Multi-dimensional data fusion interaction processing system of multi-modal intelligent health monitoring terminal

Abstract

The invention relates to the technical field of intelligent health monitoring, and discloses a multi-dimensional data fusion interaction processing system of a multi-modal intelligent health monitoring terminal. The system comprises a multi-dimensional data acquisition module, an intelligent fusion processing module and an interactive response control module. The multi-dimensional data acquisition module forms multi-modal original data through the data fusion interface unit; after receiving the data, the intelligent fusion processing module processes and outputs a feature fusion signal and an abnormal risk early warning signal through a data fusion engine, a health state evaluation unit obtains an evaluation result by combining historical health data of a user, and a parameter optimization unit outputs a personalized health guidance threshold according to the evaluation result; in the interactive response control module, an adaptive feedback unit adjusts intervention parameters, a terminal cooperative control unit adjusts working parameters of a monitoring terminal and collects feedback signals, and a human-computer interaction interface module outputs a comprehensive health risk level. According to the system, effective fusion and personalized interaction response of multi-dimensional data are realized, and the comprehensiveness and applicability of health monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent health monitoring technology, specifically to a multi-dimensional data fusion and interactive processing system for a multimodal intelligent health monitoring terminal. Background Technology

[0002] With the accelerating aging of the population and the rising incidence of chronic diseases, the demand for real-time and comprehensive monitoring of individual health status is increasing. Currently, most health monitoring terminals on the market are single-function devices that can only collect one type of physiological parameter. For example, heart rate monitoring bracelets can only acquire heart rate data, and blood pressure monitors can only measure blood pressure values, making it difficult to achieve a multi-dimensional assessment of human health status. Even though some devices attempt to integrate multiple monitoring functions, their data processing methods still have significant limitations. These devices typically simply overlay or process different types of physiological data separately, lacking an effective fusion mechanism. This leads to the neglect of correlations between data points, making it impossible to form a holistic understanding of health status. For example, when monitoring diabetic patients, focusing solely on blood glucose data while ignoring related data such as exercise and diet may fail to accurately determine the true cause of blood glucose fluctuations. The interactive response mechanisms of existing monitoring systems are rather rigid. Most devices can only issue warnings based on preset fixed thresholds, failing to dynamically adjust according to individual user differences and real-time health status. Applying the same warning standards to users of different ages and physical conditions easily leads to false alarms or missed alarms. Furthermore, the human-computer interaction interface mainly focuses on data listing, lacking in-depth interpretation of health information and personalized guidance. Ordinary users find it difficult to understand the health implications of complex monitoring data, reducing the practical value of the monitoring system. The existing system has shortcomings in data feedback and terminal collaboration. There is a lack of effective linkage mechanisms between monitoring terminals and intervention devices. When abnormal data is detected, intervention measures cannot be adjusted in a timely manner, making it difficult to achieve a closed-loop health management system. For example, when a user's blood pressure is detected to be elevated, the system cannot automatically adjust the parameters of relevant intervention devices to alleviate the abnormal blood pressure, affecting the timeliness and effectiveness of health management. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-dimensional data fusion and interactive processing system for a multimodal intelligent health monitoring terminal, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a multi-dimensional data fusion and interactive processing system for a multimodal intelligent health monitoring terminal, the system comprising: A multidimensional data acquisition module, comprising a data fusion interface unit, wherein the data fusion interface unit generates multimodal raw data; The intelligent fusion processing module receives the multimodal raw data and includes a data fusion engine, a health status assessment unit, and a parameter optimization unit. The data fusion engine processes the multimodal raw data and outputs a feature fusion signal and an abnormal risk warning signal. The health status assessment unit processes the feature fusion signal and the user's historical health data to obtain a health status assessment result. The parameter optimization unit defines intervention constraints based on the health status assessment result and outputs personalized health guidance thresholds. The interactive response control module includes an adaptive feedback unit, a terminal collaborative control unit, and a human-computer interaction interface module. The adaptive feedback unit adjusts intervention parameters according to the personalized health guidance threshold. The terminal collaborative control unit adjusts the monitoring terminal's operating parameters according to the health status assessment results and the multimodal raw data, and collects real-time physiological feedback signals. The human-computer interaction interface module processes the abnormal risk warning signals and the health status assessment results and outputs a comprehensive health risk level.

[0005] Preferably, the multidimensional data acquisition module includes a wearable biosensing unit, an environmental parameter detection unit, and a motion posture capture unit. The multidimensional data acquisition module is deployed on key parts of the user's body, clothing, and living environment, and collects raw sensor signals. The motion posture capture unit acquires the user's limb movement trajectory data through an inertial measurement unit and forms a motion state signal. The data fusion interface unit receives the raw sensor signals and the motion state signal to integrate multi-source heterogeneous data and form the multimodal raw data. The multimodal raw data includes physiological parameter data, environmental influence data, and motion characteristic data.

[0006] Preferably, the data fusion engine includes a feature extraction model based on a convolutional neural network and a Bayesian network fusion model. The physiological parameter data includes heart rate variability data, skin resistance change data, blood oxygen saturation fluctuation data, body temperature gradient distribution data, and respiratory rate cycle data. The heart rate variability data, skin resistance change data, blood oxygen saturation fluctuation data, body temperature gradient distribution data, and respiratory rate cycle data are input into the feature extraction model and output the feature fusion signal. The physiological parameter data, environmental influence data, and motion feature data are fused through the improved Bayesian network fusion model and the abnormal risk warning signal is output.

[0007] Preferably, the feature extraction model includes a multi-scale feature decomposition submodule, a spatiotemporal correlation analysis submodule, a risk feature quantification submodule, and a fusion result visualization submodule. The multi-scale feature decomposition submodule performs empirical mode decomposition on the heart rate variability data and extracts the fluctuation features of frequency bands that meet the pre-abnormal physiological state frequency band range as health risk precursor signals. The spatiotemporal correlation analysis submodule uses an attention-weighted convolutional neural network to process the skin resistance change data, the blood oxygen saturation fluctuation data, and the motion feature data and outputs the data to the risk feature quantification submodule. The risk feature quantification submodule defines risk level calculation rules and generates the feature fusion signal by comprehensively evaluating the degree of deviation of each physiological parameter from the normal range, the influence weight of environmental parameters, and the matching degree of the motion state. The fusion result visualization submodule generates a dynamic health status heatmap of the feature fusion signal based on a four-color scale, using different color scales to represent the health risk level of different body parts.

[0008] Preferably, the health status assessment unit processes the environmental impact data and the feature fusion signal using a health-intervention nonlinear relationship model, outputs a preliminary health assessment result, and uses a support vector machine algorithm to classify and optimize the preliminary health assessment result to obtain a final health status assessment result. The environmental impact data includes ambient temperature change signals, air quality index signals, and electromagnetic radiation intensity signals. The health-intervention nonlinear relationship model is constructed using the ambient temperature change signals, the feature fusion signal, and the electromagnetic radiation intensity signal. This model analyzes the correlation between environmental parameters and abnormal physiological indicators, and, combined with the environmental adaptation patterns in the user's historical health data, establishes a quantitative assessment framework for the impact of environmental factors on health status.

[0009] Preferably, the health status assessment unit further includes a health indicator database, an abnormal pattern recognition submodule, and an intervention priority ranking submodule. The health indicator database stores standard physiological parameter ranges for different age groups, genders, and health statuses. The abnormal pattern recognition submodule identifies abnormal change patterns of physiological indicators based on the feature fusion signal and the environmental impact data. The intervention priority ranking submodule obtains the health status assessment result through a decision tree algorithm based on the abnormal change patterns, the health-intervention nonlinear relationship model, and the air quality index signal.

[0010] Preferably, the parameter optimization unit defines intervention constraints based on the exercise feature data, outputs a personalized health guidance threshold, and dynamically adjusts the personalized health guidance threshold to ensure it is not lower than the health and safety benchmark value through fuzzy PID control. The exercise feature data includes exercise intensity signal, exercise duration signal, and exercise type signal. The exercise intensity signal, exercise duration signal, and exercise type signal define the intervention constraints. By comprehensively considering the degree of influence of exercise on physiological indicators, the user's current health status, and historical intervention effects, the range of health guidance parameters under different exercise scenarios is determined. The fuzzy PID control dynamically adjusts the intensity and frequency of intervention measures to ensure that the personalized health guidance threshold is not lower than the health and safety benchmark value.

[0011] Preferably, the human-computer interaction interface module includes a three-dimensional dynamic display platform and an early warning engine. The three-dimensional dynamic display platform receives the multimodal raw data and the real-time physiological feedback signal and dynamically renders the physiological indicator change process. The early warning engine processes the abnormal risk warning signal and the health status assessment result according to the improved analytic hierarchy process and outputs a comprehensive health risk level. When the comprehensive health risk level exceeds the risk level threshold, the human-computer interaction interface module controls the adaptive feedback unit to initiate intervention measures. The three-dimensional dynamic display platform supports multi-dimensional physiological data cross-sectional analysis, can display the correlation between key physiological indicators and intervention measures in real time, and simulate the physiological state change trend under different intervention schemes based on multibody dynamics.

[0012] Preferably, the early warning engine includes a three-level warning mechanism with color states: green for safety, yellow for attention, and red for emergency. When a key physiological indicator in the health status assessment deviates from the normal range to a first deviation threshold, the yellow attention state is activated, health adjustment suggestions are pushed, and the recommended exercise intensity is reduced. When a key physiological indicator in the health status assessment deviates from the normal range to a second deviation threshold, the high-intensity exercise recommendation is suspended, and the adaptive feedback unit is controlled to initiate mild intervention measures. When a key physiological indicator in the health status assessment deviates from the normal range to a third deviation threshold, the early warning engine issues an emergency health warning and controls the multimodal intelligent health monitoring terminal to automatically contact emergency contacts.

[0013] Preferably, the personalized health guidance threshold includes parameters such as intervention type, intervention intensity range, intervention frequency recommendation, and duration. The adaptive feedback unit includes an intelligent adjustment device and an integrated physiological parameter real-time monitoring component. The intelligent adjustment device can identify the personalized health guidance threshold and adjust the implementation method and intensity of the intervention as needed. The integrated physiological parameter real-time monitoring component receives the personalized health guidance threshold and monitors changes in physiological indicators during the intervention process in real time based on the photoplethysmography method. The real-time physiological feedback signal includes heart rate change rate, skin temperature fluctuation, blood oxygen saturation change, movement posture adjustment amplitude, and intervention response time.

[0014] Compared with the prior art, the beneficial effects of the present invention are: By setting up a multi-dimensional data acquisition module and leveraging a data fusion interface unit to generate multimodal raw data, the traditional single-data acquisition mode of monitoring equipment has been changed. This allows for the simultaneous acquisition of multiple physiological parameters, covering various aspects of human health status. This multi-dimensional data provides a rich source of information for a comprehensive understanding of the user's health status, enabling health monitoring to move beyond a single physiological indicator and capture changes in the body from a holistic perspective. The data fusion engine in the intelligent fusion processing module processes multimodal raw data, outputting feature fusion signals and abnormal risk warning signals, breaking down the barriers of isolated processing of different types of data. By mining the inherent connections between data, potential health risks can be identified more accurately, avoiding the one-sidedness that may result from interpreting single data points. The health status assessment unit combines feature fusion signals with users' historical health data, making the health status assessment results more continuous and targeted, reflecting the dynamic changes in users' health status rather than isolated instantaneous states. The parameter optimization unit defines intervention constraints based on the assessment results, outputting personalized health guidance thresholds, so that health guidance no longer relies on uniform standards but fully considers individual differences, making the guidance content more tailored to users' actual needs. The adaptive feedback unit of the interactive response control module adjusts intervention parameters based on personalized health guidance thresholds, achieving dynamic adaptation of intervention measures. As the user's health status changes, the intervention parameters can be adjusted in a timely manner, making the intervention more aligned with real-time health needs. The terminal collaborative control unit adjusts the monitoring terminal's operating parameters based on health status assessment results and multimodal raw data, and collects real-time physiological feedback signals, promoting dynamic interaction between the monitoring terminal and the user. This collaborative mechanism allows the monitoring terminal to optimize its operating mode according to actual conditions, while continuously improving its judgment of health status through real-time feedback signals. The human-computer interaction interface module processes abnormal risk warning signals and health status assessment results and outputs a comprehensive health risk level, transforming complex health data into intuitive and easy-to-understand level information, facilitating users' quick understanding of their own health status and enhancing their sense of participation and understanding in health management. The entire system forms a complete process from data acquisition and fusion processing to interactive response. The modules work closely together to achieve intelligent and personalized health monitoring. The comprehensiveness of data acquisition, the fusion of processing, and the adaptability of response combine to upgrade health monitoring from simple data recording to proactive health management, enhancing the practical application value of health monitoring and providing users with a health management experience that is more tailored to their individual needs. Attached Figure Description

[0015] Figure 1 This is a timing diagram of the multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal described in this invention; Figure 2 A flowchart illustrating the operation of the multidimensional data acquisition module; Figure 3 A flowchart illustrating the workings of the feature extraction model; Figure 4 A flowchart illustrating the operation of the parameter optimization unit; Figure 5 This is a flowchart illustrating how the human-computer interaction interface module works. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 This invention provides a multi-dimensional data fusion and interactive processing system for a multimodal intelligent health monitoring terminal, the system comprising: The multi-dimensional data acquisition module includes a data fusion interface unit, which integrates various types of received data to form multimodal raw data. The intelligent fusion processing module receives the multimodal raw data; its internal data fusion engine processes the data and outputs feature fusion signals and abnormal risk warning signals. The health status assessment unit combines the feature fusion signals with the user's historical health data to obtain a health status assessment result. The parameter optimization unit defines intervention constraints based on the health status assessment result and outputs personalized health guidance thresholds. In the interactive response control module, the adaptive feedback unit adjusts intervention parameters based on the personalized health guidance thresholds; the terminal collaborative control unit adjusts the monitoring terminal's operating parameters based on the health status assessment result and the multimodal raw data, and collects real-time physiological feedback signals; the human-computer interaction interface module processes the abnormal risk warning signals and health status assessment results, and outputs a comprehensive health risk level.

[0018] Example 1: See Figure 2 The multi-dimensional data acquisition module consists of a wearable biosensor unit, an environmental parameter detection unit, a motion posture capture unit, and a data fusion interface unit. The wearable biosensor unit, made of flexible electronic materials, can conform to the user's skin surface and be placed on key body parts such as the wrist, chest, and ankle, as well as in clothing and portable devices such as underwear, bracelets, and watches. It collects raw sensor signals through built-in biosensors. These sensors include photoelectric sensors, temperature sensors, and impedance sensors, which are used to capture various signals related to human physiological activities, such as changes in light intensity caused by blood flow, differences in body surface temperature, and changes in skin conductivity. The environmental parameter detection unit is distributed in the user's living environment, such as bedrooms, offices, and cars. Through temperature and humidity sensors, gas sensors, and electromagnetic sensors, it collects raw sensor signals reflecting the surrounding environmental conditions, including ambient temperature, humidity, concentration of particulate matter in the air, content of harmful gases, and intensity of electromagnetic radiation. The motion capture unit is integrated into the user's sports shoes, belt, or smart clothing. Its core is the inertial measurement unit, which consists of an accelerometer, a gyroscope, and a magnetometer. It can acquire the user's limb motion trajectory data in three-dimensional space in real time, including displacement, velocity, acceleration, and rotation angle, and process this data to form motion state signals to reflect different motion states such as walking, running, jumping, and sitting.

[0019] As the core integration component of the multi-dimensional data acquisition module, the data fusion interface unit connects with the wearable biosensor unit, environmental parameter detection unit, and motion posture capture unit via wired or wireless communication, receiving the raw sensor signals and motion state signals transmitted from them. Since these data come from different sources and have different formats, dimensions, and acquisition frequencies, they belong to multi-source heterogeneous data. The data fusion interface unit adopts a unified data format conversion protocol to convert various signals into compatible data forms. Then, through timestamp alignment technology, it matches different data within the same time period, achieving the integration of multi-source heterogeneous data and ultimately forming multimodal raw data. This multimodal raw data includes three main categories: physiological parameter data, which is information related to human physiology obtained by the wearable biosensor unit after collection and processing; environmental impact data, which is information reflecting the potential impact of the environment on the human body, collected and processed by the environmental parameter detection unit; and motion characteristic data, which is information reflecting the user's motion state generated by the motion posture capture unit.

[0020] The data fusion engine in the intelligent fusion processing module includes a feature extraction model based on convolutional neural networks and a Bayesian network fusion model. Physiological parameter data specifically covers heart rate variability data, skin resistance variation data, blood oxygen saturation fluctuation data, body temperature gradient distribution data, and respiratory rate cycle data. Heart rate variability data is obtained by analyzing the temporal changes in heartbeat intervals, reflecting the activity state of the autonomic nervous system; skin resistance variation data is collected by skin resistance sensors and is related to emotional fluctuations and sweating; blood oxygen saturation fluctuation data is obtained by monitoring changes in the ratio of oxygen-bound hemoglobin in the blood using photoelectric sensors; body temperature gradient distribution data is obtained by collecting temperatures from different parts of the body using multiple temperature sensors and calculating the temperature difference distribution; respiratory rate cycle data is obtained by monitoring the rise and fall of the chest or abdomen, or analyzing changes in airflow during respiration.

[0021] These physiological parameter data are input into a feature extraction model based on a convolutional neural network, which contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers perform sliding convolution operations on the input data using convolutional kernels of different sizes to extract local features, such as periodic fluctuations in heart rate variability data and trend changes in blood oxygen saturation fluctuation data. The pooling layers reduce the dimensionality of the features output by the convolutional layers, preserving key information while reducing the amount of data. The fully connected layers then integrate the features processed by convolution and pooling to form higher-dimensional abstract features, ultimately outputting a feature fusion signal that integrates key feature information from various physiological parameter data.

[0022] The improved Bayesian network fusion model is responsible for fusing physiological parameter data, environmental impact data, and motion characteristic data. This model pre-constructs a network structure containing multiple nodes, each node representing a data type or a feature variable, and the connections between nodes represent the probabilistic dependencies between variables. During the fusion process, the model first converts the physiological parameter data, environmental impact data, and motion characteristic data into corresponding probability distributions. Then, using Bayes' theorem, based on the existing data and the conditional probability table in the network structure, it calculates the joint probability distribution under different data combinations, thereby identifying potential abnormal patterns and ultimately outputting an abnormal risk warning signal to indicate potential health risks.

[0023] Example 2: See Figure 3 The feature extraction model consists of a multi-scale feature decomposition submodule, a spatiotemporal correlation analysis submodule, a risk feature quantification submodule, and a fusion result visualization submodule. The multi-scale feature decomposition submodule processes heart rate variability data, employing empirical mode decomposition (EMD) to decompose the data into multiple intrinsic mode functions (IMFs) of different frequencies. Through spectral analysis of these IMFs, components with frequencies falling within the pre-abnormal physiological state frequency range are selected. The fluctuation characteristics of these components, including fluctuation amplitude, periodic changes, and peak frequency, are extracted and integrated as a precursor signal of health risk.

[0024] The spatiotemporal correlation analysis submodule employs an attention-weighted convolutional neural network, which includes an input layer, an attention layer, convolutional layers, pooling layers, and an output layer. After skin resistance variation data, blood oxygen saturation fluctuation data, and motion feature data are input into the network, the attention layer automatically assigns weights based on the importance of the data to health status assessment. For example, when skin resistance variation is strongly correlated with motion status, the corresponding data will have a higher weight. The convolutional layers perform convolution operations on the weighted data using kernels of different sizes to extract local spatiotemporal features, such as the changing trend of skin resistance during movement and the fluctuation pattern of blood oxygen saturation over time. The pooling layer reduces the dimensionality of the convolutional features, retaining key information, and finally outputs the processed data to the risk feature quantification submodule.

[0025] The risk characteristic quantification submodule has pre-set risk level calculation rules that cover multiple assessment dimensions. First, it assesses the degree to which each physiological parameter deviates from the normal range; for example, it compares heart rate variability data with the standard range and calculates the proportion of the deviation relative to the standard range. Second, it analyzes the influence weight of environmental parameters; for example, the weight of body temperature gradient distribution data will be adjusted accordingly in high-temperature environments. Simultaneously, it considers the matching degree of exercise state, i.e., how well the current exercise type and intensity match the user's regular exercise state. By comprehensively evaluating the results of these three dimensions, it performs quantification according to the pre-set calculation rules to generate a feature fusion signal. This signal comprehensively reflects the risk level of each feature in numerical form.

[0026] The visualization submodule for fusion results generates a dynamic health status heatmap based on a four-color scale. Each color corresponds to a different risk level: dark blue for low risk, light blue for low to medium risk, orange for medium to high risk, and red for high risk. This submodule maps the feature fusion signals to different parts of the human body model, determining the color of the corresponding part based on the signal values. Simultaneously, the heatmap is updated in real-time with new data input, dynamically displaying changes in the health risk level of various body parts. Users can intuitively understand the health status of different areas of their body through the heatmap.

[0027] The health status assessment unit processes environmental impact data and feature fusion signals using a health-intervention nonlinear relationship model. Environmental temperature changes, electromagnetic radiation intensity, and feature fusion signals from the environmental impact data serve as input variables for the model. The model incorporates multiple nonlinear functions to describe the complex relationships between environmental parameters and abnormal physiological indicators, such as the nonlinear relationship between a sudden rise in environmental temperature and abnormal body temperature gradient distribution, and the nonlinear mapping between electromagnetic radiation intensity and heart rate variability. Through the calculation of these functions, a preliminary health assessment result is output, which includes assessment scores for multiple health indicators.

[0028] The Support Vector Machine (SVM) algorithm is used to classify and optimize the preliminary health assessment results. The algorithm finds the optimal classification hyperplane to divide the preliminary assessment results into different health status categories, such as healthy, mildly abnormal, and moderately abnormal. During the classification process, the algorithm incorporates classification features from the user's historical health data to adjust the hyperplane, improving classification accuracy and ultimately obtaining the health status assessment result. This result clearly defines the user's current health category and the specific status of each indicator.

[0029] In constructing the nonlinear relationship model between health and intervention, a large amount of data on the correlation between environmental parameters and abnormal physiological indicators is first collected, including changes in physiological indicators under different environmental temperatures and electromagnetic radiation intensities. By analyzing this data, the correlation patterns between environmental parameters and abnormal physiological indicators are determined; for example, the probability of abnormal body temperature gradient distribution increases with each increase in temperature. Simultaneously, environmental adaptation patterns are extracted from users' historical health data, such as the differences between the trends of physiological indicator changes in users under high-temperature environments and the general population. These patterns and differences are integrated into the model to establish a quantitative assessment framework for the impact of environmental factors on health status, enabling the model to output assessment results that conform to actual conditions based on specific environmental parameters and individual user characteristics.

[0030] Example 3: See Figure 4 The health status assessment unit comprises a health indicator database, an abnormal pattern recognition submodule, and an intervention priority ranking submodule. The health indicator database stores standard physiological parameter ranges categorized by age group, such as heart rate variability intervals and normal blood oxygen saturation ranges for children, adolescents, adults, and the elderly. It also records differences between men and women in areas such as skin resistance changes and body temperature gradient distribution, categorized by gender. Furthermore, it includes parameter references for different health conditions, including physiological parameter benchmarks for healthy individuals, patients with chronic diseases, and post-operative recovery groups. This data is stored in structured tabular form, supporting rapid querying and comparison.

[0031] After receiving the feature fusion signal and environmental impact data, the abnormal pattern recognition submodule first analyzes the feature fusion signal to extract deviations in various physiological indicators, such as the amplitude of heart rate variability fluctuations and the frequency of skin resistance changes. Simultaneously, it analyzes the ambient temperature change signal, air quality index signal, and electromagnetic radiation intensity signal in the environmental impact data to determine whether the current environment is undergoing drastic changes, such as a rapid rise or fall in temperature or a sudden exceedance of the air quality index. The submodule then performs correlation analysis between the deviations in physiological indicators and the environmental change state to identify any abnormal change patterns, such as abnormal fluctuations in heart rate variability when electromagnetic radiation intensity is high, or a significant decrease in blood oxygen saturation when air quality is poor.

[0032] The intervention prioritization submodule takes abnormal change patterns, a health-intervention nonlinear relationship model, and air quality index signals as inputs and processes them using a decision tree algorithm. The root node of the decision tree represents the severity of the abnormal change pattern, such as slight deviation, moderate deviation, and severe deviation; the first-level branches represent the air quality index signal levels, such as excellent, lightly polluted, moderately polluted, and heavily polluted; and the second-level branches reference the environmental impact weights output from the health-intervention nonlinear relationship model. The algorithm traverses each branch of the decision tree, calculates the implementation priority scores for different intervention measures, and ultimately obtains a health status assessment result, which clarifies the treatment order for various potential health problems.

[0033] The parameter optimization unit defines intervention constraints based on exercise characteristic data. Exercise intensity signals are represented by metabolic equivalent values, exercise duration signals are recorded in minutes, and exercise type signals are categorized into aerobic exercise, anaerobic exercise, flexibility exercise, etc. These signals collectively form the basis of the intervention constraints. For example, when the exercise intensity signal indicates high intensity, the constraints will limit the intensity of the intervention measures; when the exercise type is aerobic exercise, the constraints will focus on intervention parameters related to cardiopulmonary function.

[0034] The personalized health guidance thresholds output by the parameter optimization unit cover multiple dimensions. They comprehensively consider the impact of exercise on physiological indicators, such as the changes in heart rate and blood pressure at different exercise intensities; the user's current health status (abnormalities in the health status assessment); and historical intervention effects (records of the effectiveness of interventions implemented in similar health states in the past). This allows for the determination of health guidance parameter ranges for different exercise scenarios. For example, a lower intervention intensity range is set for high-intensity exercise scenarios, while a longer intervention duration parameter is set for light exercise scenarios for users in the recovery phase.

[0035] The parameter optimization unit dynamically adjusts the personalized health guidance threshold using fuzzy PID control to ensure it does not fall below the health and safety benchmark value. Fuzzy PID control combines fuzzy logic with the PID control algorithm, taking the deviation and rate of change between the health guidance threshold and the health and safety benchmark value as input. Fuzzy inference is performed through a fuzzy rule base to obtain the adjustment amounts of the proportional, integral, and derivative coefficients of the PID controller, thereby achieving dynamic adjustment of the intensity and frequency of intervention measures. The adjustment formula is: ; in, Let k be the control output at time k. This is the proportionality coefficient. The integral coefficient is... The differential coefficients are... Let $k$ be the deviation value at time $k$. The sampling period is defined as follows: When the personalized health guidance threshold approaches the health and safety benchmark, the adjustment sensitivity is increased by increasing the proportional coefficient; when there is a persistent deviation, the adjustment amount is accumulated through the integral coefficient; when the deviation changes drastically, overshoot is suppressed through the differential coefficient, ensuring that the personalized health guidance threshold remains above the health and safety benchmark.

[0036] Example 4: See Figure 5 The human-computer interaction interface module includes a 3D dynamic display platform and an early warning engine. After receiving multimodal raw data and real-time physiological feedback signals, the 3D dynamic display platform constructs a virtual human body model using built-in 3D modeling technology. Each part of the model corresponds to an actual human organ and system. Physiological parameters, environmental influence data, and motion characteristic data from the multimodal raw data are mapped to corresponding locations on the model. For example, heart rate variability data is associated with the heart region, environmental temperature change signals are associated with the entire body surface, and motion posture data drives the model to simulate limb movements. When real-time physiological feedback signals are received, such as an increase in heart rate variability or increased skin temperature fluctuations, the platform updates the model's display status using dynamic rendering technology. It uses a dynamic red curve to mark heart rate changes and color variations to show differences in skin temperature distribution, allowing users to intuitively see real-time changes in physiological indicators.

[0037] The 3D dynamic display platform supports multi-dimensional physiological data cross-sectional analysis. Users can zoom or swipe to view the distribution of physiological data at different angles and depths, such as the distribution of blood oxygen saturation in the thoracic cavity or changes in physiological parameters related to knee joint movement posture. The platform can display the correlation between key physiological indicators and intervention measures in real time. For example, when the adaptive feedback unit adjusts the intervention intensity, the platform will simultaneously display the change curves of indicators such as heart rate and respiratory rate under the new intervention intensity, and indicate the degree of correlation between the two. At the same time, based on the principle of multibody dynamics, the platform can simulate the trend of physiological state changes under different intervention programs, such as simulating the possible changes in body temperature gradient distribution after reducing daily exercise time, or the expected trend of skin resistance changes after increasing water intake.

[0038] The early warning engine employs an improved analytic hierarchy process (AHP) to process abnormal risk warning signals and health status assessment results. This method first decomposes abnormal risk warning signals into physiological abnormal signals, environmental abnormal signals, and motion abnormal signals, and decomposes health status assessment results into assessment levels for various health indicators. Then, it assigns corresponding weights based on the degree of influence of different signals and levels. The engine calculates a comprehensive health risk level through weighted calculations, dividing the level into multiple tiers from low to high. When the comprehensive health risk level exceeds a set risk level threshold, the human-computer interaction interface module sends a command to the adaptive feedback unit to initiate corresponding intervention measures, such as adjusting the tightness of wearable devices or changing the sampling frequency of environmental monitoring devices.

[0039] The warning engine's three-level warning mechanism corresponds to different color states and processing methods: A green "safe" state indicates all indicators are within the normal range, and the interface only displays routine health information. When key physiological indicators in the health assessment deviate from the normal range to the first deviation threshold (slight deviation), a yellow "attention" state is activated. The interface highlights the abnormal indicators in yellow and pushes health adjustment suggestions such as "increase rest time" and "avoid prolonged exposure to the current environment." Simultaneously, high-intensity exercises in the exercise recommendation list are replaced with low-to-medium intensity exercises. When key physiological indicators deviate from the normal range to the second deviation threshold (moderate deviation), all high-intensity exercise recommendations are paused. A pop-up window explains the abnormal situation, and the adaptive feedback unit initiates mild intervention measures, such as using wearable devices to emit intermittent vibrations to remind the user to adjust their posture or activating environmental control devices to improve the local microenvironment. When a key physiological indicator deviates from the normal range to the third deviation threshold, i.e., seriously deviates from the normal range, the warning engine triggers a red emergency state. The interface continuously flashes a red alarm icon and emits a warning sound. At the same time, it controls the multimodal intelligent health monitoring terminal to automatically dial the preset emergency contact number and send information including the current location, abnormal physiological indicator data, and comprehensive health risk level.

[0040] Example 5: Personalized health guidance thresholds include intervention type, intervention intensity range, recommended intervention frequency, and duration parameters. Intervention type is determined based on abnormalities in the health status assessment results, covering categories such as dietary adjustments, exercise guidance, and environmental improvements. For example, for abnormal heart rate variability, deep breathing regulation might be recommended; for abnormal body temperature gradient distribution, clothing adjustment might be recommended. Intervention intensity range is defined by specific values, such as the upper limit of daily salt intake in dietary adjustments or the walking speed range in exercise guidance. Intervention frequency recommendations are given on a daily or weekly basis, such as recommending 3 deep breathing exercises daily or 2 environmental parameter checks weekly. The duration parameter specifies the duration of each intervention, such as 5 minutes for each deep breathing exercise or 2 hours for each environmental improvement measure.

[0041] The adaptive feedback unit consists of an intelligent adjustment device and an integrated physiological parameter real-time monitoring component. The intelligent adjustment device has a built-in data analysis module that can receive and identify various parameters in the personalized health guidance thresholds, and adjust the implementation method and intensity of the intervention measures as needed based on these parameters. When the intervention type is exercise guidance, the device can adjust the specific content of the exercise plan through connected exercise equipment, such as reducing the running speed from 6 km / h to 5 km / h to match the intervention intensity range; when the intervention type is environmental improvement, the device can control the indoor air conditioning temperature setting to fluctuate within the intervention intensity range. The intelligent adjustment device can also dynamically adjust the implementation method based on real-time feedback; for example, if it finds that the user is not following the suggestions during the intervention, it can switch to voice prompts to enhance the intervention effect.

[0042] The integrated physiological parameter real-time monitoring component is synchronized with personalized health guidance thresholds, receiving monitoring indicators and time requirements. It monitors changes in physiological indicators during intervention in real time based on the photoplethysmography (PPG) method. This method uses a light-emitting diode to emit light of a specific wavelength, which penetrates human tissue and is received by a photoelectric sensor. It calculates relevant physiological indicators by utilizing changes in the light absorption characteristics of hemoglobin in the blood. Monitored indicators include heart rate variability (the amplitude of heart rate fluctuation per unit time); skin temperature fluctuation (body surface temperature changes collected by a contact temperature sensor); blood oxygen saturation change (the change in the proportion of oxygen-bound hemoglobin in the blood); movement posture adjustment amplitude (changes in limb movement angles captured by a built-in motion sensor); and intervention response time (the time from intervention initiation to the appearance of significant changes in physiological indicators).

[0043] The integrated real-time physiological parameter monitoring component preprocesses the collected real-time physiological feedback signals, including filtering to remove noise and standardizing data formats, before transmitting them wirelessly to the terminal co-control unit and the human-machine interface module. The terminal co-control unit adjusts the monitoring terminal's operating parameters based on these signals; for example, it increases the monitoring frequency when the heart rate variability exceeds the expected range. The human-machine interface module combines the real-time physiological feedback signals with health status assessment results, updating the display of the comprehensive health risk level so that users can promptly understand the effectiveness of intervention measures. Simultaneously, these signals are stored as historical data, which is used by the subsequent parameter optimization unit to adjust personalized health guidance thresholds, forming a closed-loop intervention and regulation mechanism.

[0044] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional data fusion and interactive processing system for a multimodal intelligent health monitoring terminal, characterized in that, include: A multidimensional data acquisition module, comprising a data fusion interface unit, wherein the data fusion interface unit generates multimodal raw data; The intelligent fusion processing module receives the multimodal raw data and includes a data fusion engine, a health status assessment unit, and a parameter optimization unit. The data fusion engine processes the multimodal raw data and outputs a feature fusion signal and an abnormal risk warning signal. The health status assessment unit processes the feature fusion signal and the user's historical health data to obtain a health status assessment result. The parameter optimization unit defines intervention constraints based on the health status assessment result and outputs personalized health guidance thresholds. The interactive response control module includes an adaptive feedback unit, a terminal collaborative control unit, and a human-computer interaction interface module. The adaptive feedback unit adjusts intervention parameters according to the personalized health guidance threshold. The terminal collaborative control unit adjusts the monitoring terminal's operating parameters according to the health status assessment results and the multimodal raw data, and collects real-time physiological feedback signals. The human-computer interaction interface module processes the abnormal risk warning signals and the health status assessment results and outputs a comprehensive health risk level.

2. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 1, characterized in that, The multidimensional data acquisition module includes a wearable biosensor unit, an environmental parameter detection unit, and a motion posture capture unit. The multidimensional data acquisition module is deployed on key parts of the user's body, clothing, and living environment to collect raw sensor signals. The motion posture capture unit acquires the user's limb movement trajectory data through an inertial measurement unit and forms a motion state signal. The data fusion interface unit receives the raw sensor signals and the motion state signal to integrate multi-source heterogeneous data and form the multimodal raw data. The multimodal raw data includes physiological parameter data, environmental influence data, and motion characteristic data.

3. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 2, characterized in that, The data fusion engine includes a feature extraction model based on a convolutional neural network and a Bayesian network fusion model. The physiological parameter data includes heart rate variability data, skin resistance change data, blood oxygen saturation fluctuation data, body temperature gradient distribution data, and respiratory rate cycle data. The heart rate variability data, skin resistance change data, blood oxygen saturation fluctuation data, body temperature gradient distribution data, and respiratory rate cycle data are input into the feature extraction model and output the feature fusion signal. The improved Bayesian network fusion model fuses the physiological parameter data, the environmental impact data, and the motion feature data and outputs the abnormal risk warning signal.

4. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 3, characterized in that, The feature extraction model includes a multi-scale feature decomposition submodule, a spatiotemporal correlation analysis submodule, a risk feature quantification submodule, and a fusion result visualization submodule. The multi-scale feature decomposition submodule performs empirical mode decomposition on the heart rate variability data and extracts the fluctuation features of frequency bands that meet the pre-abnormal physiological state frequency band range as health risk precursor signals. The spatiotemporal correlation analysis submodule uses an attention-weighted convolutional neural network to process the skin resistance change data, the blood oxygen saturation fluctuation data, and the motion feature data and outputs the data to the risk feature quantification submodule. The risk feature quantification submodule defines risk level calculation rules and generates the feature fusion signal by comprehensively evaluating the degree of deviation of each physiological parameter from the normal range, the influence weight of environmental parameters, and the matching degree of the motion state. The fusion result visualization submodule generates a dynamic health status heatmap of the feature fusion signal based on a four-color scale, using different color levels to represent the health risk level of different body parts.

5. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 2, characterized in that, The health status assessment unit processes the environmental impact data and the feature fusion signal using a health-intervention nonlinear relationship model, outputs a preliminary health assessment result, and uses a support vector machine algorithm to classify and optimize the preliminary health assessment result to obtain the final health status assessment result. The environmental impact data includes ambient temperature change signals, air quality index signals, and electromagnetic radiation intensity signals. The health-intervention nonlinear relationship model is constructed using the ambient temperature change signals, the feature fusion signal, and the electromagnetic radiation intensity signal. This model analyzes the correlation between environmental parameters and abnormal physiological indicators, and combines this with the environmental adaptation patterns in the user's historical health data to establish a quantitative assessment framework for the impact of environmental factors on health status.

6. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 5, characterized in that, The health status assessment unit further includes a health indicator database, an abnormal pattern recognition submodule, and an intervention priority ranking submodule. The health indicator database stores standard physiological parameter ranges for different age groups, genders, and health statuses. The abnormal pattern recognition submodule identifies abnormal change patterns of physiological indicators based on the feature fusion signal and the environmental impact data. The intervention priority ranking submodule obtains the health status assessment result through a decision tree algorithm based on the abnormal change patterns, the health-intervention nonlinear relationship model, and the air quality index signal.

7. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 2, characterized in that, The parameter optimization unit defines intervention constraints based on the exercise feature data, outputs a personalized health guidance threshold, and dynamically adjusts the personalized health guidance threshold to ensure it is not lower than the health and safety benchmark value through fuzzy PID control. The exercise feature data includes exercise intensity signal, exercise duration signal, and exercise type signal. The exercise intensity signal, exercise duration signal, and exercise type signal define the intervention constraints. By comprehensively considering the degree of influence of exercise on physiological indicators, the user's current health status, and historical intervention effects, the range of health guidance parameters under different exercise scenarios is determined. The fuzzy PID control dynamically adjusts the intensity and frequency of intervention measures to ensure that the personalized health guidance threshold is not lower than the health and safety benchmark value.

8. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 1, characterized in that, The human-computer interaction interface module includes a three-dimensional dynamic display platform and an early warning engine. The three-dimensional dynamic display platform receives the multimodal raw data and the real-time physiological feedback signals and dynamically renders the physiological indicator change process. The early warning engine processes the abnormal risk warning signals and the health status assessment results according to the improved analytic hierarchy process and outputs a comprehensive health risk level. When the comprehensive health risk level exceeds the risk level threshold, the human-computer interaction interface module controls the adaptive feedback unit to initiate intervention measures. The three-dimensional dynamic display platform supports multi-dimensional physiological data cross-sectional analysis, can display the correlation between key physiological indicators and intervention measures in real time, and simulate the physiological state change trend under different intervention schemes based on multibody dynamics.

9. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 8, characterized in that, The early warning engine includes a three-level warning mechanism with color-coded states: green for safety, yellow for attention, and red for emergency. When a key physiological indicator in the health status assessment deviates from the normal range to a first deviation threshold, the yellow attention state is activated, pushing health adjustment suggestions and reducing the recommended exercise intensity. When a key physiological indicator in the health status assessment deviates from the normal range to a second deviation threshold, high-intensity exercise recommendations are suspended, and the adaptive feedback unit is controlled to initiate mild intervention measures. When a key physiological indicator in the health status assessment deviates from the normal range to a third deviation threshold, the early warning engine issues an emergency health alert and controls the multimodal intelligent health monitoring terminal to automatically contact emergency contacts.

10. The multi-dimensional data fusion and interactive processing system of the multimodal intelligent health monitoring terminal according to claim 1, characterized in that, The personalized health guidance threshold includes parameters such as intervention type, intervention intensity range, intervention frequency recommendation, and duration. The adaptive feedback unit includes an intelligent adjustment device and an integrated physiological parameter real-time monitoring component. The intelligent adjustment device can identify the personalized health guidance threshold and adjust the implementation method and intensity of the intervention as needed. The integrated physiological parameter real-time monitoring component receives the personalized health guidance threshold and monitors changes in physiological indicators during the intervention process in real time based on the photoplethysmography method. The real-time physiological feedback signals include heart rate change rate, skin temperature fluctuation, blood oxygen saturation change, movement posture adjustment amplitude, and intervention response time.

Images