A multi-dimensional student physical health monitoring method and system

Abstract

The present invention discloses a multi-dimensional student physical health monitoring method and system, which relates to the field of health monitoring technology, including collecting data to construct a signal subset, performing time-frequency decomposition to construct a time-frequency graph matrix, performing autoregressive modeling to extract an observation matrix and a design matrix, and constructing a frequency feature vector; calculating a contribution factor and determining a dynamic health score based on the frequency characteristics of different frequency bands, calculating the posterior probability of a health state based on the conditional probability of the health score, and calculating a frequency band offset to construct an abnormal frequency band set. The method of the present invention constructs a health assessment model by utilizing an observation matrix and a design matrix, so that each row of signal values ​​is associated with its lagged historical value, which can capture the time dependence of the signal and reflect the dynamic characteristics of the health signal. The mean, variance, skewness and kurtosis are calculated based on the PSD, and the high-dimensional spectrum information is condensed into a feature vector of fixed dimension, thereby reducing data complexity and facilitating subsequent classification and health status analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring, and in particular to a multi-dimensional student physical health monitoring method and system. Background Art

[0002] In recent years, with the increasing awareness of health and the continuous development of medical technology, physical health monitoring has become an important part of modern health management, especially among students. Traditional health monitoring methods usually rely on single indicators, such as heart rate, blood pressure, and weight. Although these indicators can reflect some health conditions, they are difficult to fully reveal the full picture of complex health dynamics. With the rapid development of sensor networks and big data technology, the application of multimodal sensors has gradually become popular, making it possible to collect multi-dimensional physiological data in real time.

[0003] However, traditional data analysis methods mostly use linear or low-dimensional models, which cannot fully explore the potential nonlinear correlations in multimodal data. This is particularly evident in multi-band feature extraction and anomaly detection. Secondly, existing anomaly detection technologies are mostly based on preset thresholds, and have limited ability to distinguish the types of abnormal patterns. It is difficult to effectively identify specific anomalies and general anomalies in complex signals. In addition, the assessment of health status often relies on a static single health scoring model, ignoring the dynamic changes in health status, resulting in insufficient real-time and accuracy of the assessment results. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-dimensional student physical health monitoring method and system to solve the above problems.

[0005] The present invention is achieved through the following technical solutions:

[0006] A multi-dimensional student physical health monitoring method, comprising:

[0007] A multimodal sensor network was used to collect students' physiological data, including heart rate, blood oxygen, blood pressure, and motion trajectory data. Signal subsets were constructed from the collected data, and time-frequency decomposition was performed to construct a time-frequency graph matrix. Autoregressive modeling was performed to extract the observation matrix and design matrix, and frequency eigenvectors were constructed.

[0008] Calculate the contribution factor and determine the dynamic health score based on the frequency characteristics of different frequency bands. Calculate the posterior probability of the health state based on the conditional probability of the health score. Calculate the frequency band offset to construct the abnormal frequency band set.

[0009] Use support vector machines to classify abnormal frequency bands, draw heat maps to display frequency band anomalies, and generate abnormality reports for encrypted transmission.

[0010] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, the signal subset constructed by collecting data includes:

[0011] Use a multimodal sensor network to collect students' physiological data, including heart rate, blood oxygen, blood pressure, and movement trajectory data;

[0012] Based on different physiological data items, a corresponding data matrix is ​​established according to the collection time, and the collected data of each sensor is standardized;

[0013] The sensor data collected continuously in time is subjected to a sliding window operation, the data is segmented according to the sampling interval, and a signal subset consisting of sensor data with different sampling intervals is constructed.

[0014] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, the time-frequency decomposition is performed to construct a time-frequency graph matrix, the autoregressive modeling is performed to extract the observation matrix and the design matrix, and the frequency feature vector is constructed, including:

[0015] Perform Fourier synchronous compression transform on the subset of sensor signals to perform time-frequency decomposition and output the time-frequency map matrix of the corresponding sensor signals , represents the energy distribution of the signal at time t and frequency f;

[0016] Autoregressive spectrum AR modeling based on time-frequency diagram matrix;

[0017] The observation matrix Y and design matrix A are extracted from the signal data collected by the sensor and expressed as:

[0018] ;

[0019] ;

[0020] Where A represents the setting matrix composed of the signal history values, each row represents the previous p historical values ​​of the current signal value, each column corresponds to the data value of a lag time step, p represents the AR order, and represents the current signal value Associated with the signal value of how many historical moments, Y represents the observation matrix, containing the current signal value corresponding to each row in the design matrix A , represents the signal value at time t=p, Represents the signal value at time point t=Np, represents the target signal value after p steps of lag, and N represents the total number of sampling points in the entire signal time series;

[0021] The AR coefficient vector is calculated by the least squares method based on the observation matrix Y and the design matrix A, which is expressed as:

[0022] ;

[0023] Where a represents the AR coefficient vector, represents the transpose calculation of the design matrix;

[0024] The model noise is determined based on the AR coefficient vector and is expressed as:

[0025] ;

[0026] in represents the signal value at time t, represents the kth AR coefficient vector, represents the noise at time t;

[0027] Then calculate the power spectral density, which is expressed as:

[0028] ;

[0029] in Represents the distribution of signal power at frequency f in the frequency domain as the spectrum of the corresponding frequency, represents the variance of the noise, j represents the imaginary unit;

[0030] calculate The frequency feature vector is constructed from the mean, variance, skewness and kurtosis of .

[0031] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, wherein: the contribution factor is calculated based on the frequency characteristics of different frequency bands and the dynamic health score is determined, the posterior probability of the health state is calculated based on the conditional probability of the health score, and the frequency band offset is calculated to construct an abnormal frequency band set, including:

[0032] The contribution factor is calculated based on the statistical eigenvalues ​​of different frequency bands based on the frequency eigenvector, which is expressed as:

[0033] ;

[0034] in Indicates the contribution of the ith frequency band to the overall abnormal health status, represents the jth statistical eigenvalue of the i-th frequency band, represents the jth statistical eigenvalue of the i-th frequency band in the healthy state, which is determined by calculating the sum of the mean and standard deviation of the healthy data set, n represents the total number of frequency bands, and m represents the number of statistical features of the frequency eigenvector;

[0035] Combining the contribution factors and spectrum characteristics of each frequency band, the dynamic health score is calculated and expressed as:

[0036] ;

[0037] Where DHS represents the overall health status value;

[0038] Construct a health status classification model based on the dynamic Bayesian formula, calibrate the health status of students according to the historical health status value data, including normal, sub-health and abnormal, and define the kth health status as The occurrence frequency of each state in the historical health status value data is statistically analyzed as the prior probability of the corresponding student health state;

[0039] The health score DHS data under each student's health status are distributed and fitted, and the mean and standard deviation are calculated. The conditional probability of the health score DHS under the premise that the student is currently in a healthy state is calculated, which is expressed as:

[0040] ;

[0041] in represents the probability that the health score belongs to the kth health state, represents the standard deviation of health scores under the kth health state, represents the mean health score under the kth health state;

[0042] Calculate the posterior probability of the student's health status under different health conditions, expressed as:

[0043] ;

[0044] in represents the posterior probability that the health score belongs to the kth health state, represents the prior probability of the kth health state, K represents the total number of health states, represents the kth health state, represents the probability that the health score belongs to the kth health state;

[0045] The health state with the maximum probability is selected based on the posterior probability to determine the predicted health state of the student with the corresponding health score;

[0046] Count different health states and corresponding spectra in historical data, and use the spectra as reference spectra for the corresponding health states;

[0047] The discrimination of health states was verified by the distribution distance between the mean and standard deviation of the reference spectrum;

[0048] Based on the predicted health status, the corresponding reference spectrum is selected to calculate the frequency band offset, which is expressed as:

[0049] ;

[0050] in represents the spectrum offset of the i-th frequency band, and represents the frequency range of the i-th frequency band and the i-1-th frequency band, represents the spectrum of frequency f in the i-th frequency band, represents the reference spectrum of frequency f in the i-th frequency band;

[0051] The deviation threshold is calculated based on the mean and double standard deviation of the offset data of historical samples. If the spectrum offset is greater than or equal to the deviation threshold, it means that the deviation is large, and the frequency band i is marked as abnormal. The judgment of the marked abnormality and the corresponding offset data and frequency feature vector are combined into an abnormal frequency band set.

[0052] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, the abnormal classification of abnormal frequency bands using a support vector machine includes:

[0053] Use the support vector machine (SVM) pre-trained with historically labeled anomaly type data to perform anomaly classification on each abnormal frequency band i in the abnormal frequency band set;

[0054] Specific anomaly types and general anomaly types are identified as classification results for each anomaly frequency band.

[0055] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, wherein: drawing a heat map to display frequency band abnormalities includes:

[0056] Plot the position and offset of the abnormal frequency band on the frequency axis, and construct a heat map to show the degree of frequency band abnormality. The horizontal axis represents time, the vertical axis represents frequency, and the color represents the corresponding frequency offset.

[0057] Based on the classification results of the abnormal frequency bands, different colors are used to distinguish and mark specific anomalies and general anomalies.

[0058] As a preferred solution of the multi-dimensional student physical health monitoring method of the present invention, the generating of abnormality reports for encrypted transmission includes:

[0059] Corresponding abnormality reports are generated for the heat map data through a spreadsheet tool, and sent to the student health supervision terminal using SSL / TLS protocol encrypted communication. At the same time, the generated abnormality report data is stored in the cloud, and the health scores, abnormal frequency band sets and abnormal classification results of different students are sorted out based on time series to generate health reports for the corresponding students through a spreadsheet tool.

[0060] The present invention provides a system for a multi-dimensional student physical health monitoring method, comprising:

[0061] The data acquisition module uses multimodal sensors to collect students' real-time physiological data and pre-processes the collected raw data;

[0062] The autoregressive modeling module performs synchronous compression transformation on the signal subset, generates a time-frequency map matrix, and builds an autoregressive AR model based on the time-frequency features;

[0063] The health scoring module calculates dynamic health scores based on frequency eigenvectors and frequency band contribution factors. It also calculates posterior probabilities based on the dynamic Bayesian formula and the conditional probability of the health scores to predict students' health status.

[0064] The anomaly detection module calculates the frequency band offset based on the reference spectrum corresponding to the healthy state, identifies abnormal frequency bands, constructs an abnormal frequency band set, uses a support vector machine (SVM) to classify the abnormal frequency bands, and generates a heat map showing the time, frequency, and offset of the abnormal frequency bands;

[0065] The report generation module generates abnormal reports and transmits them to the student health monitoring terminal through the SSL / TLS encryption protocol. All health scores, abnormal frequency band sets and classification results are stored in the cloud to support long-term health management.

[0066] The present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the multi-dimensional student physical health monitoring method as described in the present invention is implemented.

[0067] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the multi-dimensional student physical health monitoring method as described in the present invention is implemented.

[0068] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0069] 1. The present invention constructs a health assessment model by using an observation matrix and a design matrix, so that each row of signal values ​​is associated with its lagged historical value. This can capture the time dependence of the signal and reflect the dynamic characteristics of the health signal. Based on the PSD calculation of the mean, variance, skewness and kurtosis, the high-dimensional spectrum information is condensed into a fixed-dimensional feature vector, reducing data complexity and facilitating subsequent classification and health status analysis. By estimating the model noise, the deviation between the signal and the model prediction value is quantified, sudden abnormal signals are identified, and the sensitivity and robustness of the health monitoring system to signal anomalies are enhanced.

[0070] 2. This invention constructs a dynamic health score, combines the contribution factors of each frequency band and the spectrum characteristics, and dynamically reflects the overall fluctuation of health status. This makes the health score time-sensitive, can capture the changing trend of health status in real time, and provides quantitative indicators to facilitate medical personnel to quickly understand the individual's health level;

[0071] 3. The present invention uses a classification model based on the dynamic Bayesian formula to accurately predict individual health status. It combines dynamic health scores to calculate conditional probabilities, improves the classification model's ability to distinguish health status, and enhances its sensitivity to abnormal health status. Based on the reference spectrum corresponding to the predicted health status, the frequency band offset is calculated. Identifying abnormal frequency bands can quickly locate the source of abnormal signals and support refined health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0073] Figure 1 This is a flowchart of the multi-dimensional student physical health monitoring method;

[0074] Figure 2 This is a structural diagram of the multi-dimensional student physical health monitoring system. DETAILED DESCRIPTION

[0075] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0076] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a multi-dimensional student physical health monitoring method, comprising the following steps:

[0077] S1, collect data to construct signal subsets, perform time-frequency decomposition to construct a time-frequency map matrix, perform autoregressive modeling to extract the observation matrix and design matrix, calculate the power spectral density, and construct the frequency eigenvector;

[0078] Preferably, collecting data to construct a signal subset includes:

[0079] Use a multimodal sensor network to collect students' physiological data, including heart rate, blood oxygen, blood pressure, and movement trajectory data;

[0080] Based on different physiological data items, a corresponding data matrix is ​​established according to the collection time, and the collected data of each sensor is standardized;

[0081] The sensor data collected continuously in time is subjected to a sliding window operation, the data is segmented according to the sampling interval, and a signal subset consisting of sensor data with different sampling intervals is constructed.

[0082] Through a multimodal sensor network (heart rate, blood oxygen, blood pressure, and motion trajectory), students' physiological data can be collected from multiple angles, covering different aspects of health status, improving the comprehensiveness of health assessments, and avoiding misjudgments that may result from single-indicator monitoring. By standardizing data from different sensors, analysis biases caused by different dimensions and units are eliminated. Through sliding window operations, the timing characteristics of the signal are retained, making it suitable for the assessment of dynamic health status and improving the accuracy of subsequent health assessments and anomaly detection.

[0083] Furthermore, time-frequency decomposition is performed to construct a time-frequency graph matrix, autoregressive modeling is performed to extract the observation matrix and design matrix, and the frequency feature vector is constructed, including:

[0084] Perform Fourier synchronous compression transform on the subset of sensor signals to perform time-frequency decomposition and output the time-frequency map matrix of the corresponding sensor signals , represents the energy distribution of the signal at time t and frequency f;

[0085] FSST reduces the blurring effect of non-stationary signals by compressing the spectrum energy and has higher spectral resolution than traditional STFT.

[0086] Autoregressive spectrum AR modeling is performed based on the time-frequency matrix (the order p is determined by the Akaike Information Criterion AIC);

[0087] The observation matrix Y and design matrix A are extracted from the signal data collected by the sensor and expressed as:

[0088] ;

[0089] ;

[0090] Where A represents the setting matrix composed of the signal history values, each row represents the previous p historical values ​​of the current signal value, each column corresponds to the data value of a lag time step, p represents the AR order, and represents the current signal value Associated with the signal value of how many historical moments, Y represents the observation matrix, containing the current signal value corresponding to each row in the design matrix A , represents the signal value at time t=p, Represents the signal value at time point t=Np, represents the target signal value after p steps of lag, and N represents the total number of sampling points in the entire signal time series;

[0091] The AR coefficient vector is calculated by the least squares method based on the observation matrix Y and the design matrix A, which is expressed as:

[0092] ;

[0093] Where a represents the AR coefficient vector, represents the transpose calculation of the design matrix;

[0094] The model noise is determined based on the AR coefficient vector and is expressed as:

[0095] ;

[0096] in represents the signal value at time t, represents the kth AR coefficient vector, represents the noise at time t;

[0097] Then calculate the power spectral density, which is expressed as:

[0098] ;

[0099] in Represents the distribution of signal power at frequency f in the frequency domain as the spectrum of the corresponding frequency, represents the variance of the noise, j represents the imaginary unit, and satisfies ;

[0100] calculate The frequency feature vector is constructed from the mean, variance, skewness and kurtosis of .

[0101] In the monitoring of students' physical health, the signal power spectrum at frequency f Providing information on the distribution of signal energy in the frequency domain is an important representation of time-varying signals. It further condenses important features in the spectrum into a fixed-dimensional description through statistical quantities including mean, variance, skewness, and kurtosis, retaining the main characteristics of the signal while reducing data complexity. The mean represents the overall energy level of the frequency band, the variance reflects the degree of discreteness of the spectrum distribution and captures energy changes, the skewness describes the asymmetry of the spectrum distribution and reveals abnormal signals, and the kurtosis quantifies the sharpness of the spectrum distribution and can be used to detect local anomalies.

[0102] Compared with the traditional short-time Fourier transform (STFT), FSST can reduce the blurring effect of non-stationary signals by compressing the spectral energy through Fourier synchronous compression transform, improve the spectral resolution of the signal, and more accurately capture the distribution of dynamic health signals in time and frequency. It also enhances the ability to analyze complex or sudden signals, such as heart rate fluctuations and drastic changes in blood pressure.

[0103] By using the observation matrix and the design matrix to construct a health assessment model, each row of signal values ​​is associated with its lagged historical value, which can capture the time dependence of the signal and reflect the dynamic characteristics of the health signal. By fitting the design matrix and the observation matrix through the least squares method, the time correlation characteristics of the signal are obtained, providing clear parameter support for the prediction and abnormal analysis of health signals. The spectrum distribution is calculated according to the AR model to reflect the distribution of signal energy in the frequency domain, accurately quantify the spectrum characteristics of health signals, and capture energy anomalies at specific frequencies. The mean, variance, skewness and kurtosis are calculated based on the PSD, and the high-dimensional spectrum information is condensed into a fixed-dimensional feature vector, reducing data complexity and facilitating subsequent classification and health status analysis. By estimating the model noise, the deviation between the signal and the model prediction value is quantified, sudden abnormal signals are identified, and the sensitivity and robustness of the health monitoring system to signal anomalies are enhanced.

[0104] S2, calculates the contribution factor and determines the dynamic health score based on the frequency characteristics of different frequency bands, calculates the posterior probability of the health state based on the conditional probability of the health score, and calculates the frequency band offset to construct the abnormal frequency band set;

[0105] Preferably, the contribution factor is calculated based on the frequency characteristics of different frequency bands and a dynamic health score is determined. The posterior probability of the health state is calculated based on the conditional probability of the health score, and the frequency band offset is calculated to construct an abnormal frequency band set, including:

[0106] The contribution factor is calculated based on the statistical eigenvalues ​​of different frequency bands based on the frequency eigenvector, which is expressed as:

[0107] ;

[0108] in Indicates the contribution of the ith frequency band to the overall abnormal health status, represents the jth statistical eigenvalue of the i-th frequency band, represents the jth statistical eigenvalue of the i-th frequency band in the healthy state, which is determined by calculating the sum of the mean and standard deviation of the healthy data set, n represents the total number of frequency bands, and m represents the number of statistical features of the frequency eigenvector;

[0109] Combining the contribution factors and spectrum characteristics of each frequency band, the dynamic health score is calculated and expressed as:

[0110] ;

[0111] Where DHS represents the overall health status value;

[0112] Construct a health status classification model based on the dynamic Bayesian formula, calibrate the health status of students according to the historical health status value data, including normal, sub-health and abnormal, and define the kth health status as The occurrence frequency of each state in the historical health status value data is statistically analyzed as the prior probability of the corresponding student health state;

[0113] The health score DHS data under each student's health status are distributed and fitted, and the mean and standard deviation are calculated. The conditional probability of the health score DHS under the premise that the student is currently in a healthy state is calculated, which is expressed as:

[0114] ;

[0115] in represents the probability that the health score belongs to the kth health state, represents the standard deviation of health scores under the kth health state, represents the mean health score under the kth health state;

[0116] Calculate the posterior probability of the student's health status under different health conditions, expressed as:

[0117] ;

[0118] in represents the posterior probability that the health score belongs to the kth health state, represents the prior probability of the kth health state, K represents the total number of health states, represents the kth health state, represents the probability that the health score belongs to the kth health state;

[0119] The health state with the maximum probability is selected based on the posterior probability to determine the predicted health state of the student with the corresponding health score;

[0120] Count different health states and corresponding spectra in historical data, and use the spectra as reference spectra for the corresponding health states;

[0121] The health status discrimination is verified by the distribution distance between the mean and standard deviation of the reference spectrum. The sum of the mean and standard deviation of the calibrated reference spectrum distribution distance value is used as the verification threshold to calculate the distribution distance verification, which is expressed as:

[0122] ;

[0123] Where D represents the distribution distance value, and Respectively represent the reference spectrum of the frequency f of the ith frequency band of the sequence number 1 and sequence number 2 in the health state, and Respectively represent the variance of the reference spectrum of the health state of sequence number 1 and sequence number 2;

[0124] If the distribution distance value is greater than or equal to the verification threshold, it is determined that the two health states have obvious distinction, and the corresponding reference spectrum is used accordingly;

[0125] Based on the predicted health status, the corresponding reference spectrum is selected to calculate the frequency band offset, which is expressed as:

[0126] ;

[0127] in represents the spectrum offset of the i-th frequency band, and represents the frequency range of the i-th frequency band and the i-1-th frequency band, represents the spectrum of frequency f in the i-th frequency band, represents the reference spectrum of frequency f in the i-th frequency band;

[0128] The deviation threshold is calculated based on the mean and double standard deviation of the offset data of historical samples. If the spectrum offset is greater than or equal to the deviation threshold, it means that the deviation is large, and the frequency band i is marked as abnormal. The judgment of the marked abnormality and the corresponding offset data and frequency feature vector are combined into an abnormal frequency band set.

[0129] Based on the contribution factor calculation of frequency eigenvectors, the statistical eigenvalues ​​of each frequency band are used to calculate its contribution to overall abnormal health status, making the Dynamic Health Score (DHS) more targeted, accurately quantifying the abnormal contribution of frequency bands, and improving the credibility of health status assessments. Through the construction of the dynamic health score, combining the contribution factors of each frequency band with the spectrum characteristics, it dynamically reflects the overall fluctuations of health status, making the health score time-sensitive and able to capture changing trends in health status in real time, providing quantitative indicators to facilitate medical personnel to quickly understand individual health levels. The classification model based on the dynamic Bayesian formula accurately predicts individual health status (normal, sub-healthy, abnormal), supporting personalized health management. It uses historical health status value data for calibration and combines prior probability, conditional probability, and posterior probability to classify student health status. By fitting the distribution of historical health scores (mean and standard deviation) and combining it with the dynamic health score to calculate conditional probability, the classification model's ability to distinguish health states and enhances its sensitivity to abnormal health states. Based on the reference spectrum corresponding to the predicted health state, the frequency band offset is calculated. Identifying abnormal frequency bands can quickly locate the source of abnormal signals and support refined health monitoring.

[0130] S3 uses support vector machines to classify abnormal frequency bands, draws a heat map to show frequency band anomalies, and generates anomaly reports for encrypted transmission;

[0131] Preferably, using a support vector machine to perform abnormal classification on abnormal frequency bands includes:

[0132] Using a support vector machine (SVM) pre-trained with historically labeled anomaly type data, we perform anomaly classification on each abnormal frequency band i in the abnormal frequency band set, identifying specific anomaly types, including signal mutations (such as equipment failure or sudden health conditions), obvious deviations within the frequency range, and significant changes in power spectrum concentration. Conventional anomaly types, including signal changes caused by motion and slight spectral deviations, are also identified as the classification results for each abnormal frequency band.

[0133] Through support vector machines (SVM), abnormality types are classified, abnormal frequency bands are classified, specific abnormalities (such as signal mutations and changes in spectrum concentration) and common abnormalities (such as slight fluctuations) are identified, and the abnormality types are accurately classified to provide targeted guidance for medical intervention and improve the intelligence level of system abnormality detection. Through the annotation and learning of historical data, the system's adaptability to personalized health status is enhanced, supporting long-term health monitoring and trend analysis.

[0134] Furthermore, the position and offset of the abnormal frequency band are plotted on the frequency axis to construct a heat map to show the degree of frequency band abnormality, where the horizontal axis represents time, the vertical axis represents frequency, and the color represents the corresponding frequency offset;

[0135] Based on the classification results of the abnormal frequency bands, different colors are used to distinguish and mark specific anomalies and general anomalies.

[0136] By plotting the position and offset of abnormal frequency bands on the frequency axis and marking the degree of abnormality with color, an intuitive visualization of abnormal frequency bands is provided, making it easier for users to quickly understand the monitoring results. Different colors are used to distinguish specific abnormalities from general abnormalities, enhancing the interactive experience of the monitoring system.

[0137] Furthermore, an exception report is generated for encrypted transmission, including:

[0138] Corresponding abnormality reports are generated for the heat map data through a spreadsheet tool, and sent to the student health supervision terminal using SSL / TLS protocol encrypted communication. At the same time, the generated abnormality report data is stored in the cloud, and the health scores, abnormal frequency band sets and abnormal classification results of different students are sorted out based on time series to generate health reports for the corresponding students through a spreadsheet tool.

[0139] By utilizing heat map data, detailed reports of abnormal frequency bands are automatically generated, including frequency offset, abnormal classification results and time information. The abnormal reports are sent to the student health monitoring terminal through the secure SSL / TLS protocol to ensure the confidentiality and integrity of the data during transmission. The health scores, abnormal frequency band sets and classification results of different students are sorted out based on time series to form the health trajectory of individual students, provide data support for personalized health interventions, and improve health management effects.

[0140] This embodiment also provides a system for a multi-dimensional student physical health monitoring method, including:

[0141] The data acquisition module uses multimodal sensors to collect students' real-time physiological data and pre-processes the collected raw data;

[0142] The autoregressive modeling module performs synchronous compression transformation on the signal subset, generates a time-frequency map matrix, and builds an autoregressive AR model based on the time-frequency features;

[0143] The health scoring module calculates dynamic health scores based on frequency eigenvectors and frequency band contribution factors. It also calculates posterior probabilities based on the dynamic Bayesian formula and the conditional probability of the health scores to predict students' health status.

[0144] The anomaly detection module calculates the frequency band offset based on the reference spectrum corresponding to the healthy state, identifies abnormal frequency bands, constructs an abnormal frequency band set, uses a support vector machine (SVM) to classify the abnormal frequency bands, and generates a heat map showing the time, frequency, and offset of the abnormal frequency bands;

[0145] The report generation module generates abnormal reports and transmits them to the student health monitoring terminal through the SSL / TLS encryption protocol. All health scores, abnormal frequency band sets and classification results are stored in the cloud to support long-term health management.

[0146] This embodiment also provides a computer device suitable for the multi-dimensional student physical health monitoring method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the multi-dimensional student physical health monitoring method proposed in the above embodiment.

[0147] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0148] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-dimensional student physical health monitoring method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0149] In summary, the present invention constructs a health assessment model by using the observation matrix and the design matrix, so that each row of signal values ​​is associated with its lagged historical value, which can capture the time dependence of the signal and reflect the dynamic characteristics of the health signal. The mean, variance, skewness and kurtosis are calculated based on the PSD, and the high-dimensional spectrum information is condensed into a fixed-dimensional feature vector, reducing data complexity and facilitating subsequent classification and health status analysis. By estimating the model noise, the deviation between the signal and the model prediction value is quantified, sudden abnormal signals are identified, and the sensitivity and robustness of the health monitoring system to signal anomalies are enhanced. By constructing a dynamic health score, combining the contribution factors of each frequency band and the spectrum characteristics, the overall fluctuation of the health status is dynamically reflected, making the health score time-sensitive, and the changing trend of the health status can be captured in real time. It provides a quantitative indicator to facilitate medical personnel to quickly understand the individual's health level. The classification model based on the dynamic Bayesian formula accurately predicts the individual's health status. The conditional probability is calculated in combination with the dynamic health score, which improves the classification model's ability to distinguish health states and enhances its sensitivity to abnormal health states. Based on the reference spectrum corresponding to the predicted health state, the frequency band offset is calculated. Identifying abnormal frequency bands can quickly locate the source of abnormal signals and support refined health monitoring.

[0150] Example 2, referring to Table 1, is the second embodiment of the present invention. In order to further verify the technical solution of the present invention, experimental simulation data of a multi-dimensional student physical health monitoring method is provided.

[0151] The students' physiological status data are collected through multimodal sensors, as shown in Table 1 below:

[0152]

[0153] Table 1 Physiological status data table

[0154] Standardize each column of data, perform sliding window processing, and calculate the average value within each window;

[0155] By performing Fourier transform on the data in the sliding window, the energy distribution of multiple time points and frequency points is extracted, as shown in Table 2 below:

[0156]

[0157] Table 2 Energy distribution data

[0158] The frequency power spectrum matrix generated by FSST is used as input to further build the AR model. The order p is determined to be 2 by the Akaike Information Criterion (AIC) to generate the observation matrix Y and the design matrix A.

[0159] ;

[0160] ;

[0161] Compute the AR coefficient vector, where:

[0162] ;

[0163] ;

[0164] ;

[0165] Calculate the power spectral density, expressed as:

[0166] ;

[0167] The power spectrum density calculation results are shown in Table 3 below:

[0168]

[0169] Table 3 Power spectral density data table

[0170] Calculate the contribution factors of different frequency bands, as shown in Table 4 below:

[0171]

[0172] Table 4 Contribution factor data table

[0173] And calculate the dynamic health score, expressed as:

[0174] ;

[0175]

[0176] Based on historical health status data, three health states are defined as normal, subhealthy, and abnormal, and the prior probabilities are calculated as 0.5, 0.3, and 0.2 respectively;

[0177] A normal distribution was fitted for the DHS value of each health state, the mean and standard deviation were calculated, and the conditional probability was calculated, which was expressed as:

[0178] ;

[0179] ;

[0180] ;

[0181] According to the state distribution of historical data, the posterior probability is calculated and expressed as:

[0182] ;

[0183] ;

[0184] ;

[0185] The details are shown in Table 5 below:

[0186]

[0187] Table 5 Probability calculation data table

[0188] Based on the maximum posterior probability, the health status is predicted to be normal;

[0189] For each frequency band, the offset is calculated based on the difference between the current spectrum and the reference spectrum, and the deviation threshold of each frequency band is calculated based on the mean and twice the standard deviation of historical samples for abnormal judgment. The pre-trained support vector machine (SVM) pre-trained model is used to classify and predict each abnormal frequency band, and generate a heat map data table, as shown in Table 6 below:

[0190]

[0191] Table 6 Heat map data table

[0192] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-dimensional student physical health monitoring method, characterized in that: include: A multimodal sensor network was used to collect students' physiological data, including heart rate, blood oxygen, blood pressure, and motion trajectory data. Signal subsets were constructed from the collected data, and time-frequency decomposition was performed to construct a time-frequency graph matrix. Autoregressive modeling was performed to extract the observation matrix and design matrix, and frequency eigenvectors were constructed. Calculate the contribution factor and determine the dynamic health score based on the frequency characteristics of different frequency bands. Calculate the posterior probability of the health state based on the conditional probability of the health score. Calculate the frequency band offset to construct the abnormal frequency band set. Use support vector machines to classify abnormal frequency bands, draw heat maps to show frequency band anomalies, and generate abnormality reports for encrypted transmission; The calculation of contribution factors and determination of dynamic health scores based on frequency characteristics of different frequency bands, calculation of posterior probability of health status based on conditional probability of health scores, and calculation of frequency band offsets include: Calculate contribution factors based on the statistical eigenvalues ​​of different frequency bands based on frequency eigenvectors; The dynamic health score is calculated by combining the contribution factors and spectrum characteristics of each frequency band; A health status classification model based on the dynamic Bayesian formula is constructed to calibrate the health status of students according to historical health status value data, including normal, sub-healthy and abnormal. The frequency of occurrence of each state in the historical health status value data is statistically calculated as the prior probability of the corresponding student health status. Fit the distribution of the health score DHS data under each student's health status, calculate the mean and standard deviation, and calculate the conditional probability of the health score DHS under the premise that the student is currently in a healthy state; Calculate the posterior probability of students’ health status under different health conditions; The health state with the maximum probability is selected based on the posterior probability to determine the predicted health state of the student with the corresponding health score; Count different health states and corresponding spectra in historical data, and use the spectra as reference spectra for the corresponding health states; The discrimination of health states was verified by the distribution distance between the mean and standard deviation of the reference spectrum; Based on the predicted health status, the corresponding reference spectrum is selected to calculate the frequency band offset, which is expressed as: ； in represents the spectrum offset of the i-th frequency band, and represents the frequency range of the i-th frequency band and the i-1-th frequency band, represents the spectrum of frequency f in the i-th frequency band, represents the reference spectrum of frequency f in the i-th frequency band.

2. A multi-dimensional student physical health monitoring method according to claim 1, characterized in that: The collecting data to construct a signal subset includes: Use a multimodal sensor network to collect students' physiological data, including heart rate, blood oxygen, blood pressure, and movement trajectory data; Based on different physiological data items, a corresponding data matrix is ​​established according to the collection time, and the collected data of each sensor is standardized; The sensor data collected continuously in time is subjected to a sliding window operation, the data is segmented according to the sampling interval, and a signal subset consisting of sensor data with different sampling intervals is constructed.

3. A multi-dimensional student physical health monitoring method according to claim 2, characterized in that: The time-frequency decomposition is performed to construct a time-frequency graph matrix, the autoregressive modeling is performed to extract the observation matrix and the design matrix, and the frequency feature vector is constructed, including: Perform Fourier synchronous compression transform on the subset of sensor signals to perform time-frequency decomposition and output the time-frequency map matrix of the corresponding sensor signals , represents the energy distribution of the signal at time t and frequency f; Autoregressive spectrum AR modeling based on time-frequency diagram matrix; The observation matrix Y and design matrix A are extracted from the signal data collected by the sensor and expressed as: ； ； Where A represents the setting matrix composed of the signal history values, each row represents the previous p historical values ​​of the current signal value, each column corresponds to the data value of a lag time step, p represents the AR order, and represents the current signal value Associated with the signal value of how many historical moments, Y represents the observation matrix, containing the current signal value corresponding to each row in the design matrix A , represents the signal value at time t=p, Represents the signal value at time point t=Np, represents the target signal value after p steps of lag, and N represents the total number of sampling points in the entire signal time series; The AR coefficient vector is calculated by the least squares method based on the observation matrix Y and the design matrix A, which is expressed as: ； Where a represents the AR coefficient, represents the transpose calculation of the design matrix; The model noise is determined based on the AR coefficient vector and is expressed as: ； in represents the signal value at time t, represents the kth AR coefficient, represents the noise at time t; Then calculate the power spectral density, which is expressed as: ； in Represents the distribution of signal power at frequency f in the frequency domain as the spectrum of the corresponding frequency, represents the variance of the noise, j represents the imaginary unit; calculate The frequency feature vector is constructed from the mean, variance, skewness and kurtosis of .

4. A multi-dimensional student physical health monitoring method according to claim 3, characterized in that: The step of calculating the contribution factor and determining the dynamic health score based on the frequency characteristics of different frequency bands, calculating the posterior probability of the health state based on the conditional probability of the health score, and calculating the frequency band offset to construct an abnormal frequency band set includes: The contribution factor is calculated based on the statistical eigenvalues ​​of different frequency bands based on the frequency eigenvector, which is expressed as: ； in Indicates the contribution of the ith frequency band to the overall abnormal health status, represents the jth statistical eigenvalue of the i-th frequency band, represents the jth statistical eigenvalue of the i-th frequency band in the healthy state, which is determined by calculating the sum of the mean and standard deviation of the healthy data set, n represents the total number of frequency bands, and m represents the number of statistical features of the frequency eigenvector; Combining the contribution factors and spectrum characteristics of each frequency band, the dynamic health score is calculated and expressed as: ； Where DHS represents the overall health status value; Calculate the conditional probability of the health score DHS under the premise that the student is currently in a healthy state, expressed as: ； in represents the probability that the health score belongs to the kth health state, represents the standard deviation of health scores under the kth health state, represents the mean health score under the kth health state; Calculate the posterior probability of the student's health status under different health conditions, expressed as: ； in represents the posterior probability that the health score belongs to the healthy state, Indicates health status The prior probability of K represents the total number of health states, represents the kth health state, represents the probability that the health score belongs to the kth health state; Based on the predicted health status, the corresponding reference spectrum is selected to calculate the frequency band offset; The deviation threshold is calculated based on the mean and double standard deviation of the offset data of historical samples. If the spectrum offset is greater than or equal to the deviation threshold, it means that the deviation is large, and the frequency band i is marked as abnormal. The judgment of the marked abnormality and the corresponding offset data and frequency feature vector are combined into an abnormal frequency band set.

5. A multi-dimensional student physical health monitoring method according to claim 4, characterized in that: The use of a support vector machine to perform abnormal classification on abnormal frequency bands includes: Use the support vector machine (SVM) pre-trained with historically labeled anomaly type data to perform anomaly classification on each abnormal frequency band i in the abnormal frequency band set; Specific anomaly types and general anomaly types are identified as classification results for each anomaly frequency band.

6. A multi-dimensional student physical health monitoring method according to claim 5, characterized in that: The heat map shows frequency band anomalies, including: Plot the position and offset of the abnormal frequency band on the frequency axis, and construct a heat map to show the degree of frequency band abnormality. The horizontal axis represents time, the vertical axis represents frequency, and the color represents the corresponding frequency offset. Based on the classification results of the abnormal frequency bands, different colors are used to distinguish and mark specific anomalies and general anomalies.

7. A multi-dimensional student physical health monitoring method according to claim 6, characterized in that: The generating of the abnormal report for encrypted transmission includes: Corresponding abnormality reports are generated for the heat map data through a spreadsheet tool, and sent to the student health supervision terminal using SSL / TLS protocol encrypted communication. At the same time, the generated abnormality report data is stored in the cloud, and the health scores, abnormal frequency band sets and abnormal classification results of different students are sorted out based on time series to generate health reports for the corresponding students through a spreadsheet tool.

8. A system for a multi-dimensional student physical health monitoring method, based on the multi-dimensional student physical health monitoring method according to any one of claims 1 to 7, characterized in that: include: The data acquisition module uses multimodal sensors to collect students' real-time physiological data and pre-processes the collected raw data; The autoregressive modeling module performs synchronous compression transformation on the signal subset, generates a time-frequency map matrix, and builds an autoregressive AR model based on the time-frequency features; The health scoring module calculates dynamic health scores based on frequency eigenvectors and frequency band contribution factors. It also calculates posterior probabilities based on the dynamic Bayesian formula and the conditional probability of the health scores to predict students' health status. The anomaly detection module calculates the frequency band offset based on the reference spectrum corresponding to the healthy state, identifies abnormal frequency bands, constructs an abnormal frequency band set, uses a support vector machine (SVM) to classify the abnormal frequency bands, and generates a heat map showing the time, frequency, and offset of the abnormal frequency bands; The report generation module generates abnormal reports and transmits them to the student health monitoring terminal through the SSL / TLS encryption protocol. All health scores, abnormal frequency band sets and classification results are stored in the cloud to support long-term health management.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-dimensional student physical health monitoring method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-dimensional student physical health monitoring method according to any one of claims 1 to 7 are implemented.

Images