluster centers, and select the cluster center with the smallest Euclidean distance as the user's representative historical pulse feature; The extracted real-time pulse features are weightedly fused with representative historical pulse features to obtain the user's comprehensive pulse feature vector.

10. A traditional Chinese medicine pulse diagnosis data analysis system, characterized in that: The system includes a monitoring device and a pulse diagnosis and analysis device, and the monitoring device and the pulse diagnosis and analysis device are communicatively connected; The monitoring device includes a pulse diagnosis sensor and a physiological state sensor, wherein the pulse diagnosis sensor is used to collect the user's pulse data, and the physiological state sensor is used to synchronously collect the user's physiological state parameters; the pulse data includes pulse rate, pulse strength, pulse position and pulse rhythm, and the physiological state parameters include activity level, heart rate and respiratory rate; The pulse diagnosis and analysis device comprises: A receiving module is used for receiving the real-time physiological state parameters and real-time pulse data collected by the monitoring device in real time; The impact factor evaluation module is used to obtain the user's historical monitoring data, analyze the correlation between the changes in each physiological state parameter and the changes in each pulse data, and obtain the physiological noise impact factor of each physiological state parameter on each pulse data; the historical monitoring data includes the historical pulse data and historical physiological state parameters collected synchronously; An impact degree evaluation module is used to calculate the theoretical impact value of each real-time physiological state parameter on each real-time pulse data according to the physiological noise impact factor; The noise suppression module is used to subtract the corresponding theoretical impact value from each real-time pulse data to obtain the pulse data after physiological noise suppression; A feature analysis module is used to extract real-time pulse characteristics based on the pulse data after physiological noise suppression; the real-time pulse characteristics include real-time pulse rate variability, real-time pulse strength change, real-time pulse position deviation degree and real-time pulse arrhythmia degree; The risk assessment module is used to assess the health risk type and corresponding risk level based on the extracted real-time pulse characteristics combined with the user's historical pulse characteristics; historical pulse characteristics include historical pulse rate variability, historical pulse strength changes, historical pulse position deviation degree and historical pulse arrhythmia degree.

Images