Search

CN-121971053-A - Physical and mental index dynamic characteristic multi-mode recognition analysis system based on artificial intelligence

CN121971053ACN 121971053 ACN121971053 ACN 121971053ACN-121971053-A

Abstract

The invention discloses a body and mind index dynamic characteristic multi-mode identification analysis system based on artificial intelligence, which relates to the technical field of biomedical signal processing and comprises a multi-source data acquisition module, a dynamic risk assessment module, a blood flow abnormality analysis module, a multi-mode risk layering module and an intelligent early warning module, wherein the multi-source data acquisition module is used for collecting respiratory, heart rate and blood oxygen data, the dynamic risk assessment module is used for predicting risks by using a long-term and short-term memory network, the blood flow abnormality analysis module is used for identifying blood flow abnormality by using a convolutional neural network, the multi-mode risk layering module is used for assessing risk levels, and the intelligent early warning module is used for generating reports and taking early warning measures. According to the invention, through multi-mode identification analysis, comprehensive, real-time and accurate monitoring of physical and psychological indexes is realized, the risk assessment accuracy is improved, potential health risks are early-warned in time, powerful support is provided for clinical decisions, the public health management level is improved, the medical cost is reduced, and the life quality is improved.

Inventors

  • LIU JICHUN
  • WANG GUANGQI
  • ZHANG ZEXING
  • LIU ZUGUANG
  • LI KEJIAN

Assignees

  • 上海轩任科技集团有限公司

Dates

Publication Date
20260505
Application Date
20260210

Claims (10)

  1. 1. The system is characterized by comprising a multi-source data acquisition module, a dynamic risk assessment module, a blood flow abnormality analysis module, a multi-mode risk layering module and an intelligent early warning module; The multi-source data acquisition module acquires physical and psychological index dynamic characteristics of a detected person, wherein the physical and psychological index dynamic characteristics comprise respiratory frequency data, heart rate data and blood oxygen saturation data; The dynamic risk assessment module is used for constructing a risk assessment model based on respiratory frequency data and heart rate data of a detected person by using a long-short-period memory network algorithm, capturing a dynamic change rule of respiratory frequency and heart rate and outputting a preliminary risk score; the blood flow abnormality analysis module is used for constructing a blood flow abnormality analysis model based on blood oxygen saturation data by utilizing a convolutional neural network algorithm and extracting a blood flow distribution abnormality mode of a detected person from blood oxygen saturation time sequence characteristics; The multi-mode risk layering module is used for constructing a dynamic risk layering model based on the risk assessment model and the blood flow abnormality analysis model, importing the latest collected preliminary risk score and blood flow distribution abnormality mode data set into the dynamic risk layering model, and outputting a risk grade; The intelligent early warning module generates a risk thermodynamic diagram and an individual trend report based on the preliminary risk score and the risk grade, and adopts a corresponding early warning mechanism according to the risk grade.
  2. 2. The system for dynamically identifying and analyzing the multiple modes of physical and mental index based on artificial intelligence according to claim 1, wherein the multi-source data acquisition module is characterized in that the respiratory rate data, heart rate data and blood oxygen saturation data acquisition process of the detected person comprises the following steps: The integrated millimeter wave radar sensor is deployed at the front end of detection equipment by adopting the Doppler effect principle, keeps a distance of 0.5 to 1.5 meters from a detected person and is opposite to a thoracic cavity region, the working frequency band of the radar is set to be 60 to 64 GHz, the transmitting power is not more than 10mW, the frequency spectrum peak frequency is calculated as the respiratory frequency by transmitting millimeter waves and receiving micro signals reflected by the surface of the thoracic cavity, the millimeter wave radar multiplexes the respiratory detection channel, and heart rate data is acquired by a signal time division multiplexing technology; The high-frame-rate infrared camera is deployed above a screen of the detection device and is aligned to the face area of the detected person, the infrared camera is integrated with a dual-wavelength light source, subcutaneous blood flow signals of the detected person are collected through reflection type photoplethysmography, the absorption ratio of red light to infrared light is calculated, and the blood oxygen saturation is deduced based on an empirical calibration curve.
  3. 3. The system for dynamically identifying and analyzing the multiple modes of physical and mental indicators based on artificial intelligence according to claim 2, wherein the process of preprocessing the collected respiratory rate data, heart rate data and blood oxygen saturation data by the multi-source data collecting module comprises the following steps: Band-pass filtering is carried out on thoracic cavity micro-motion signals received by the millimeter wave radar, signal frequency spectrums are calculated through Fourier transformation, and frequencies corresponding to frequency spectrum peaks are extracted to serve as respiratory fundamental frequencies; Multiplexing a respiration detection channel of the millimeter wave radar, separating a heartbeat signal by a time division multiplexing technology, extracting a heartbeat cycle signal by time-frequency analysis based on chest micro motion caused by heartbeat, and dividing the heartbeat cycle signal by adopting a sliding window with the length of 5 seconds; Carrying out normalization processing on the dual-wavelength reflection signals acquired by the infrared camera, extracting arterial blood pulse components of photoelectric volume pulse waves through filtering, separating alternating current components and direct current components of red light and infrared light, and analyzing blood oxygen saturation data by adopting a sliding window with the length of 10 seconds; The micro-vibration of the skin surface is detected through the phase change of the millimeter wave radar heartbeat cycle signal, and the skin conductivity data and the body temperature data of the detected person are collected by combining the light absorption characteristic of the infrared camera dual-wavelength reflection signal.
  4. 4. The artificial intelligence based dynamic body and mind index characteristic multi-modal identification and analysis system as claimed in claim 3, wherein the multi-source data acquisition module performs characteristic extraction on the preprocessed respiratory rate data, heart rate data and blood oxygen saturation data, and the process comprises the following steps: On the basis of respiratory fundamental frequency, harmonic signals of integer multiple frequency bands are extracted, the ratio of the sum of harmonic energy to fundamental frequency energy is calculated, the respiratory harmonic energy ratio is obtained, the standard deviation of adjacent heartbeat intervals is calculated, the heartbeat signals are decomposed into low-frequency components and high-frequency components, the power ratio of the low-frequency components and the high-frequency components is calculated, the standard deviation of blood oxygen saturation in a sliding window is calculated, the dynamic change amplitude of the blood oxygen saturation is quantized, the local blood flow non-uniformity of a detected person is calculated based on the ratio of alternating current components to direct current components of photoelectric volume pulse waves, the fluctuation standard deviation of skin conductivity data is extracted, the dynamic trend slope of body temperature data is combined, the outward direction index of the detected person is obtained, and the logic thinking potential of the detected person is estimated based on the pearson correlation coefficient of the blood oxygen saturation and the heart rate variability.
  5. 5. The system for dynamically identifying and analyzing the multiple modes of physical and mental indicators based on artificial intelligence according to claim 4, wherein the process of constructing a risk assessment model and outputting a preliminary risk score by the dynamic risk assessment module comprises the following steps: Designing a framework of a long-period memory network model to construct a risk assessment model, wherein the long-period memory network model comprises an input layer, a hidden layer and an output layer, and the hidden layer comprises a forgetting gate, an input gate, a memory unit and an output gate; Inputting a time sequence characteristic sequence comprising respiratory fundamental frequency, respiratory harmonic energy ratio, adjacent heartbeat interval standard deviation and low-frequency and high-frequency power ratio to an input layer of a risk assessment model, and searching the dynamic relevance of respiratory and heart rate characteristics by a hidden layer, wherein a forgetting gate dynamically adjusts the weight of the respiratory and heart rate long-term dependence according to a historical memory state, the input gate screens short-term characteristics sensitive to the emotional state in the current input, updates a memory unit, and outputs the hidden state of the current moment to an output layer of a long-term memory network model; the output layer maps the hidden states into four-dimensional emotion probability distribution and three-dimensional character potential distribution, calculates the percentage difference value between the standard deviation of adjacent heartbeat intervals and the corresponding baseline value of healthy people to obtain heart rate variability deviation degree, adds normalization results of the heart rate variability deviation degree, the respiratory harmonic energy ratio and the blood oxygen fluctuation range according to weights to generate a preliminary risk score of 0-100, and outputs a dominant emotion label and a dominant character label through a normalization function, wherein the emotion label comprises calm, anxiety, depression and pressure, and the character label comprises exogenousness, logic thinking and pilot force; Calculating a pearson correlation coefficient of the respiration and the heart rate signals, inputting the pearson correlation coefficient as an additional characteristic into a risk assessment model, and adaptively adjusting a judgment threshold value of the respiration harmonic energy ratio and the heart rate variability based on historical data of a detected person; and inputting the new acquired data into a trained risk assessment model, outputting a preliminary risk score, an emotion label and a character label within 5 seconds, and predicting emotion and character deterioration trend of the detected person by the risk assessment model according to the score change rate of the continuous window.
  6. 6. The artificial intelligence based body and mind index dynamic characteristic multi-mode recognition analysis system according to claim 5, wherein the blood flow anomaly analysis module, using a convolutional neural network algorithm, constructs a blood flow anomaly analysis model comprising: adopting a convolutional neural network architecture, receiving a two-dimensional feature map of time step length X blood oxygen value converted by blood oxygen saturation time sequence data, dividing the blood oxygen saturation time sequence data according to a time window with the length of 10 seconds, generating a spectrogram through short-time Fourier transform, mapping the spectrogram into a gray image with 128X 128 pixels, constructing a blood flow anomaly analysis model, extracting a local blood flow fluctuation mode by a convolutional kernel in the convolutional neural network, identifying high-frequency anomaly signals, fusing space-time information, capturing the spatial correlation of blood flow distribution, positioning the abnormal region of the face of a detected person based on the abnormal ratio of alternating current components to direct current components of photoelectric volume pulse waves, combining the feature map output by a risk evaluation model, and outputting the abnormal probability of the blood flow of the detected person through a fully connected network; calculating a pearson correlation coefficient of blood oxygen fluctuation and heart rate variability, if the pearson correlation coefficient of the blood oxygen fluctuation and the heart rate variability is lower than 0.3, judging that the blood flow abnormality and the autonomic nervous disorder have a synergistic effect, enhancing the abnormal region sensitivity of a blood flow abnormality analysis model, and adaptively setting a blood flow non-uniformity threshold based on historical data of a detected person; The newly acquired blood oxygen data is preprocessed and then input into a trained blood flow abnormality analysis model, a blood flow abnormality score and a blood flow distribution thermodynamic diagram are output within 5 seconds, and a blood flow abnormality area of a detected person is marked.
  7. 7. The system for dynamically identifying and analyzing the multiple modes of physical and mental indicators based on artificial intelligence according to claim 6, wherein the process for constructing the dynamic risk stratification model comprises the following steps: Splicing the preliminary risk score, the blood flow abnormality feature and the character potential index into a fusion feature vector, and adjusting the weights of the preliminary risk score, the blood flow abnormality feature and the character potential index according to the pearson correlation coefficient of the respiratory harmonic energy ratio and the blood oxygen fluctuation standard deviation to construct a dynamic risk layering fusion model; And inputting the fusion feature vector into a dynamic risk layering model, if the pearson correlation coefficient is lower than 0.3, judging that respiration and blood flow are abnormal and act synergistically, and increasing the weight of blood flow abnormal features, if the pearson correlation coefficient is higher than 0.3, distributing time sequence weight to be 0.6 according to a base line, blood flow abnormal feature weight to be 0.3, character potential index weight to be 0.1, and calculating comprehensive risk score= (preliminary risk score multiplied by corresponding weight) + (blood flow abnormal score multiplied by corresponding weight) - (character potential index multiplied by corresponding weight) based on the fusion feature vector and the weight distribution result, wherein the outward direction score is more than 70 time, the comprehensive risk score is reduced by 10%, the logic thinking score is more than 75 time, and the risk grade is automatically reduced by one level.
  8. 8. The system for dynamically identifying and analyzing the multiple modes of physical and mental indicators based on artificial intelligence according to claim 7, wherein the process of outputting the risk level comprises the steps of: the method comprises the steps of inputting a preliminary risk score comprising a respiratory harmonic energy ratio, a character potential index, a heart rate variability deviation degree and a blood oxygen fluctuation range and an abnormal region label comprising a blood oxygen standard deviation, local blood flow non-uniformity and a blood flow distribution thermodynamic diagram into a dynamic risk layering model, outputting a comprehensive risk score, dividing the comprehensive risk score into red early warning with the value higher than 75 according to historical data of a tested person, orange early warning with the comprehensive risk score between 50 and 75, yellow early warning with the comprehensive risk score between 30 and 50, and green with the comprehensive risk score lower than 30.
  9. 9. The system for dynamically identifying and analyzing multiple modes of physical and mental index based on artificial intelligence according to claim 8, wherein the intelligent early warning module generates a risk thermodynamic diagram and an individual trend report, and the process comprises the following steps: Integrating comprehensive risk scores output by a dynamic risk layering model, combining geographic position data of detected persons in the region, counting the number of risk persons in each region according to time window based on respiration, heart rate and blood oxygen time sequence data aligned by hardware time stamps, generating a risk distribution matrix, mapping the risk distribution matrix into a thermodynamic diagram through a geographic information system of a cloud platform according to a color coding rule, updating an abnormally high sending region in real time, calculating an average pearson correlation coefficient of respiration-heart rate-blood oxygen in the region, marking as physiological cooperativity abnormality if the average pearson correlation coefficient is lower than 0.3, superposing the physiological cooperativity abnormality to the risk thermodynamic diagram, and lifting the risk level and carrying out data resampling verification if the average pearson correlation coefficient is higher than 0.3; Drawing a time sequence index curve and a comprehensive risk score line graph showing the respiratory harmonic energy ratio, the heart rate variability deviation degree, the blood oxygen fluctuation standard deviation and the outward sex index dynamic change, marking event nodes, analyzing the comprehensive risk score change rate based on a sliding window, judging as a mood deterioration period if the single-day comprehensive risk score rises to 15 minutes, marking with red shadows in an individual trend report, and marking out-of-range data points in the abnormal period as yellow warning symbols by combining the physiological relevance of the respiratory frequency and the heart rate; If the logic thinking score in the personality potential distribution is more than 75 points, the high logic thinking potential is marked in the report, and a logic reasoning training tool is recommended, and if the leader score is more than 70 points, the testee is recommended to participate in team collaborative simulation training.
  10. 10. The artificial intelligence based body and mind index dynamic characteristic multi-mode recognition analysis system according to claim 9, wherein the intelligent early warning module adopts a corresponding early warning mechanism process comprising: If the comprehensive risk score exceeds 75 minutes and the conditions that the respiratory harmonic energy ratio is more than 35%, the blood oxygen fluctuation standard deviation is more than 3% and the respiratory-heart rate pearson correlation coefficient is less than 0.2 appear, the red early warning is judged, the dominant risk factors are marked, and the hardware equipment is triggered to automatically lock abnormal period data; if the comprehensive risk score is between 50 and 75 minutes and the heart rate variability deviation is more than 60 percent, judging that the heart rate variability deviation is orange early warning, and adjusting a threshold value based on historical data of a tested person; if the comprehensive risk score is between 30 and 50 minutes, the standard deviation of blood oxygen fluctuation is between 1 and 2 percent, and the respiratory-heart rate correlation coefficient is between 0.2 and 0.4, judging that the comprehensive risk score is yellow early warning, and marking an individual trend report of a tested person with a label to be observed; If the comprehensive risk score is lower than 30 minutes, the respiratory harmonic energy ratio is less than 15 percent, the standard deviation of blood oxygen fluctuation is less than 1 percent, the standard deviation is judged to be green and normal, and the baseline file is updated monthly.

Description

Physical and mental index dynamic characteristic multi-mode recognition analysis system based on artificial intelligence Technical Field The invention relates to the technical field of biomedical signal processing, in particular to a body and mind index dynamic characteristic multi-mode recognition analysis system based on artificial intelligence. Background The dynamic characteristics of the physical and psychological indexes are mainly represented by the differences, continuity and variability of the physical and psychological indexes of the individuals along with the time and situation changes, firstly, the differences are the differences of the physical and psychological indexes of different individuals, the differences can be caused by various factors such as inheritance, environment, life style and the like, so that the physical and psychological states of each person are unique and need to be individually concerned and evaluated, secondly, the continuity is a continuous changing process rather than static unchanged, the continuity requires that when the physical and psychological indexes are evaluated, people pay attention to the development trend and the change rule of the physical and psychological indexes so as to discover problems in time and take measures, finally, the variability means that the physical and psychological states can be changed, by means of adjusting the life style, improving the environment, carrying out psychological treatment and the like, the physical and psychological states can be positively influenced, the physical and psychological states can be promoted to develop towards healthier and more positive directions, and the dynamic characteristics of the physical and psychological indexes can remind people to treat the physical and psychological states of the individuals in a dynamic and comprehensive view angle, so that the physical and psychological states of the individuals can be better promoted to develop the health. In order to solve the problems of insufficient accuracy and comprehensiveness of physical and mental health monitoring and risk assessment, the prior art mainly adopts regular physical examination, questionnaire investigation and other modes for processing, however, the mode can only capture static physiological parameters and psychological states, and cannot comprehensively and real-timely reflect physical and mental dynamic characteristics of an individual, so that in practical application, the conditions of untimely and inaccurate early warning of potential health risks often occur, further problems of delayed illness state, poor treatment effect and the like are caused, and in order to overcome the limitations, a multi-mode identification and analysis system for physical and mental index dynamic characteristics based on artificial intelligence is provided. Disclosure of Invention The invention aims to provide a body and mind index dynamic characteristic multi-mode recognition analysis system based on artificial intelligence so as to solve the problems in the background technology. In order to solve the technical problems, the invention adopts the technical scheme that the body and mind index dynamic characteristic multi-mode identification analysis system based on artificial intelligence comprises a multi-source data acquisition module, a dynamic risk assessment module, a blood flow abnormality analysis module, a multi-mode risk layering module and an intelligent early warning module; The multi-source data acquisition module acquires physical and psychological index dynamic characteristics of a detected person, wherein the physical and psychological index dynamic characteristics comprise respiratory frequency data, heart rate data and blood oxygen saturation data; The dynamic risk assessment module is used for constructing a risk assessment model based on respiratory frequency data and heart rate data of a detected person by using a long-short-period memory network algorithm, capturing a dynamic change rule of respiratory frequency and heart rate and outputting a preliminary risk score; the blood flow abnormality analysis module is used for constructing a blood flow abnormality analysis model based on blood oxygen saturation data by utilizing a convolutional neural network algorithm and extracting a blood flow distribution abnormality mode of a detected person from blood oxygen saturation time sequence characteristics; The multi-mode risk layering module is used for constructing a dynamic risk layering model based on the risk assessment model and the blood flow abnormality analysis model, importing the latest collected preliminary risk score and blood flow distribution abnormality mode data set into the dynamic risk layering model, and outputting a risk grade; The intelligent early warning module generates a risk thermodynamic diagram and an individual trend report based on the preliminary risk score and the risk grade, and adopts a corresponding early w