Whole-cycle electrocardiogram acquisition and diagnosis auxiliary system based on millimeter wave radar

Abstract

The invention relates to the technical field of health monitoring and management, and discloses a full-cycle electrocardiogram acquisition and diagnosis auxiliary system based on a millimeter wave radar, which comprises a hardware acquisition and real-time signal processing module used for acquiring vital sign data of a user through the millimeter wave radar and a multi-mode sensor and preprocessing the acquired original signal; the crowd classification and feature modeling module performs classification according to information input by the user and the collected data and constructs a personalized health portrait; and the edge calculation and real-time monitoring module is used for extracting electrocardiosignals and other vital sign data in real time and monitoring abnormal events. High-precision and multi-dimensional acquisition and personalized management of user health data are realized through a hardware acquisition and real-time signal processing module for acquiring vital sign data by a millimeter-wave radar and a multi-modal sensor and by combining a technical scheme of user classification and dynamic updating of health portraits.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring and management, and specifically to a full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar. Background Art

[0002] With the continuous improvement of people's health awareness, the demand for long-term monitoring of vital signs and health management is increasing day by day. However, existing health monitoring technologies and systems still have many deficiencies and cannot fully meet people's needs for high-precision, multi-dimensional, and personalized health management.

[0003] Existing health monitoring systems mostly rely on a single sensor for data collection. Heart rate belts are used to measure heart rate, respiratory belts are used to record respiratory rate, or activity data is recorded through smart bracelets. Although these technologies can collect single parameters under specific conditions, due to their limited collection dimensions, it is difficult to comprehensively reflect the comprehensive health status of users. At the same time, single sensors have poor adaptability in complex scenarios. For example, wearable devices are difficult to continuously collect reliable data when the user is not wearing them or for special populations such as pregnant women. Such deficiencies limit the integrity and accuracy of monitoring data and affect the effectiveness of subsequent health management.

[0004] In the prior art, health intervention programs usually rely on a preset fixed threshold trigger strategy to issue alarms or suggestions when the user's heart rate exceeds a specific value. However, this static trigger mechanism cannot adapt to the personalized needs and dynamic changes in the user's health status. For example, the heart rate fluctuation range of athletes is significantly different from that of the elderly or pregnant women, but existing systems often do not effectively classify and adjust for these differences.

[0005] In view of the deficiencies of the prior art, the present invention provides a full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar, which solves the problems of single-dimensional monitoring data of vital signs, inaccurate analysis of health status, lack of dynamic optimization and personalization of intervention strategies in the prior art.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar, including a hardware acquisition and real-time signal processing module, which is used to collect the user's vital sign data through a millimeter-wave radar and a multi-modal sensor, and preprocess the collected original signals; A population classification and feature modeling module, which classifies according to the information input by the user and the collected data, and constructs a personalized health profile; Edge computing and real-time monitoring module, for extracting electrocardiogram signals and other vital sign data in real time and monitoring abnormal events; Health trend analysis and personalized reporting module, for performing trend analysis on long-term health data and generating personalized health reports for users; Dynamic intervention and long-term feedback module, for generating health intervention suggestions based on the trend analysis results and optimizing the intervention strategy according to user feedback; Population segmentation and behavior pattern recognition module, for optimizing personalized health management by further segmenting user categories and identifying behavior patterns; Health anomaly prediction and personalized threshold optimization module, for predicting the risk of health anomalies and dynamically adjusting the monitoring threshold; Personalized health intervention simulation and optimization module, for simulating and optimizing health intervention strategies; Report interaction and data analysis module, for generating and presenting multi-level health reports and supporting user interaction analysis.

[0008] Preferably, the hardware acquisition and real-time signal processing module includes: Acquisition sub-module based on FMCW millimeter-wave radar, for acquiring chest micro-vibration signals; Infrared sensor sub-module, for acquiring the body surface temperature of the user; Sound sensor sub-module, for acquiring snoring signals and analyzing the breathing pattern of the user.

[0009] Preferably, the population classification and feature modeling module includes: User classification sub-module, for classifying into the elderly, pregnant women or athletes according to the information input by the user (including age, gender, occupation and health status) and the acquired health data, and further refining into categories such as good mobility and limited mobility; Health portrait dynamic update sub-module, for dynamically adjusting the user's health portrait by combining the real-time acquired data through the following formula:

[0010] where, is the current health portrait status, is the latest acquired data, is the update coefficient.

[0011] Preferably, the edge computing and real-time monitoring module includes: Signal separation sub-module, for extracting respiratory signals and heart rate signals through a band-pass filter, where the frequency range of the respiratory signal is 0.1–0.4Hz and the frequency range of the heart rate signal is 0.8–2.5Hz; Real-time anomaly detection sub-module, for monitoring the deviation of the user's heart rate and breathing pattern.

[0012] Preferably, the health trend analysis and personalized report module analyzes the long-term health trends of users based on a recurrent neural network (RNN), and its model is:

[0013]

[0014] where, is the hidden state, is the input health data, is the predicted health trend result.

[0015] Preferably, the health trend analysis and personalized report module analyzes the long-term health trends of users based on a recurrent neural network, and its model is:

[0016]

[0017] where, is the hidden state, is the input health data, is the predicted health trend result.

[0018] Preferably, the personalized health intervention simulation and optimization module optimizes the intervention strategy based on reinforcement learning, and its objective function is:

[0019] where, is the quality value of the current state and action of, is the immediate reward, is the discount factor.

[0020] Preferably, the dynamic intervention and long-term feedback module includes: An intervention strategy generation sub-module, which is used to generate intervention suggestions based on user classification; A user feedback collection sub-module, which is used to update the health profile according to the user's feedback on the intervention effect.

[0021] Preferably, the population segmentation and behavior pattern recognition module analyzes the user behavior pattern through a long short-term memory network LSTM, and its model is:

[0022] where, is the current hidden state, is the input behavior data, is the bias term.

[0023] Preferably, the personalized health report generated by the report interaction and data analysis module includes: User health trend chart; Time distribution of abnormal events; Health suggestions for user classification.

[0024] The present invention provides a full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar, having the following beneficial effects: 1. The present invention adopts a hardware acquisition and real-time signal processing module for collecting vital sign data through millimeter-wave radar and multi-modal sensors, and a technical solution combining user classification and dynamic update of health portraits, realizing high-precision and multi-dimensional acquisition and personalized management of user health data. Compared with the technical solutions in the prior art with insufficient monitoring accuracy and poor adaptability of single sensors, the present invention solves the problems of insufficient integrity and reliability of data collection in different scenarios, and effectively improves the comprehensiveness and stability of the health monitoring system.

[0025] 2. The present invention adopts a health trend analysis module based on recurrent neural network and a dynamic intervention and long-term feedback module optimized by reinforcement learning. Through time series modeling and adaptive optimization strategies, it realizes accurate prediction of users' long-term health trends and continuous optimization of personalized interventions. Compared with the technical solutions in the prior art that only rely on fixed thresholds to trigger interventions, the present invention solves the problems of slow response to changes in users' health status and inflexible intervention strategies, and significantly improves the scientificity of the health management plan and the user experience.

[0026] 3. The present invention adopts a population segmentation and dynamic classification technology combining user behavior pattern recognition, and a visualization and interactive health report generation module, realizing refined analysis of the health status of different populations and output of operable health management suggestions. Compared with the technical solutions in the prior art that lack in-depth exploration of user behavior patterns and classification characteristics, the present invention solves the problems of single content and insufficient guidance of health reports, improves the practicality of the reports and the user's understanding depth of health data, and provides users with more personalized health management services that meet their actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a system diagram of the present invention; Figure 2 It is a schematic diagram of the hardware acquisition and real-time signal processing module of the present invention; Figure 3 It is a schematic diagram of the population classification and feature modeling module of the present invention; Figure 4 It is a schematic diagram of the edge computing and real-time detection module of the present invention; Figure 5Schematic diagram of the dynamic intervention and long-term feedback module of the present invention. Detailed implementation manners

[0028] Next, in combination with the accompanying drawings of the present invention specification, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0029] Please refer to the attached Figure 1 -attached Figure 5 , the embodiment of the present invention provides a full-cycle electrocardiogram acquisition and diagnostic assistance system based on a millimeter-wave radar, including: A hardware acquisition and real-time signal processing module, which is used to collect the vital sign data of the user through a millimeter-wave radar and a multi-modal sensor, and preprocess the collected original signal; A population classification and feature modeling module, which classifies according to the information input by the user and the collected data, and constructs a personalized health portrait; An edge computing and real-time monitoring module, which is used to extract electrocardiogram signals and other vital sign data in real time and monitor abnormal events; A health trend analysis and personalized report module, which is used to perform trend analysis on long-term health data and generate a personalized health report for the user; A dynamic intervention and long-term feedback module, which generates health intervention suggestions according to the trend analysis results and optimizes the intervention strategy according to the user feedback; A population segmentation and behavior pattern recognition module, which optimizes personalized health management by further segmenting user categories and recognizing behavior patterns; A health anomaly prediction and personalized threshold optimization module, which is used to predict the health anomaly risk and dynamically adjust the monitoring threshold; A personalized health intervention simulation and optimization module, which is used to simulate and optimize health intervention strategies; A report interaction and data analysis module, which is used to generate and display multi-level health reports and support user interaction analysis.

[0030] Specifically, for the hardware acquisition and real-time signal processing module: this module collects the vital sign data of the user through millimeter-wave radar technology and multi-modal sensors (such as infrared sensors, sound sensors, etc.), including heart rate, respiratory signal, body surface temperature, motion state, etc., and performs denoising, artifact removal and signal standardization processing on the original signal to provide high-quality basic data for subsequent data analysis.

[0031] Population Classification and Feature Modeling Module: Based on the user's input information (such as age, gender, occupation, health status, etc.) and the collected vital sign data, the users are classified into different population categories (such as the elderly, pregnant women, athletes, etc.), and personalized health portraits are constructed according to the classification results. This module supports dynamic updating of the user's health portrait to continuously optimize it with the changes in time and new data.

[0032] Edge Computing and Real-Time Monitoring Module: This module extracts key vital sign parameters such as the user's electrocardiogram signal, respiratory rate, heart rate variability, etc. in real time through efficient edge computing, and at the same time detects abnormal events (such as arrhythmia, apnea, etc.). Edge computing not only reduces the burden of cloud analysis but also ensures rapid response to abnormal events.

[0033] Health Trend Analysis and Personalized Report Module: This module conducts trend analysis on the user's long-term health data, including changes in vital sign parameters, distribution of abnormal events, overall trends in health status, etc., and generates personalized health reports that contain information such as the user's health score, trend charts, abnormal analysis, etc., providing intuitive health insights for users and doctors.

[0034] Dynamic Intervention and Long-Term Feedback Module: According to the results of health trend analysis, this module generates targeted health intervention suggestions, such as improving lifestyle, adjusting exercise volume, or drug reminders. At the same time, this module continuously optimizes the intervention strategy through user feedback data to form a personalized closed-loop health management plan.

[0035] Population Subdivision and Behavior Pattern Recognition Module: On the basis of the basic population classification, the user categories are further subdivided (for example, the elderly are divided into those with limited mobility and those with normal mobility according to their activity ability), and combined with data from accelerometers, millimeter-wave radars, and other sensors, the user's behavior patterns (such as sitting still, walking, sleeping, etc.) are recognized, thereby optimizing personalized health management.

[0036] Health Abnormality Prediction and Personalized Threshold Optimization Module: Based on the analysis of long-term data, this module predicts the user's potential health risks (such as the likelihood of a heart attack), and dynamically adjusts the thresholds of monitoring parameters to adapt to the changes in the user's health status, thereby improving the sensitivity and accuracy of monitoring.

[0037] Personalized Health Intervention Simulation and Optimization Module: Before implementing health intervention measures, this module simulates the effects of different intervention strategies (such as environmental adjustment, exercise plan optimization, or drug management), and selects the optimal intervention plan. Through continuous learning from user feedback, the intervention decision-making is continuously optimized to make the intervention plan more accurate and effective.

[0038] Report Interaction and Data Analysis Module: Generate multi-level health reports suitable for different user needs, including core health data for basic users, detailed training analysis for athletes, diagnostic assistance reports for doctors, etc. The reports support interactive functions, allowing users to deeply analyze health trends, abnormal events, and key indicators to obtain customized health advice.

[0039] The hardware acquisition and real-time signal processing module includes: An acquisition sub-module based on FMCW millimeter-wave radar, used to acquire chest micro-vibration signals; An infrared sensor sub-module, used to acquire the user's body surface temperature; A sound sensor sub-module, used to acquire snoring signals and analyze the user's breathing pattern.

[0040] Specifically, the acquisition sub-module based on FMCW millimeter-wave radar: This sub-module uses frequency-modulated continuous-wave (FMCW) millimeter-wave radar technology. By transmitting microwave signals and receiving reflected signals, it accurately captures the user's chest micro-vibration information. The micro-vibration signal reflects the user's breathing frequency and heartbeat signal characteristics. The millimeter-wave radar has the characteristics of high sensitivity and non-contact, and can penetrate objects such as clothes and quilts for monitoring, which is especially suitable for long-term monitoring of users in daily life.

[0041] Infrared sensor sub-module: This sub-module is used to acquire the user's body surface temperature and calculate the change of the body surface temperature by measuring the intensity of infrared radiation emitted by the human body. The infrared sensor has the advantages of fast response and high precision, and can capture the user's body temperature fluctuation information in real time, providing important reference for monitoring health status (such as fever, low temperature, etc.).

[0042] Sound sensor sub-module: This sub-module is used to acquire the user's breathing and snoring signals and analyze the breathing pattern and its possible abnormalities (such as apnea, snoring intensity, etc.). The sound sensor combines signal processing algorithms to separate background noise and the user's breathing-related sounds, providing data support for identifying breathing disorders and sleep quality analysis.

[0043] The population classification and feature modeling module includes: A user classification sub-module, used to classify into the elderly, pregnant women, or athletes according to the information input by the user (including age, gender, occupation, and health status) and the acquired health data, and further refine into categories such as good mobility and limited mobility; A health portrait dynamic update sub-module, used to dynamically adjust the user's health portrait in combination with the real-time acquired data through the following formula:

[0044] Where, is the current health portrait status, For the latest collected data, is the update coefficient.

[0045] Specifically, the user classification sub-module: This sub-module preliminarily classifies users based on the basic information input by users (such as age, gender, occupation, and health status) and the real-time health data collected. The classification results include but are not limited to three major groups: the elderly, pregnant women, and athletes. At the same time, on this basis, the classification is further refined. For example, for the elderly group, it can be divided into two categories: good mobility and limited mobility according to the mobility ability; for pregnant women, it can be divided into early, middle, and late pregnancy stages according to the pregnancy stage; for athletes, it can be divided into endurance type, strength type, or mixed type according to the training type. This classification process dynamically combines the user input information and the collected health data to make the classification more accurate and lay a foundation for subsequent personalized health management.

[0046] The dynamic health profile update sub-module: This sub-module constructs a personalized health profile for users based on the classification results of users and combines the real-time collected health data (such as heart rate, breathing pattern, body temperature fluctuation, etc.). The health profile contains the core health characteristic parameters and dynamic change trends of users, and can comprehensively reflect the health status of users. At the same time, this module supports dynamic updates. As the continuous accumulation of users' long-term monitoring data and the input of real-time data, the health profile will be automatically adjusted based on the latest data. The update process ensures that the user profile can reflect their current health status and potential risk changes, providing a scientific basis for personalized health management and intervention suggestions.

[0047] The edge computing and real-time monitoring module includes: A signal separation sub-module, which is used to extract the breathing signal and heart rate signal through a band-pass filter, where the frequency range of the breathing signal is 0.1–0.4 Hz, and the frequency range of the heart rate signal is 0.8–2.5 Hz; A real-time anomaly detection sub-module, which is used to monitor the deviation of the user's heart rate and breathing pattern.

[0048] Specifically, the signal separation sub-module: This sub-module separates and processes the collected comprehensive vital sign signals by applying a band-pass filter, and extracts the breathing signal and heart rate signal from them. The breathing signal is mainly concentrated in the low-frequency range (0.1–0.4 Hz) and is used to reflect the user's breathing frequency and pattern; the heart rate signal is mainly distributed in the higher frequency band (0.8–2.5 Hz) and is used to reflect the user's heart rate. During the signal separation process, the band-pass filter can effectively remove irrelevant background noise and artifact signals, ensuring that the extracted vital sign signals have high signal-to-noise ratio and accuracy, and providing high-quality data input for subsequent monitoring and analysis.

[0049] Real-time anomaly detection sub-module: Based on the isolated respiratory signal and heart rate signal, this sub-module monitors the deviation of the user's vital signs in real time. By analyzing the frequency, amplitude, and change pattern of the signals, it identifies possible abnormal states of the user, such as significant changes in the breathing pattern (such as apnea or hyperventilation), abnormal fluctuations in heart rate (too high or too low), etc. Once an anomaly is detected, this sub-module can generate an immediate warning signal and transmit key data to the upper-level analysis module or the user device so that the user can take corresponding measures in time.

[0050] The health trend analysis and personalized report module analyzes the long-term health trend of the user based on a Recurrent Neural Network (RNN), and its model is:

[0051]

[0052] where is the hidden state, is the input health data, is the predicted health trend result.

[0053] The health trend analysis and personalized report module analyzes the long-term health trend of the user based on a Recurrent Neural Network, and its model is:

[0054]

[0055] where is the hidden state, is the input health data, is the predicted health trend result.

[0056] Specifically, for the long-term health trend analysis: The historical health data of the user is modeled and analyzed through a Recurrent Neural Network. The Recurrent Neural Network can effectively capture the time-varying correlation characteristics in the health data, thereby revealing the long-term trend of the health state. The input data includes key vital sign parameters such as the user's heart rate, breathing pattern, sleep quality, body temperature changes, etc., as well as the abnormal event records extracted from the anomaly detection module. The model generates the health state prediction for each time point by gradually processing the time series data, further reflecting the dynamic change trend of the user's health condition.

[0057] Trend evaluation and comparison: Based on the analysis of the user's own health trend, combined with the statistical average level of the health data of the same type of population (such as the same age group, gender, or classification category), a horizontal comparison of the user's health state is carried out. The evaluation results include the deviation degree of the health indicators, the proximity to the target state, etc., providing a basis for personalized intervention suggestions.

[0058] Personalized health report generation: Generate a personalized health report for the user based on the trend analysis results. The report content includes: A time trend graph of health indicators; The evaluation results of important vital sign parameters; The time distribution and occurrence frequency of abnormal events; Personalized health suggestions, such as lifestyle adjustments, optimized exercise plans, or medical advice. The report structure is intuitive and clear, facilitating user understanding and enabling doctors to refer to the user's long-term health data during the diagnosis process.

[0059] The personalized health intervention simulation and optimization module optimizes the intervention strategy based on reinforcement learning, and its objective function is:

[0060] where is the quality value of the current state and action respectively, is the immediate reward, is the discount factor.

[0061] Specifically, for health intervention strategy generation: The module generates a preliminary intervention strategy based on user classification and health portraits, combined with health trend analysis and risk assessment results. The intervention strategy may include lifestyle adjustment suggestions (such as increasing exercise and improving sleep), environmental optimization measures (such as adjusting indoor temperature and humidity), dietary advice (such as increasing specific nutrient intake), and specific medical reminders (such as taking medicine or seeing a doctor on time).

[0062] Intervention simulation and effect prediction: Before implementing the intervention, the module simulates the potential effects of different intervention strategies through reinforcement learning techniques. Each strategy is regarded as an "action", and the model evaluates its possible benefits (i.e., intervention effects) based on the user's current health status. This predictive ability can help the module screen out the most beneficial intervention plan for the user's health, avoiding potential side effects or ineffective operations caused by blind intervention.

[0063] Reinforcement learning to optimize the strategy: The module uses reinforcement learning techniques to continuously optimize the intervention decision. Evaluate the actual effect of the strategy through the user's real-time feedback (such as changes in health data after the intervention), and input the feedback results into the learning model to continuously update and improve the intervention strategy. Through repeated trials and adjustments, the reinforcement learning model makes the intervention decision gradually tend to be optimal.

[0064] Multi-dimensional health feedback input: The module collects users' feedback data from multiple dimensions, including the improvement of vital sign parameters (such as the heart rate returning to normal and the respiratory rate tending to be stable), users' subjective feelings (such as a decrease in fatigue level and an improvement in sleep quality), and the acceptability of interventions (such as the user's cooperation with the intervention content). This feedback information is used as training data for reinforcement learning to further improve the accuracy of policy optimization.

[0065] Implementation of personalized intervention plan: After completing the optimization of the intervention strategy, the module converts the optimal strategy into specific actionable suggestions and transmits them to the user through the user terminal (such as a smartphone, smartwatch, etc.). For example, for a user detected to be overly fatigued, the module may suggest reducing the amount of exercise and increasing the rest frequency; for a user found to have sleep disorders, it may suggest adjusting the sleep environment and performing specific relaxation training.

[0066] Continuous learning and dynamic adjustment: The module is designed to support continuous learning and dynamic update, and can adaptively adjust the intervention strategy as the user's health status changes. For example, when the user enters a new health stage (such as a pregnant woman entering the late pregnancy) or the health status improves significantly, the module will automatically update the intervention goals and strategy priorities.

[0067] The dynamic intervention and long-term feedback module includes: An intervention strategy generation sub-module for generating intervention suggestions based on user classification; A user feedback collection sub-module for updating the health profile according to the user's feedback on the intervention effect.

[0068] The population segmentation and behavior pattern recognition module analyzes the user's behavior pattern through a long short-term memory network (LSTM), and its model is:

[0069] where, is the current hidden state, is the input behavior data, is the bias term.

[0070] The personalized health report generated by the report interaction and data analysis module includes: User health trend charts; Time distribution of abnormal events; Health suggestions for user classification.

[0071] Specifically, the intervention strategy generation sub-module: This sub-module generates personalized intervention suggestions based on the user's classification (such as the elderly, pregnant women, athletes, etc.) and health profile, combined with the results of health trend analysis and anomaly detection. The intervention content may include lifestyle adjustments (such as optimizing the work and rest time), exercise program suggestions (such as adjusting the exercise intensity and frequency), diet suggestions (such as supplementing specific nutrients), and health environment optimization (such as improving the temperature and humidity conditions of the sleep environment). The intervention strategy is dynamically adjusted by analyzing the specific needs and health status of the user to ensure the practicality and scientific nature of the suggestions.

[0072] User feedback collection sub-module: This sub-module is used to collect the actual responses and effectiveness feedback of users to the intervention measures, including objective health data changes (such as the improvement of heart rate and respiratory rate) and the subjective feelings of users (such as the reduction of fatigue and the improvement of sleep quality). The feedback data will update the user's health profile in real time to ensure that the health management plan can be adaptively optimized as the user's status changes.

[0073] Population segmentation: Based on the user's input information and the collected health data, the classification is further refined. For example, for the athlete group, it can be subdivided into endurance type, strength type, and mixed type according to the type of exercise; for the elderly group, it can be divided into those with good mobility and those with limited mobility according to the mobility ability. The segmentation provides refined support for the personalized intervention strategy of health management.

[0074] Behavior pattern recognition: This module analyzes the collected time series data through the LSTM model to capture the dynamic changes of the user's behavior patterns. Behavior patterns include static behaviors (such as sleeping, sitting still), dynamic behaviors (such as walking, running), and special behaviors (such as fetal movement of pregnant women). The LSTM model can accurately identify the user's behavior categories based on the context association of the time series data and predict possible health risks according to the behavior change trend, providing decision-making support for subsequent interventions.

[0075] User health trend chart: It visually displays the time change trends of the user's key health indicators (such as heart rate, respiratory rate, heart rate variability, etc.). The trend chart intuitively presents the dynamic changes of the health status and helps users understand their long-term health conditions.

[0076] Time distribution of abnormal events: By reporting the time points and frequencies of the occurrence of marked abnormal events, including important health events such as apnea and abnormal heart rate, it provides key health risk warnings for users and doctors.

[0077] Health suggestions for user classification: According to the user's classification and health profile, personalized health suggestions are provided. For example, for athletes, the report may include suggestions for optimizing training intensity; for the elderly, it may provide a plan for improving the sleep environment; for pregnant women, it may prompt diet adjustments or rest arrangements.

[0078] Interactive analysis function: Users can interact with the report content through terminal devices. For example, they can click on the trend chart to view detailed data at specific time points or compare the health status in different time periods. The interactive function enhances users' understanding and application of health data and improves their participation in health management.

[0079] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A full-cycle electrocardiogram acquisition and diagnosis assistance system based on a millimeter-wave radar, characterized in that Including: A hardware acquisition and real-time signal processing module, which is used to collect the user's vital sign data through a millimeter-wave radar and multi-modal sensors, and preprocess the collected original signals; A crowd classification and feature modeling module, which classifies according to the information input by the user and the collected data, and constructs a personalized health profile; An edge computing and real-time monitoring module, which is used to extract electrocardiogram signals and other vital sign data in real time and monitor abnormal events; A health trend analysis and personalized report module, which is used to perform trend analysis on long-term health data and generate a personalized health report for the user; A dynamic intervention and long-term feedback module, which generates health intervention suggestions according to the trend analysis results and optimizes the intervention strategy according to the user's feedback; A crowd segmentation and behavior pattern recognition module, which optimizes personalized health management by further segmenting user categories and recognizing behavior patterns; A health anomaly prediction and personalized threshold optimization module, which is used to predict the health anomaly risk and dynamically adjust the monitoring threshold; A personalized health intervention simulation and optimization module, which is used to simulate and optimize health intervention strategies; A report interaction and data analysis module, which is used to generate and display multi-level health reports and support user interaction analysis.

2. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The hardware acquisition and real-time signal processing module includes: An acquisition sub-module based on an FMCW millimeter-wave radar, which is used to collect chest micro-vibration signals; An infrared sensor sub-module, which is used to collect the user's body surface temperature; A sound sensor sub-module, which is used to collect snoring signals and analyze the user's breathing pattern.

3. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, characterized in that, The crowd classification and feature modeling module includes: A user classification sub-module, which classifies into the elderly, pregnant women or athletes according to the information input by the user (including age, gender, occupation and health status) and the collected health data, and is further refined into categories such as good mobility and limited mobility; A health profile dynamic update sub-module, which is used to dynamically adjust the user's health profile by combining the real-time collected data through the following formula: Among them, is the current health portrait status, is the latest collected data, is the update coefficient.

4. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The edge computing and real-time monitoring module includes: A signal separation sub-module, which is used to extract respiratory signals and heart rate signals through a band-pass filter, where the frequency range of the respiratory signal is 0.1–0.4Hz and the frequency range of the heart rate signal is 0.8–2.5Hz; A real-time anomaly detection sub-module, which is used to monitor the deviation of the user's heart rate and breathing pattern.

5. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, characterized in that, The health trend analysis and personalized report module analyzes the user's long-term health trend based on a recurrent neural network (RNN), and its model is: Among them, is in a hidden state, is the input health data, is the result of predicting the health trend.

6. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The health trend analysis and personalized report module analyzes the user's long-term health trend based on a recurrent neural network, and its model is: Among them, is in a hidden state, is the input health data, is the result of predicting the health trend.

7. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The personalized health intervention simulation and optimization module optimizes the intervention strategy based on reinforcement learning, and its objective function is: Among them, is the current state and the action 's quality value, is the immediate reward, is the discount factor.

8. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The dynamic intervention and long-term feedback module includes: An intervention strategy generation sub-module, which is used to generate intervention suggestions based on user classification; A user feedback collection sub-module, which is used to update the health profile according to the user's feedback on the intervention effect.

9. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, characterized in that The crowd segmentation and behavior pattern recognition module analyzes the user's behavior pattern through a long short-term memory network LSTM, and its model is: Among them, is the current hidden state, is the input behavior data, is the bias term.

10. The full-cycle electrocardiogram acquisition and diagnosis assistance system based on millimeter-wave radar according to claim 1, wherein The personalized health report generated by the report interaction and data analysis module includes: User health trend charts; Abnormal event time distribution; Health advice for user classification.

Images