Intelligent accompanying machine interaction method and device based on physiological monitoring

Abstract

The embodiment of the invention provides an intelligent accompanying machine interaction method and device based on physiological monitoring, and the method and device achieve the quality optimization of monitoring data through innovatively constructing a multi-source physiological data processing mechanism and integrating heart rate, body temperature, heartbeat waveform and sleep state data. And designing a disease correlation analysis model based on a chronic disease map, and establishing a health risk assessment strategy for intelligent prediction in combination with a graph convolutional network and medical expert knowledge. A multi-level early warning response mechanism is introduced, and linkage response and accurate early warning of different risk levels are realized through classification processing of an emergency level, a severity level and a general level. According to the method, the defects of the traditional technology in the aspects of data processing, risk assessment, early warning response and the like are effectively overcome, and the service quality and reliability of intelligent accompanying are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of data processing, specifically to an intelligent companion machine interaction method and device based on physiological monitoring. Background Technology

[0002] Existing intelligent companionship machine interaction methods have significant shortcomings. Traditional systems lack systematicity in physiological data processing, making it difficult to effectively integrate heart rate, body temperature, heart rate waveform, and sleep status data, thus affecting the accuracy of monitoring.

[0003] Furthermore, existing technologies face bottlenecks in disease analysis. Most systems fail to fully utilize the knowledge of medical experts and lack a health risk assessment mechanism based on chronic disease atlases, resulting in less than ideal early warning effects.

[0004] Existing systems have technical shortcomings in early warning response. They lack the ability to classify health risks at multiple levels, making it difficult to achieve accurate early warnings at different levels through coordinated response mechanisms, thus affecting service reliability. Solving these problems is crucial for improving the level of intelligent companionship services. Summary of the Invention

[0005] To address the problems in existing technologies, this application provides an intelligent companionship machine interaction method and device based on physiological monitoring, which can effectively solve the shortcomings of traditional technologies in data processing, risk assessment and early warning response, and significantly improve the service quality and reliability of intelligent companionship.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] In a first aspect, this application provides an intelligent companion machine interaction method based on physiological monitoring, comprising:

[0008] Physiological data of the elderly is collected by using a smart bracelet to collect heart rate and body temperature data, using millimeter-wave radar to collect heartbeat waveform data, and using a smart pillow to collect sleep state data. The physiological data is segmented by a sliding window, a dynamic threshold model is constructed to calculate the data reliability score, abnormal data is corrected based on the reliability score, historical monitoring data is obtained from the user's health record to establish personalized baseline parameters, and the corrected data is normalized based on the personalized baseline parameters to obtain physiological feature vectors.

[0009] A symptom association analysis model is constructed. A chronic disease atlas is built based on medical expert knowledge. The symptom nodes and relation edges in the chronic disease atlas are input into a graph convolutional network to extract disease features. Symptom matching analysis is performed on the physiological feature vectors to obtain symptom combination features. A health risk matrix is ​​generated based on symptom correlation. The health risk matrix is ​​input into a prediction model to generate a health risk score. The health risk score is then time-series modeled to obtain the risk development trend.

[0010] The system implements a multi-level early warning response, classifying abnormal states into emergency, severe, and general levels based on the health risk score and risk development trend. For emergency-level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institutions, triggering a linkage response mechanism. For severe-level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general-level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0011] Furthermore, it also includes: inputting the heart rate signal collected by the photoplethysmography sensor built into the smart bracelet and the body temperature signal collected by the temperature sensor into the data acquisition module; the data acquisition module performs bandpass filtering on the heart rate signal to remove baseline drift and power frequency interference to obtain heart rate data; performs low-pass filtering on the body temperature signal to eliminate high-frequency noise to obtain body temperature data; uses millimeter-wave radar to transmit frequency-modulated continuous wave signals and receive echo signals; performs phase demodulation on the echo signals to extract Doppler frequency shift to obtain heartbeat waveform data; and collects pressure distribution data based on the pressure sensor array built into the smart pillow; extracts body movement information and respiratory rhythm from the pressure distribution data to obtain sleep state data.

[0012] The heart rate data, body temperature data, heartbeat waveform data, and sleep state data are segmented according to the sampling time. Each data segment is input into a dynamic threshold model. The dynamic threshold model calculates the mean and standard deviation based on the statistical characteristics of the data, constructs an adaptive threshold boundary by combining historical data distribution, judges the validity of the sampling points in the data segment, calculates the proportion of valid sampling points to obtain the data reliability score, and uses the data reliability score as a weighting coefficient to perform a weighted average of the data segments.

[0013] Furthermore, it also includes: comparing the data reliability score with a preset threshold, marking data segments with scores below the threshold as abnormal data segments, correcting the abnormal data segments using a piecewise linear interpolation method, constructing a Kalman filter to perform state estimation and smoothing on the corrected data, recombining the smoothed data according to the time series to obtain a corrected data sequence, retrieving the target user's recent physiological monitoring data from the user's health record, and calculating the statistical parameters of the physiological monitoring data based on a time-weighted average method to obtain personalized baseline parameters, wherein the personalized baseline parameters include the mean level and fluctuation range of physiological indicators;

[0014] The mean level in the personalized baseline parameters is subtracted from the corrected data sequence to obtain the fluctuation amount. The fluctuation amount is normalized based on the fluctuation range in the personalized baseline parameters to obtain standardized features. A feature extraction network is constructed. The feature extraction network performs dimensionality reduction mapping on the standardized features based on a multilayer perceptron structure. The mapping result is numerically stabilized through a batch normalization layer. The processing result is converted into a physiological feature vector through an activation function.

[0015] Furthermore, it also includes: constructing a chronic disease atlas based on medical expert knowledge, using symptom descriptions and risk characteristics of common chronic diseases as node information, and using the correlation between symptoms and disease evolution patterns as edge information; semantically encoding the node information to generate node feature vectors; calculating the weight coefficients of edges based on co-occurrence frequency to generate an adjacency matrix; inputting the node feature vectors and the adjacency matrix into a graph convolutional network; the graph convolutional network propagates node information on the graph structure through a message passing mechanism; using an attention mechanism to weight the contributions of different nodes; and extracting local structural features and global semantic features to obtain disease features.

[0016] The physiological feature vector is input into the disease matching module. The disease matching module calculates the matching degree between the physiological feature vector and the disease feature based on cosine similarity, constructs a symptom correlation matrix to perform correlation analysis on the matching results, uses a deep neural network to combine features of the symptom correlation matrix, and captures the dependency relationship between symptoms through a multi-head self-attention mechanism to generate symptom combination features.

[0017] Furthermore, it also includes: calculating the correlation strength between different symptoms based on the symptom combination features, constructing a symptom transition probability matrix to characterize the symptom evolution pattern, multiplying the symptom transition probability matrix with symptom risk weights to obtain an initial risk matrix, iteratively updating the initial risk matrix using a Markov chain model, calculating the risk accumulation effect of different symptom combinations based on a steady-state distribution, correcting the risk accumulation effect with an expert rule base to obtain a health risk matrix, inputting the health risk matrix into a deep neural network, the deep neural network extracting risk features through multiple convolutions, and generating a health risk score through feature fusion via a fully connected layer;

[0018] A temporal analysis network is constructed, which performs sequence modeling of the health risk score based on long short-term memory units. It selectively retains historical state information through a gating mechanism and captures temporal dependencies by combining an attention mechanism. It predicts the changing trend of the health risk score and calculates trend features using a sliding time window. The trend features are then input into a conditional random field model, which infers the risk development trend based on sequence labeling methods.

[0019] Furthermore, it also includes: constructing an early warning classification model, mapping the health risk score to a preset score range, setting a risk level threshold based on the experience of medical experts, inputting the risk development trend into a trend prediction network, the trend prediction network using a recurrent neural network structure to dynamically update the trend features, calculating the risk evolution speed and acceleration, classifying abnormal states based on a combination rule of risk score and evolution features, the combination rule prioritizing emergency situations where the risk score exceeds the threshold, assessing the persistence of risk in combination with evolution features, and classifying the assessment results into three levels: emergency, severe, and general.

[0020] A tiered response mechanism is adopted, triggering differentiated early warning processes for different levels of abnormal states. An emergency plan database is built for emergency-level abnormalities, which includes contact information and handling procedures. A guardian feedback channel is established for severe-level abnormalities, which supports two-way information interaction. A health management module is set up for general-level abnormalities, which tracks and records the abnormal state in real time.

[0021] Furthermore, it also includes: constructing an early warning information distribution module to generate early warning information containing location information and symptom descriptions for emergency-level anomalies; pushing the early warning information in parallel to the emergency center dispatch system, the guardian's mobile terminal, and the community medical management platform based on a message queue; initiating a linkage response mechanism, which automatically allocates treatment tasks based on a preset process; constructing a status monitoring engine for severe-level anomalies, which continuously collects physiological data and extracts key indicators, pushes abnormal fluctuation information of key indicators to the guardian, and sends a follow-up application to the community medical management platform; and initiating a health management task for general-level anomalies, which records the time of occurrence of the anomaly and the symptoms.

[0022] A data storage module is constructed to uniformly store the early warning information distribution records, the status monitoring data, and the health management records into an interactive data table. The data in the interactive data table is indexed according to the time and function dimensions. Invalid information is filtered based on data cleaning rules, and the cleaned data is written into the user's health record using an incremental update strategy. The user's health record adopts a distributed storage structure to achieve data backup and disaster recovery.

[0023] Secondly, this application provides an intelligent companion robot interaction device based on physiological monitoring, comprising:

[0024] The physiological data acquisition module is used to collect physiological data of the elderly. It collects heart rate and body temperature data through a smart bracelet, heartbeat waveform data through millimeter-wave radar, and sleep state data based on a smart pillow. The physiological data is segmented by a sliding window, a dynamic threshold model is constructed to calculate the data reliability score, abnormal data is corrected based on the reliability score, historical monitoring data is obtained from the user's health record to establish personalized baseline parameters, and the corrected data is normalized based on the personalized baseline parameters to obtain physiological feature vectors.

[0025] The symptom analysis module is used to construct a symptom association analysis model, build a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relation edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend.

[0026] The multi-level response module is used to execute multi-level early warning responses. Based on the health risk score and the risk development trend, abnormal states are divided into emergency, severe, and general levels. For emergency-level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institutions, and a linkage response mechanism is triggered. For severe-level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general-level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0027] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the intelligent companion machine interaction method based on physiological monitoring.

[0028] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent companion machine interaction method based on physiological monitoring.

[0029] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the aforementioned intelligent companion machine interaction method based on physiological monitoring.

[0030] As described above, this application provides an intelligent companion machine interaction method and device based on physiological monitoring. It innovatively constructs a multi-source physiological data processing mechanism, integrating heart rate, body temperature, heartbeat waveform, and sleep state data to optimize the quality of monitoring data. A symptom association analysis model based on a chronic disease atlas is designed, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, classifying risks into emergency, severe, and general levels to achieve coordinated responses and accurate early warnings for different risk levels. This method effectively addresses the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the intelligent companion machine interaction method based on physiological monitoring in the embodiments of this application;

[0033] Figure 2 This is a structural diagram of the intelligent companion robot interaction device based on physiological monitoring in the embodiments of this application;

[0034] Figure 3 This is a schematic diagram of the structure of the electronic device in the embodiments of this application.

[0035] Figure label:

[0036] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0039] To address the shortcomings of existing technologies, this application provides an intelligent companionship machine interaction method and device based on physiological monitoring. It innovatively constructs a multi-source physiological data processing mechanism, integrating heart rate, body temperature, heartbeat waveform, and sleep state data to optimize the quality of monitoring data. A symptom association analysis model based on a chronic disease atlas is designed, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, classifying risks into emergency, severe, and general levels to achieve coordinated responses and accurate early warnings for different risk levels. This method effectively solves the deficiencies of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0040] To effectively address the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, and to significantly improve the service quality and reliability of intelligent companionship, this application provides an embodiment of an intelligent companionship machine interaction method based on physiological monitoring. See [link to embodiment]. Figure 1The intelligent companion machine interaction method based on physiological monitoring specifically includes the following:

[0041] Step S101: Collect physiological data of the elderly. Collect heart rate and body temperature data through a smart bracelet, collect heartbeat waveform data using millimeter-wave radar, and collect sleep state data based on a smart pillow. Perform sliding window segmentation on the physiological data, construct a dynamic threshold model to calculate the data reliability score, correct abnormal data based on the reliability score, obtain historical monitoring data from the user's health record to establish personalized baseline parameters, and normalize the corrected data based on the personalized baseline parameters to obtain physiological feature vectors.

[0042] Optionally, this embodiment innovatively designs a multi-source data acquisition and processing scheme to address issues such as unstable data quality and inaccurate personalized feature extraction during the collection of physiological data from the elderly. This embodiment first constructs a multimodal data acquisition network and collects heart rate signals through the photoplethysmography sensor built into the smart bracelet. The system designs a signal processing mechanism: Clean_Signal = BPF(Raw_Signal, f_low, f_high), where Raw_Signal is the raw signal, and f_low and f_high are the cutoff frequencies of the bandpass filter, respectively. This mechanism effectively removes baseline drift and power frequency interference. In practical applications in nursing homes, due to changes in the activity intensity and environmental conditions of the elderly, this adaptive filtering method can significantly improve the reliability of heart rate data.

[0043] This embodiment deeply optimizes the heartbeat detection strategy of millimeter-wave radar. Addressing the signal interference problem caused by significant body movement in the elderly, a heartbeat waveform extraction method based on the Doppler effect was designed. The system transmits a frequency-modulated continuous wave signal and receives the echo, extracting the weak heartbeat signal through phase demodulation technology. Special attention is paid to the activity status of the elderly; when vigorous body movement is detected, the system automatically adjusts the signal processing parameters to ensure the continuity of the heartbeat waveform. For example, when the elderly turn over or cough, the system can maintain the stability of the heartbeat signal through a dynamic compensation mechanism. This non-contact monitoring method greatly improves the user comfort for the elderly.

[0044] This embodiment innovatively implements a sleep monitoring mechanism for a smart pillow. A high-precision pressure distribution map is constructed based on a pressure sensor array, and body movement information and respiratory rhythm are extracted through time-series analysis. The system employs a matrix factorization algorithm to separate pressure signals from different sources, accurately identifying pressure changes caused by breathing, heartbeat, and body movement. In particular, for elderly individuals with sleep disorders, the system can accurately record their number of turns, sleep apnea, and other abnormalities, providing objective evidence for medical personnel to assess sleep quality.

[0045] This embodiment deeply optimizes the data reliability assessment mechanism. Multi-source data is segmented using a sliding time window, and a dynamic threshold model is constructed to achieve anomaly detection. The system establishes adaptive threshold boundaries by calculating data statistical characteristics and combining them with historical data distribution. Different evaluation standards are adopted for different types of physiological data. For example, for heart rate data, the system focuses on the continuity and magnitude of change; for body temperature data, it emphasizes the stability and degree of deviation. This multi-dimensional reliability assessment ensures the data quality for subsequent analysis.

[0046] This embodiment achieves data smoothing through a Kalman filter. For abnormal data segments, the system first uses piecewise linear interpolation for preliminary correction, and then achieves smooth data transition through state estimation: State_t = AState_t-1 + K(Measure_t - H × State_t-1), where State_t is the current state estimate, Measure_t is the measured value, and K is the Kalman gain. This state estimation-based smoothing effectively eliminates data jumps and maintains the continuity of physiological indicators.

[0047] This innovative design not only solves the data quality problems of traditional methods but also establishes a continuously optimized personalized monitoring framework. Through multi-source data fusion and dynamic threshold adjustment, the system can accurately capture changes in the physiological state of the elderly. This data processing mechanism based on personalized baselines ensures that the system maintains efficient monitoring capabilities and reliable assessment results when dealing with different elderly individuals. In elderly care service practice, this intelligent physiological monitoring solution significantly improves the accuracy of health management.

[0048] This embodiment achieves an upgrade in health monitoring for the elderly population by establishing a complete data acquisition and processing chain. The system can dynamically adjust processing strategies based on historical data, avoiding the limitations of traditional fixed-parameter schemes. Through multi-dimensional data fusion and normalization processing, the accuracy and adaptability of physiological feature extraction are significantly improved, providing reliable data support for subsequent health risk assessments. This intelligent monitoring mechanism demonstrates strong adaptability and service effectiveness in elderly care service scenarios.

[0049] Step S102: Construct a symptom association analysis model, construct a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relationship edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend;

[0050] Optionally, this embodiment addresses the problems of incomplete symptom association analysis and inaccurate health risk assessment in chronic disease monitoring of the elderly by innovatively designing a symptom analysis scheme based on graph neural networks. This embodiment first constructs a medical knowledge graph, encoding information such as the symptom characteristics and evolutionary patterns of common chronic diseases into a graph structure through expert experience. The system designs a node encoding mechanism: Node_Feature = Encoder(Symptom_Description, Risk_Pattern), where Symptom_Description is the symptom description text and Risk_Pattern is the risk feature pattern. This mechanism achieves a structured expression of symptom information. In practical applications in nursing homes, this knowledge graph-based analysis method can comprehensively characterize the features and evolutionary patterns of common chronic diseases in the elderly.

[0051] This embodiment deeply optimizes the disease feature extraction strategy. Addressing the complex interrelationships of chronic diseases, a feature extraction module based on graph convolutional networks is designed. The system propagates node information across the graph structure via a message passing mechanism, enabling information exchange between symptom nodes. Particular attention is paid to the correlations and transformation patterns between symptoms, progressively extracting high-order features through multi-layer graph convolutional operations. For example, when elderly individuals exhibit symptoms such as palpitations and shortness of breath, the system can infer possible causes based on symptom correlation analysis and predict symptom progression trends. This graph-based analysis method effectively captures the complex dependencies between symptoms.

[0052] This embodiment innovatively implements a symptom matching analysis mechanism. Based on extracted disease features, the system uses cosine similarity to calculate the degree of matching between physiological feature vectors and disease features. By constructing a symptom correlation matrix, the system can analyze the combination patterns and order of occurrence of different symptoms. In particular, for chronic diseases prevalent among the elderly, the system pays more attention to the gradual changes in symptoms and their long-term cumulative effects. For example, for patients with hypertension, the system not only focuses on fluctuations in blood pressure but also analyzes the frequency and severity of related symptoms such as headaches and dizziness.

[0053] This embodiment deeply optimizes the health risk assessment strategy. The system constructs a health risk matrix based on symptom combination characteristics: Risk_Matrix = Symptom_Matrix × Transfer_Probability, where Symptom_Matrix is ​​the symptom combination matrix and Transfer_Probability is the symptom transfer probability. This risk assessment mechanism can comprehensively consider the severity and evolution trend of symptoms, providing a reliable basis for health risk early warning. In particular, the system iteratively updates the risk matrix using a Markov chain model to calculate the cumulative risk effect of different symptom combinations.

[0054] This innovative design not only solves the problem of incomplete symptom analysis in traditional methods but also establishes a continuously optimized health risk assessment framework. Through the combination of graph neural networks and deep learning, the system can accurately grasp changes in the health status of the elderly. This knowledge graph-based analysis mechanism ensures that the system maintains high analytical capabilities and reliable predictive performance even when faced with complex combinations of symptoms. In elderly care practice, this intelligent health monitoring solution significantly improves the accuracy and foresight of chronic disease prevention and control.

[0055] This embodiment upgrades health monitoring for the elderly population by establishing a complete disease analysis and risk assessment chain. The system can dynamically adjust assessment strategies based on historical data, avoiding the limitations of traditional fixed models. Through multi-dimensional feature analysis and risk assessment, it significantly improves the accuracy and predictability of health monitoring, providing reliable technical support for chronic disease management in the elderly. This intelligent monitoring mechanism demonstrates powerful analytical capabilities and early warning effects in elderly care service scenarios.

[0056] This embodiment not only improves the accuracy of health risk assessment but also establishes an ever-evolving intelligent system through continuous data accumulation and model optimization. This knowledge graph-based analysis mechanism ensures that the system continuously improves with use, providing increasingly precise health monitoring services for the elderly. In practical applications, this self-optimization mechanism significantly enhances the system's long-term service quality and early warning effectiveness, providing strong technical support for intelligent elderly care services.

[0057] Step S103: Execute a multi-level early warning response. Based on the health risk score and the risk development trend, the abnormal state is divided into emergency, severe, and general levels. For emergency level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institution, and a linkage response mechanism is triggered. For severe level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0058] Optionally, this embodiment innovatively designs a multi-level early warning and response scheme to address the issue of timely detection and rapid response to abnormal health conditions in the elderly. This embodiment first constructs an early warning grading model, converting health risk scores to a standard range using a risk mapping function: Risk_Level = f(Score, Trend), where Score is the risk score and Trend is the development trend. The system sets grading thresholds based on medical expert experience and dynamically adjusts them in conjunction with the risk evolution speed and acceleration. In practical applications in nursing homes, this multi-dimensional risk assessment method can accurately identify health abnormalities of different degrees, achieving precise triggering of tiered early warnings.

[0059] This embodiment deeply optimizes the handling process for emergency-level anomalies. A multi-party collaborative emergency response mechanism is designed for emergencies that may endanger life. The system automatically generates a warning information package containing geographical location, symptom description, and medical history, and pushes it to various response units in parallel via a message queue. Special attention is paid to the real-time nature of information transmission, employing a heartbeat mechanism to ensure message delivery and requiring confirmation feedback from the recipient. For example, when a severe heart rate abnormality or persistent breathing difficulty is detected in an elderly person, the system immediately activates the emergency plan, simultaneously notifying the emergency center, guardian, and community medical institutions to ensure that rescue forces can arrive at the scene in the shortest possible time.

[0060] This embodiment innovatively implements a monitoring scheme for severe abnormalities. Based on a real-time status monitoring engine, the system continuously collects and analyzes changes in physiological indicators. When an abnormal trend is detected, an early warning message is pushed to the caregiver via a mobile terminal. The message includes the specific value of the abnormal indicator, its changing trend, and potential health risks. Simultaneously, the system automatically sends a follow-up request to the community healthcare platform, coordinating with doctors for timely examination. This preventative monitoring mechanism effectively prevents further deterioration of health conditions. For example, for elderly individuals with persistently high blood pressure, the system regularly reports measurement data to family members and arranges for regular checkups by community doctors.

[0061] This embodiment significantly optimizes the tracking and management of general-level anomalies. The system includes a health management module that digitally records the time, duration, and specific manifestations of anomalies. Based on this data, the system periodically generates health analysis reports, which include trends in physiological indicators, patterns of anomaly occurrence, and health recommendations. Particular attention is paid to the correlation between the elderly's daily habits and their health status, using data analysis to identify potential health risk factors. This continuous health management approach helps caregivers better understand the health status of the elderly.

[0062] This embodiment achieves reliable storage of interactive data through a distributed storage architecture. The system employs a multi-dimensional index structure to organize early warning records, monitoring data, and management logs, ensuring rapid data retrieval and efficient utilization. Special attention is paid to the timeliness of data management, with differentiated storage strategies set for different types of interactive data. For example, emergency early warning records need to be permanently stored, while general status data can be archived periodically. Through an incremental update mechanism, the system can continuously optimize user health records, providing more comprehensive historical data for subsequent health assessments.

[0063] This innovative design not only solves the problems of untimely early warning response and non-standardized handling procedures in traditional methods, but also establishes a sustainable and optimized health management framework. Through the combination of multi-level early warning and coordinated response, the system can take corresponding measures based on the severity of the anomaly. This risk-level-based response mechanism ensures that the system maintains efficient response capabilities and reliable handling effects when facing different types of health anomalies. In elderly care service practice, this intelligent early warning scheme significantly improves the effectiveness of health risk prevention and control.

[0064] This embodiment achieves an upgrade in health monitoring for the elderly population by establishing a complete early warning response and data management chain. The system can dynamically adjust response strategies based on real-time status, avoiding the limitations of traditional fixed procedures. Through multi-layered early warning mechanisms and data accumulation, the accuracy and foresight of health management are significantly improved, providing reliable health protection for the elderly. This intelligent monitoring mechanism demonstrates strong practical value and social benefits in elderly care service scenarios.

[0065] As described above, the intelligent companion machine interaction method based on physiological monitoring provided in this application can optimize the quality of monitoring data by innovatively constructing a multi-source physiological data processing mechanism and integrating heart rate, body temperature, heartbeat waveform, and sleep state data. A symptom association analysis model based on a chronic disease atlas is designed, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, achieving coordinated response and accurate early warning for different risk levels through emergency, severe, and general classification processing. This method effectively solves the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0066] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0067] Step S201: The heart rate signal collected by the photoplethysmography sensor built into the smart bracelet and the body temperature signal collected by the temperature sensor are input into the data acquisition module. The data acquisition module performs bandpass filtering on the heart rate signal to remove baseline drift and power frequency interference to obtain heart rate data, and performs low-pass filtering on the body temperature signal to eliminate high-frequency noise to obtain body temperature data. The module uses millimeter-wave radar to transmit frequency-modulated continuous wave signals and receive echo signals. The echo signals are phase demodulated to extract Doppler frequency shift to obtain heartbeat waveform data. Pressure distribution data is collected based on the pressure sensor array built into the smart pillow. Body movement information and respiratory rhythm are extracted from the pressure distribution data to obtain sleep state data.

[0068] Step S202: The heart rate data, body temperature data, heartbeat waveform data, and sleep state data are segmented according to the sampling time. Each data segment is input into a dynamic threshold model. The dynamic threshold model calculates the mean and standard deviation based on the statistical characteristics of the data, constructs an adaptive threshold boundary by combining historical data distribution, judges the validity of the sampling points in the data segment, calculates the proportion of valid sampling points to obtain the data reliability score, and uses the data reliability score as a weighting coefficient to perform a weighted average of the data segments.

[0069] Optionally, this embodiment innovatively designs a multimodal data acquisition and preprocessing scheme to address issues such as unstable signal quality and asynchronous multi-source data during the physiological data collection process for the elderly. This embodiment first constructs a smart bracelet data acquisition module, which collects heart rate signals in real time using a photoplethysmography (PPG) sensor. The system employs a signal processing mechanism: Clean_Signal = BPF(Raw_Signal, f1, f2), where Raw_Signal is the raw signal, and f1 and f2 are the low-frequency and high-frequency cutoff frequencies of the bandpass filter, respectively. This mechanism effectively removes baseline drift and power frequency interference. In practical applications in nursing homes, due to the low activity levels of the elderly, heart rate signals are easily affected by body movement and environmental noise; this adaptive filtering method can significantly improve signal quality.

[0070] This embodiment deeply optimizes the heartbeat detection strategy of millimeter-wave radar. Addressing the potential discomfort that traditional contact sensors may cause to the elderly, a non-contact heartbeat detection method based on the Doppler effect was designed. The system transmits a frequency-modulated continuous wave signal and receives the reflected echo, using phase demodulation technology to extract the weak heartbeat signal. Special attention is paid to the elderly person's activity level; when vigorous body movement is detected, the system automatically adjusts the signal processing parameters to ensure the continuity of the heartbeat waveform. For example, when the elderly person turns over or coughs, the system can maintain the stability of the heartbeat signal through a dynamic compensation mechanism. This non-contact monitoring method greatly improves the user comfort for the elderly.

[0071] This embodiment innovatively implements a sleep monitoring mechanism for a smart pillow. A high-precision pressure distribution map is constructed based on a pressure sensor array, and pressure signals from different sources are separated using a matrix factorization algorithm. The system employs principal component analysis (PCA) to extract body movement information and respiratory rhythm: Movement_Pattern = PCA(Pressure_Matrix), where Pressure_Matrix is ​​the pressure distribution matrix, and PCA is the principal component analysis operation. This pressure signal analysis mechanism can accurately identify pressure changes caused by breathing, heartbeat, and body movement. For elderly people with sleep disorders, the system can accurately record their number of times they turn over, apnea, and other abnormal conditions.

[0072] This embodiment deeply optimizes the data segmentation and quality assessment strategies. To address the issue of inconsistent sampling frequencies among multi-source data, the system designs a timestamp-based data alignment method. Multimodal data is synchronously segmented using a sliding time window, ensuring consistency across different sources in the temporal dimension. Specifically, the system dynamically adjusts the size of the time window based on the changing characteristics of different physiological indicators. For example, a smaller time window is used for rapidly changing heart rate data, while a larger time window is used for relatively stable body temperature data.

[0073] This innovative design not only solves the data acquisition problems of traditional methods but also establishes a reliable data quality assessment framework. Through a dynamic threshold model for anomaly detection, the system can accurately identify and label unreliable data segments. This statistically based assessment mechanism ensures that the system maintains efficient data acquisition capabilities and reliable quality assessment results even in complex monitoring environments. In elderly care service practice, this intelligent data acquisition solution significantly improves the accuracy of health monitoring.

[0074] This embodiment achieves an upgrade in physiological monitoring for the elderly population by establishing a complete data acquisition and quality control chain. The system can dynamically adjust processing strategies based on historical data, avoiding the limitations of traditional fixed-parameter schemes. Through multi-dimensional data acquisition and quality assessment, the reliability and continuity of physiological data are significantly improved, providing high-quality data support for subsequent health risk assessments. This intelligent monitoring mechanism demonstrates strong adaptability and service effectiveness in elderly care service scenarios.

[0075] This embodiment not only improves the accuracy of data collection but also establishes a continuously evolving monitoring system through ongoing parameter optimization and quality control. This multimodal acquisition mechanism ensures that the system continuously improves with use, providing increasingly accurate health monitoring services for the elderly. In practical applications, this self-optimization mechanism significantly enhances the long-term service quality and user experience of the system, providing a reliable data foundation for intelligent elderly care services.

[0076] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0077] Step S301: Compare the data reliability score with a preset threshold, mark data segments with scores below the threshold as abnormal data segments, correct the abnormal data segments using a piecewise linear interpolation method, construct a Kalman filter to perform state estimation and smoothing on the corrected data, recombine the smoothed data according to the time series to obtain the corrected data sequence, retrieve the target user's recent physiological monitoring data from the user's health record, calculate the statistical parameters of the physiological monitoring data based on the time-weighted average method to obtain personalized baseline parameters, the personalized baseline parameters include the mean level and fluctuation range of physiological indicators;

[0078] Step S302: Subtract the mean level in the personalized baseline parameters from the corrected data sequence to obtain the fluctuation amount. Normalize the fluctuation amount based on the fluctuation range in the personalized baseline parameters to obtain standardized features. Construct a feature extraction network. The feature extraction network performs dimensionality reduction mapping on the standardized features based on a multilayer perceptron structure. Perform numerical stabilization processing on the mapping results through a batch normalization layer. Convert the processing results into physiological feature vectors through an activation function.

[0079] Optionally, this embodiment innovatively designs a data correction and feature extraction scheme to address issues such as unstable data quality and inaccurate personalized feature extraction during the collection of physiological data from the elderly. This embodiment first constructs a data quality assessment module, identifying abnormal data segments through a reliability threshold mechanism. The system designs a scoring mechanism: Quality_Score = f(Stability, Continuity, Range), where Stability represents data stability, Continuity represents data continuity, and Range represents the reasonableness of the data range. In practical applications in nursing homes, due to changes in the activity status and environmental conditions of the elderly, this multi-dimensional quality assessment method can accurately identify data segments that need correction.

[0080] This embodiment deeply optimizes the correction strategy for abnormal data. For data segments with scores below a threshold, a preliminary correction method based on piecewise linear interpolation is designed. The system reconstructs the data by analyzing the effective values ​​before and after the data segment and constructing a local interpolation function. Special attention is paid to the continuity requirements of the data, and the variation patterns of physiological indicators are considered during the interpolation process. For example, for the correction of heart rate data, the system references the variation range of normal physiological rhythms to avoid generating unrealistic values. This interpolation method based on physiological characteristics ensures the rationality of the corrected data.

[0081] This embodiment innovatively implements a data smoothing mechanism. A state estimation model is constructed based on a Kalman filter: State_t = AState_t-1 + K(Measure_t - H × State_t-1), where State_t is the current state estimate, Measure_t is the measured value, K is the Kalman gain, and A and H are the state transition matrix and observation matrix, respectively. This state estimation mechanism can effectively eliminate data fluctuations and achieve a smooth numerical sequence. In particular, the system adjusts the filtering parameters according to the characteristics of different physiological indicators to ensure the targeted processing effect.

[0082] This embodiment deeply optimizes the strategy for establishing personalized baseline parameters. The system retrieves recent monitoring data from the user's health record and calculates statistical characteristics using a time-weighted average method. Particular attention is paid to the timeliness of the data, with recent data receiving higher weight in the calculation. In this way, the system can accurately grasp the personalized characteristics of the elderly's physiological indicators, including the mean level and normal fluctuation range of the indicators. This personalized modeling based on historical data provides an important reference for subsequent feature extraction.

[0083] This embodiment achieves efficient representation of physiological features through a feature extraction network. A feature mapping model is designed based on a multilayer perceptron structure, mapping standardized features to a low-dimensional feature space through nonlinear transformation. The system pays particular attention to the expressive power of features, ensuring the stability of numerical distribution through batch normalization layers and employing appropriate activation functions to enhance the model's nonlinear expressive power. This deep learning framework can extract effective health features from complex physiological data.

[0084] This innovative design not only solves the data quality problems of traditional methods but also establishes a continuously optimizing feature extraction framework. Through a combination of data correction and feature learning, the system can accurately capture the health status characteristics of the elderly. This deep learning-based feature extraction mechanism ensures that the system maintains high-efficiency feature extraction capabilities and reliable representation effects when dealing with different elderly individuals. In elderly care service practice, this intelligent feature extraction scheme significantly improves the accuracy of health monitoring.

[0085] This embodiment achieves an upgrade in health monitoring for the elderly population by establishing a complete data processing and feature extraction chain. The system can dynamically adjust processing strategies based on historical data, avoiding the limitations of traditional fixed-parameter schemes. Through multi-dimensional data correction and feature extraction, the accuracy and adaptability of health feature expression are significantly improved, providing reliable feature support for subsequent health risk assessment. This intelligent feature extraction mechanism demonstrates strong adaptability and analytical effectiveness in elderly care service scenarios.

[0086] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0087] Step S401: Construct a chronic disease atlas based on medical expert knowledge. Use the symptom descriptions and risk characteristics of common chronic diseases as node information, and the correlation between symptoms and the disease evolution pattern as edge information. Semantically encode the node information to generate node feature vectors. Calculate the weight coefficients of the edges based on co-occurrence frequency to generate an adjacency matrix. Input the node feature vectors and the adjacency matrix into a graph convolutional network. The graph convolutional network propagates node information on the graph structure through a message passing mechanism. Use an attention mechanism to weight the contribution of different nodes and extract local structural features and global semantic features to obtain disease features.

[0088] Step S402: Input the physiological feature vector into the disease matching module. The disease matching module calculates the matching degree between the physiological feature vector and the disease feature based on cosine similarity, constructs a symptom correlation matrix to perform correlation analysis on the matching results, uses a deep neural network to combine features of the symptom correlation matrix, captures the dependency relationship between symptoms through a multi-head self-attention mechanism, and generates symptom combination features.

[0089] Optionally, this embodiment addresses the problems of incomplete symptom association analysis and inaccurate symptom matching in chronic disease monitoring of the elderly by innovatively designing a symptom analysis scheme based on knowledge graphs. This embodiment first constructs a medical knowledge graph, encoding information such as the symptom characteristics and evolutionary patterns of common chronic diseases into a graph structure through expert experience. The system designs a node encoding mechanism: Node_Feature = Encoder(Symptom_Text, Risk_Pattern), where Symptom_Text is the symptom description text and Risk_Pattern is the risk feature pattern. This mechanism enables the structured expression of symptom information. In practical applications in nursing homes, this knowledge graph-based analysis method can comprehensively characterize the features and evolutionary patterns of common chronic diseases in the elderly.

[0090] This embodiment deeply optimizes the graph structure construction strategy. Addressing the complex relationships between chronic disease symptoms, a method for calculating edge weights based on co-occurrence frequency is designed. The system analyzes a large number of medical records, statistically analyzing the co-occurrence frequency of different symptoms, and transforms this statistical characteristic into edge weight information. Particular attention is paid to the temporal relationships between symptoms; for example, certain symptoms often appear in a specific order, or the occurrence of certain symptoms increases the probability of other symptoms. This data-driven graph construction method accurately reflects the dependencies between symptoms.

[0091] This embodiment innovatively implements a feature extraction mechanism for graph convolutional networks. Based on a message-passing mechanism, a node information propagation rule is designed: Node_Update = Aggregate(Neighbor_Features, Edge_Weights), where Neighbor_Features are the features of neighboring nodes, and Edge_Weights are the edge weights. This information propagation mechanism can effectively integrate local structural features and global semantic information. In particular, the system dynamically adjusts the importance weights of different nodes through an attention mechanism to ensure that key symptoms receive sufficient attention during feature extraction.

[0092] This embodiment deeply optimizes the symptom matching strategy. The system uses cosine similarity to calculate the degree of matching between physiological feature vectors and disease features. This similarity measurement method can effectively capture semantic associations in the feature space. Special attention is paid to multi-dimensional feature matching, considering not only the matching degree of individual symptoms, but also analyzing the overall similarity of symptom combinations. For example, when multiple related symptoms are detected in an elderly person, the system will comprehensively evaluate the degree of matching between these symptoms and different disease patterns.

[0093] This embodiment utilizes a deep neural network to generate symptom combination features. The system employs a multi-head self-attention mechanism, capturing feature dependencies across different dimensions through parallel attention computation. Each attention head is responsible for focusing on a specific type of symptom association. Through this multi-dimensional feature extraction approach, the system can comprehensively understand the clinical significance of symptom combinations. For example, the simultaneous occurrence of certain symptoms may indicate the development trend of a specific disease, while the sequential appearance of certain symptoms may reflect the evolutionary process of the disease.

[0094] This innovative design not only solves the problem of incomplete symptom analysis in traditional methods but also establishes a continuously optimizing knowledge graph framework. Through the combination of graph neural networks and deep learning, the system can accurately grasp the health status characteristics of the elderly. This knowledge graph-based analysis mechanism ensures that the system maintains high analytical capabilities and reliable predictive performance even when faced with complex combinations of symptoms. In elderly care practice, this intelligent health monitoring solution significantly improves the accuracy and foresight of chronic disease prevention and control.

[0095] This embodiment upgrades health monitoring for the elderly population by establishing a complete symptom analysis and feature matching chain. The system can dynamically adjust its analysis strategy based on historical data, avoiding the limitations of traditional fixed models. Through multi-dimensional feature analysis and matching calculations, the accuracy and predictability of health monitoring are significantly improved, providing reliable technical support for the management of chronic diseases in the elderly. This intelligent monitoring mechanism demonstrates powerful analytical capabilities and early warning effects in elderly care service scenarios.

[0096] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0097] Step S501: Calculate the correlation strength between different symptoms based on the symptom combination features, construct a symptom transition probability matrix to characterize the symptom evolution pattern, multiply the symptom transition probability matrix with the symptom risk weight to obtain an initial risk matrix, use a Markov chain model to iteratively update the initial risk matrix, calculate the risk accumulation effect of different symptom combinations based on the steady-state distribution, and correct the risk accumulation effect with an expert rule base to obtain a health risk matrix. Input the health risk matrix into a deep neural network, which extracts risk features through multiple convolutions and performs feature fusion through a fully connected layer to generate a health risk score.

[0098] Step S502: Construct a temporal analysis network. The temporal analysis network performs sequence modeling on the health risk score based on long short-term memory units, selectively retains historical state information through a gating mechanism, captures temporal dependencies by combining an attention mechanism, predicts the changing trend of the health risk score, calculates trend features using a sliding time window, and inputs the trend features into a conditional random field model. The conditional random field model infers the risk development trend based on the sequence labeling method.

[0099] Optionally, this embodiment addresses the problems of insufficient symptom correlation analysis and inaccurate risk prediction in chronic disease risk assessment for the elderly by innovatively designing a health risk assessment scheme based on time series analysis. This embodiment first constructs a symptom correlation analysis module, building a correlation matrix by calculating the correlation coefficients between symptom combination features. The system designs a transition probability calculation mechanism: Trans_Prob = Count(Si→Sj) / Count(Si), where Si and Sj represent different symptom states, and Count represents the count of state transitions. In practical applications in nursing homes, this statistical correlation analysis method can accurately characterize the evolutionary patterns between different symptoms.

[0100] This embodiment deeply optimizes the risk matrix construction strategy. An iterative update mechanism based on Markov chains is designed to address the cumulative risk effect of different symptom combinations. The system combines the symptom transition probability matrix with expert-defined risk weights to generate an initial risk matrix. Through multiple iterative calculations, the system can accurately assess the long-term risk level of symptom combinations. Particular attention is paid to the rate of symptom evolution and severity; when a rapidly worsening trend is detected, the system increases the risk weight of the corresponding symptom combination. For example, for elderly individuals experiencing both palpitations and shortness of breath, the system pays closer attention to the potential cardiovascular risks associated with this combination.

[0101] This embodiment innovatively implements a risk feature extraction mechanism. A feature extraction model is constructed based on a deep neural network, capturing local and global features in the risk matrix through multi-layer convolutional operations: Risk_Feature = CNN(Risk_Matrix), where Risk_Matrix is ​​the health risk matrix, and CNN represents the convolutional neural network operation. This deep learning framework can automatically learn risk features at different levels, providing comprehensive feature support for health risk scoring. Specifically, the system achieves feature fusion through fully connected layers to generate a comprehensive health risk score.

[0102] This embodiment deeply optimizes the time-series analysis strategy. The system constructs a time-series analysis model based on a Long Short-Term Memory (LSTM) network and uses a gating mechanism to selectively retain historical state information. It pays particular attention to the time-series change patterns of risk scores; when abnormal trends are detected, the system triggers more in-depth analysis. For example, for elderly individuals with persistently high blood pressure, the system combines changes in other physiological indicators to predict potential health risks. This time-series-based analysis method can effectively provide early warnings of potential health problems.

[0103] This embodiment achieves sequence labeling of risk trends using a Conditional Random Field (CRF) model. The system designs a trend feature extraction method based on a sliding time window, constructing a trend feature vector by analyzing changes in risk scores over continuous time periods. Particular attention is paid to the acceleration and persistence of risk changes, as these features directly reflect the evolution trend of health status. Through CRF sequence labeling, the system can accurately infer the development direction of health risks.

[0104] This innovative design not only solves the problem of inaccurate risk assessment in traditional methods but also establishes a continuously optimized health monitoring framework. Through the combination of deep learning and time-series analysis, the system can accurately predict the health risk trends of the elderly. This multi-dimensional analysis-based assessment mechanism ensures that the system maintains high predictive efficiency and reliable assessment results even when facing complex health conditions. In elderly care service practice, this intelligent risk assessment scheme significantly improves the foresight of health monitoring.

[0105] This embodiment upgrades health monitoring for the elderly population by establishing a complete risk assessment and trend prediction chain. The system can dynamically adjust assessment strategies based on real-time data, avoiding the limitations of traditional fixed models. Through multi-level risk analysis and trend prediction, the accuracy and predictability of health monitoring are significantly improved, providing reliable health early warning services for the elderly. This intelligent assessment mechanism demonstrates powerful predictive capabilities and practical value in elderly care service scenarios.

[0106] This embodiment not only improves the accuracy of health risk assessment but also establishes an ever-evolving early warning system through continuous data accumulation and model optimization. This deep learning-based assessment mechanism ensures that the system continuously improves with use, providing increasingly precise health monitoring services for the elderly. In practical applications, this self-optimization mechanism significantly enhances the system's long-term service quality and early warning effectiveness, providing strong technical support for intelligent elderly care services.

[0107] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0108] Step S601: Construct an early warning classification model, map the health risk score to a preset score range, set a risk level threshold based on medical expert experience, input the risk development trend into a trend prediction network, the trend prediction network uses a recurrent neural network structure to dynamically update the trend features, calculate the risk evolution speed and acceleration, classify the abnormal state based on the combination rules of risk score and evolution features, the combination rules give priority to emergency situations where the risk score exceeds the threshold, combine the evolution features to assess the persistence of risk, and divide the assessment results into three levels: emergency, severe, and general.

[0109] Step S602: Adopt a graded response mechanism to trigger differentiated early warning processes for different levels of abnormal states. Construct an emergency plan database for emergency-level abnormalities, which includes contact information and handling procedures. Establish a guardian feedback channel for severe-level abnormalities, which supports two-way information interaction. Set up a health management module for general-level abnormalities, which tracks and records the abnormal state in real time.

[0110] Optionally, this embodiment innovatively designs a multi-level early warning classification and response scheme to address the issue of timely detection and rapid response to abnormal health conditions in the elderly. This embodiment first constructs an early warning classification model, transforming health risk scores into a standard range using a risk mapping function: Risk_Level = f(Score, Trend_Features), where Score is the risk score and Trend_Features is the trend feature vector. The system sets classification thresholds based on the experience of medical experts and dynamically adjusts them in conjunction with risk evolution characteristics. In practical applications in nursing homes, this multi-dimensional risk assessment method can accurately identify health abnormalities of different degrees, achieving precise triggering of graded early warnings.

[0111] This embodiment deeply optimizes the trend prediction strategy. Considering the gradual changes in the health status of the elderly, a trend analysis module based on a recurrent neural network was designed. The system captures the changing patterns of risk indicators through time-series modeling, paying particular attention to the speed and acceleration characteristics of risk evolution. For example, when a sustained rise in blood pressure is detected with an increasing rate of increase, the system raises the warning level. This prediction method based on dynamic features can promptly identify potential health risks. Through the design of long short-term memory units, the system can effectively handle long-term dependencies and accurately predict risk development trends.

[0112] This embodiment innovatively implements a tiered mechanism for abnormal states. Based on a combined analysis of risk scores and evolutionary characteristics, the system designs multi-level tiered rules. When a risk score exceeds an emergency threshold, the system immediately triggers the highest-level warning. Simultaneously, the system assesses the persistence and development trend of the abnormal state by incorporating risk evolutionary characteristics. For example, when multiple physiological indicators of an elderly person are found to be abnormal simultaneously and showing a deteriorating trend, the system will raise the warning level even if a single indicator does not reach the emergency threshold. This comprehensive assessment mechanism can effectively prevent sudden deterioration of health conditions.

[0113] This embodiment deeply optimizes the tiered response strategy. The system designs differentiated handling procedures for different levels of abnormal states. For emergency-level abnormalities, a complete emergency plan library has been built, including detailed contact information and handling steps. Special attention is paid to the timeliness of the response; the system automatically notifies relevant personnel in a preset order. For severe-level abnormalities, a two-way interactive feedback mechanism has been established, enabling caregivers to promptly understand changes in the elderly person's condition and provide necessary care guidance. For general-level abnormalities, the system continuously tracks and records the occurrence patterns and evolution of the abnormalities through the health management module.

[0114] This innovative design not only solves the problems of delayed early warning and untimely response in traditional methods, but also establishes a continuously optimized health monitoring framework. Through the combination of multi-level early warning and coordinated response, the system can take appropriate measures based on the severity of the anomaly. This risk-level-based response mechanism ensures that the system maintains high-efficiency response capabilities and reliable handling effects when facing different types of health anomalies. In elderly care service practice, this intelligent early warning scheme significantly improves the effectiveness of health risk prevention and control.

[0115] This embodiment upgrades health monitoring for the elderly population by establishing a complete early warning and response chain. The system can dynamically adjust its response strategy based on real-time status, avoiding the limitations of traditional fixed procedures. Through multi-layered early warning mechanisms and handling processes, the accuracy and timeliness of health management are significantly improved, providing reliable health protection for the elderly. This intelligent monitoring mechanism demonstrates strong practical value and social benefits in elderly care service scenarios.

[0116] This embodiment not only improves the timeliness of health monitoring but also establishes an ever-evolving intelligent system through continuous data accumulation and process optimization. This tiered early warning-based monitoring mechanism ensures that the system continuously improves with use, providing increasingly precise health protection services for the elderly. In practical applications, this self-optimization mechanism significantly enhances the system's long-term service quality and early warning effectiveness, providing strong technical support for intelligent elderly care services.

[0117] In one embodiment of the intelligent companion machine interaction method based on physiological monitoring in this application, it may further include the following:

[0118] Step S701: Construct an early warning information distribution module to generate early warning information containing location information and symptom description for emergency-level anomalies. Based on a message queue, push the early warning information in parallel to the emergency center dispatch system, the guardian's mobile terminal, and the community medical management platform, and activate the linkage response mechanism. The linkage response mechanism automatically allocates treatment tasks based on a preset process. For severe-level anomalies, construct a status monitoring engine. The status monitoring engine continuously collects physiological data and extracts key indicators. Push abnormal fluctuation information of key indicators to the guardian and send a follow-up application to the community medical management platform. For general-level anomalies, activate a health management task. The health management task records the time of occurrence of the anomaly and the symptoms.

[0119] Step S702: Construct a data storage module to uniformly store the early warning information distribution record, the status monitoring data, and the health management record into an interactive data table. Index the data in the interactive data table according to the time dimension and the functional dimension. Filter invalid information based on data cleaning rules. Use an incremental update strategy to write the cleaned data into the user's health record. The user's health record adopts a distributed storage structure to realize data backup and disaster recovery.

[0120] Optionally, this embodiment innovatively designs a multi-level early warning distribution and data management scheme to address issues such as untimely early warning response and poor information transmission in cases of health abnormalities among the elderly. This embodiment first constructs an early warning information distribution module, obtaining the location information of the elderly through real-time location services. The system designs an early warning information generation mechanism: Alert_Message = Package(Location, Symptoms, Severity), where Location is the geographical location information, Symptoms are symptom descriptions, and Severity is the urgency level. In practical applications in nursing homes, this structured early warning information can help rescue personnel quickly understand the situation and respond.

[0121] This embodiment deeply optimizes the emergency warning distribution strategy. For life-threatening emergencies, a parallel push mechanism based on message queues is designed. The system employs a highly available message middleware to ensure reliable information transmission and monitors message delivery status through a heartbeat detection mechanism. Particular attention is paid to the timeliness of information; when a message push delay is detected, the system automatically activates a backup channel. For example, when an elderly person experiences severe arrhythmia, the system will simultaneously send warning information through multiple channels to ensure timely response from rescue forces. This multi-channel warning mechanism significantly improves response speed in emergency situations.

[0122] This embodiment innovatively implements a status monitoring engine. For severe anomalies, the system constructs a real-time monitoring framework to continuously collect and analyze key physiological indicators. Particular attention is paid to abnormal fluctuations in indicators; when an indicator exceeds the safe range, the system immediately generates an anomaly report. For example, for elderly individuals with abnormal blood pressure, the system monitors blood pressure trends in real time and promptly pushes information about abnormal fluctuations to their caregivers. Simultaneously, the system automatically sends follow-up requests to the community healthcare platform to ensure timely intervention by professional physicians.

[0123] This embodiment deeply optimizes the execution strategy for health management tasks. For general-level abnormalities, the system is designed with a detailed recording mechanism, including information such as the time of occurrence, duration, and symptom presentation. Particular attention is paid to the evolution of symptoms, constructing a complete disease development profile through continuous recording. This meticulous health management approach provides important reference for doctors' diagnoses. For example, when an elderly person experiences mild discomfort, the system will record the sequence of symptom appearance and changes in detail, helping doctors accurately determine the cause.

[0124] This embodiment achieves reliable data management through a distributed storage architecture. The system uses a multi-dimensional index structure to organize interactive data, enabling rapid retrieval by time and function dimensions. Special attention is paid to data quality control, filtering out duplicate, erroneous, or invalid information through cleaning rules. For example, when multiple sensors report anomalies simultaneously, the system uses data correlation analysis to remove redundant information, retaining only the most valuable data records.

[0125] This innovative design not only solves the problem of untimely early warning response in traditional methods but also establishes a sustainable and optimized health management framework. Through the combination of multi-level early warning and data management, the system can take appropriate measures based on the severity of the anomaly. This risk-level-based response mechanism ensures that the system maintains high-efficiency response capabilities and reliable handling effects when facing different types of health anomalies. In elderly care service practice, this intelligent early warning scheme significantly improves the effectiveness of health risk prevention and control.

[0126] This embodiment achieves an upgrade in health monitoring for the elderly population by establishing a complete early warning response and data management chain. The system can dynamically adjust response strategies based on real-time status, avoiding the limitations of traditional fixed procedures. Through multi-layered early warning mechanisms and data accumulation, the accuracy and foresight of health management are significantly improved, providing reliable health protection for the elderly. This intelligent monitoring mechanism demonstrates strong practical value and social benefits in elderly care service scenarios.

[0127] This embodiment not only improves the timeliness of early warning response but also establishes a continuously evolving health management system through ongoing data accumulation and analysis. This multi-level response-based management mechanism ensures that the system continuously improves with use, providing increasingly precise health monitoring services for the elderly. In practical applications, this self-optimization mechanism significantly enhances the system's long-term service quality and early warning effectiveness, providing strong technical support for intelligent elderly care services.

[0128] To effectively address the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, and to significantly improve the service quality and reliability of intelligent companionship, this application provides an embodiment of a physiological monitoring-based intelligent companionship machine interaction device for implementing all or part of the aforementioned physiological monitoring-based intelligent companionship machine interaction method. See [link to embodiment]. Figure 2 The intelligent companion robot interaction device based on physiological monitoring specifically includes the following components:

[0129] The physiological data acquisition module 10 is used to collect physiological data of the elderly. It collects heart rate and body temperature data through a smart bracelet, heartbeat waveform data through millimeter-wave radar, and sleep state data based on a smart pillow. The physiological data is segmented by a sliding window, a dynamic threshold model is constructed to calculate the data reliability score, abnormal data is corrected based on the reliability score, historical monitoring data is obtained from the user's health record to establish personalized baseline parameters, and the corrected data is normalized based on the personalized baseline parameters to obtain a physiological feature vector.

[0130] The symptom analysis module 20 is used to construct a symptom association analysis model, construct a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relation edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend.

[0131] The multi-level response module 30 is used to execute multi-level early warning responses. Based on the health risk score and the risk development trend, the abnormal state is divided into emergency, severe, and general levels. For emergency level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institutions, and a linkage response mechanism is triggered. For severe level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with the community doctor is scheduled. For general level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0132] As described above, the intelligent companion robot interaction device based on physiological monitoring provided in this application can optimize the quality of monitoring data by innovatively constructing a multi-source physiological data processing mechanism and integrating heart rate, body temperature, heartbeat waveform, and sleep state data. A symptom association analysis model based on a chronic disease atlas is designed, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, achieving coordinated response and accurate early warning for different risk levels through emergency, severe, and general classification processing. This method effectively solves the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0133] From a hardware perspective, in order to effectively address the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, and significantly improve the service quality and reliability of intelligent companionship, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned intelligent companionship machine interaction method based on physiological monitoring. The electronic device specifically includes the following components:

[0134] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the intelligent companion robot interaction device based on physiological monitoring and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the intelligent companion robot interaction method based on physiological monitoring and the embodiments of the intelligent companion robot interaction device based on physiological monitoring in the embodiments, the content of which is incorporated herein, and repeated details will not be described again.

[0135] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0136] In practical applications, some parts of the intelligent companion robot interaction method based on physiological monitoring can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific choice depends on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0137] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0138] Figure 3 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 3 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 3This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0139] In one embodiment, the intelligent companion machine interaction method based on physiological monitoring can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following controls:

[0140] Step S101: Collect physiological data of the elderly. Collect heart rate and body temperature data through a smart bracelet, collect heartbeat waveform data using millimeter-wave radar, and collect sleep state data based on a smart pillow. Perform sliding window segmentation on the physiological data, construct a dynamic threshold model to calculate the data reliability score, correct abnormal data based on the reliability score, obtain historical monitoring data from the user's health record to establish personalized baseline parameters, and normalize the corrected data based on the personalized baseline parameters to obtain physiological feature vectors.

[0141] Step S102: Construct a symptom association analysis model, construct a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relationship edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend;

[0142] Step S103: Execute a multi-level early warning response. Based on the health risk score and the risk development trend, the abnormal state is divided into emergency, severe, and general levels. For emergency level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institution, and a linkage response mechanism is triggered. For severe level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0143] As described above, the electronic device provided in this application innovatively constructs a multi-source physiological data processing mechanism, integrating heart rate, body temperature, heartbeat waveform, and sleep state data to optimize the quality of monitoring data. It designs a symptom association analysis model based on a chronic disease atlas, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, classifying risks into emergency, severe, and general levels to achieve coordinated responses and accurate early warnings for different risk levels. This method effectively addresses the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0144] In another embodiment, the intelligent companion robot interaction device based on physiological monitoring can be configured separately from the central processing unit 9100. For example, the intelligent companion robot interaction device based on physiological monitoring can be configured as a chip connected to the central processing unit 9100, and the function of the intelligent companion robot interaction method based on physiological monitoring can be realized through the control of the central processing unit.

[0145] like Figure 3 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 3 All components shown; in addition, the electronic device 9600 may also include Figure 3 For components not shown, please refer to existing technologies.

[0146] like Figure 3 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0147] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0148] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0149] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0150] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0151] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0152] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 (transmitter / receiver) is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0153] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the intelligent companion machine interaction method based on physiological monitoring, in which the execution subject is a server or client, as described in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the intelligent companion machine interaction method based on physiological monitoring, in which the execution subject is a server or client, as described in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0154] Step S101: Collect physiological data of the elderly. Collect heart rate and body temperature data through a smart bracelet, collect heartbeat waveform data using millimeter-wave radar, and collect sleep state data based on a smart pillow. Perform sliding window segmentation on the physiological data, construct a dynamic threshold model to calculate the data reliability score, correct abnormal data based on the reliability score, obtain historical monitoring data from the user's health record to establish personalized baseline parameters, and normalize the corrected data based on the personalized baseline parameters to obtain physiological feature vectors.

[0155] Step S102: Construct a symptom association analysis model, construct a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relationship edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend;

[0156] Step S103: Execute a multi-level early warning response. Based on the health risk score and the risk development trend, the abnormal state is divided into emergency, severe, and general levels. For emergency level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institution, and a linkage response mechanism is triggered. For severe level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0157] As described above, the computer-readable storage medium provided in this application innovatively constructs a multi-source physiological data processing mechanism, integrating heart rate, body temperature, heartbeat waveform, and sleep state data to optimize the quality of monitoring data. A symptom association analysis model based on a chronic disease atlas is designed, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, classifying risks into emergency, severe, and general levels to achieve coordinated responses and accurate early warnings for different risk levels. This method effectively addresses the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0158] Embodiments of this application also provide a computer program product capable of implementing all steps of the physiological monitoring-based intelligent companion machine interaction method in the above embodiments, where the execution subject is a server or a client. When executed by a processor, this computer program / instruction implements the steps of the physiological monitoring-based intelligent companion machine interaction method. For example, the computer program / instruction implements the following steps:

[0159] Step S101: Collect physiological data of the elderly. Collect heart rate and body temperature data through a smart bracelet, collect heartbeat waveform data using millimeter-wave radar, and collect sleep state data based on a smart pillow. Perform sliding window segmentation on the physiological data, construct a dynamic threshold model to calculate the data reliability score, correct abnormal data based on the reliability score, obtain historical monitoring data from the user's health record to establish personalized baseline parameters, and normalize the corrected data based on the personalized baseline parameters to obtain physiological feature vectors.

[0160] Step S102: Construct a symptom association analysis model, construct a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relationship edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend;

[0161] Step S103: Execute a multi-level early warning response. Based on the health risk score and the risk development trend, the abnormal state is divided into emergency, severe, and general levels. For emergency level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institution, and a linkage response mechanism is triggered. For severe level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

[0162] As described above, the computer program product provided in this application innovatively constructs a multi-source physiological data processing mechanism, integrating heart rate, body temperature, heartbeat waveform, and sleep state data to optimize the quality of monitoring data. It designs a symptom association analysis model based on a chronic disease atlas, combining graph convolutional networks and medical expert knowledge to establish a health risk assessment strategy for intelligent prediction. A multi-level early warning response mechanism is introduced, classifying risks into emergency, severe, and general levels to achieve coordinated responses and accurate early warnings for different risk levels. This method effectively addresses the shortcomings of traditional technologies in data processing, risk assessment, and early warning response, significantly improving the service quality and reliability of intelligent companionship.

[0163] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0164] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0165] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0166] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0167] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for intelligent companionship machine interaction based on physiological monitoring, characterized in that, The method includes: Physiological data of the elderly is collected by using a smart bracelet to collect heart rate and body temperature data, using millimeter-wave radar to collect heartbeat waveform data, and using a smart pillow to collect sleep state data. The physiological data is segmented by a sliding window, a dynamic threshold model is constructed to calculate the data reliability score, abnormal data is corrected based on the reliability score, historical monitoring data is obtained from the user's health record to establish personalized baseline parameters, and the corrected data is normalized based on the personalized baseline parameters to obtain physiological feature vectors. A symptom association analysis model is constructed. A chronic disease atlas is built based on medical expert knowledge. The symptom nodes and relation edges in the chronic disease atlas are input into a graph convolutional network to extract disease features. Symptom matching analysis is performed on the physiological feature vectors to obtain symptom combination features. A health risk matrix is ​​generated based on symptom correlation. The health risk matrix is ​​input into a prediction model to generate a health risk score. The health risk score is then time-series modeled to obtain the risk development trend. The system implements a multi-level early warning response, classifying abnormal states into emergency, severe, and general levels based on the health risk score and risk development trend. For emergency-level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institutions, triggering a linkage response mechanism. For severe-level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general-level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

2. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The collection of physiological data from the elderly involves using a smart bracelet to collect heart rate and body temperature data, millimeter-wave radar to collect heartbeat waveform data, and a smart pillow to collect sleep state data. The physiological data is then segmented using a sliding window, and a dynamic threshold model is constructed to calculate a data reliability score, including: The heart rate signal collected by the photoplethysmography sensor built into the smart bracelet and the body temperature signal collected by the temperature sensor are input into the data acquisition module. The data acquisition module performs bandpass filtering on the heart rate signal to remove baseline drift and power frequency interference to obtain heart rate data, and performs low-pass filtering on the body temperature signal to eliminate high-frequency noise to obtain body temperature data. The module uses millimeter-wave radar to transmit frequency-modulated continuous wave signals and receive echo signals. The echo signals are phase demodulated to extract Doppler frequency shift to obtain heartbeat waveform data. Pressure distribution data is collected based on the pressure sensor array built into the smart pillow. Body movement information and respiratory rhythm are extracted from the pressure distribution data to obtain sleep state data. The heart rate data, body temperature data, heartbeat waveform data, and sleep state data are segmented according to the sampling time. Each data segment is input into a dynamic threshold model. The dynamic threshold model calculates the mean and standard deviation based on the statistical characteristics of the data, constructs an adaptive threshold boundary by combining historical data distribution, judges the validity of the sampling points in the data segment, calculates the proportion of valid sampling points to obtain the data reliability score, and uses the data reliability score as a weighting coefficient to perform a weighted average of the data segments.

3. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The process of correcting abnormal data based on the reliability score involves obtaining historical monitoring data from user health records to establish personalized baseline parameters, and then normalizing the corrected data using these personalized baseline parameters to obtain a physiological feature vector, including: The data reliability score is compared with a preset threshold, and data segments with scores below the threshold are marked as abnormal data segments. Piecewise linear interpolation is used to correct the abnormal data segments. A Kalman filter is constructed to perform state estimation and smoothing on the corrected data. The smoothed data is recombined according to the time series to obtain the corrected data sequence. Recent physiological monitoring data of the target user is retrieved from the user's health record. The statistical parameters of the physiological monitoring data are calculated based on the time-weighted average method to obtain personalized baseline parameters. The personalized baseline parameters include the mean level and fluctuation range of physiological indicators. The mean level in the personalized baseline parameters is subtracted from the corrected data sequence to obtain the fluctuation amount. The fluctuation amount is normalized based on the fluctuation range in the personalized baseline parameters to obtain standardized features. A feature extraction network is constructed. The feature extraction network performs dimensionality reduction mapping on the standardized features based on a multilayer perceptron structure. The mapping result is numerically stabilized through a batch normalization layer. The processing result is converted into a physiological feature vector through an activation function.

4. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The aforementioned disease association analysis model constructs a chronic disease atlas based on medical expert knowledge. Symptom nodes and relationship edges from this atlas are input into a graph convolutional network to extract disease features. Disease matching analysis is then performed on the physiological feature vectors to obtain symptom combination features, including: A chronic disease atlas is constructed based on medical expert knowledge. Symptom descriptions and risk characteristics of common chronic diseases are used as node information, and the correlation between symptoms and disease evolution patterns are used as edge information. The node information is semantically encoded to generate node feature vectors. The weight coefficients of the edges are calculated based on the co-occurrence frequency to generate an adjacency matrix. The node feature vectors and the adjacency matrix are input into a graph convolutional network. The graph convolutional network propagates node information on the graph structure through a message passing mechanism. The contribution of different nodes is weighted using an attention mechanism, and local structural features and global semantic features are extracted to obtain disease features. The physiological feature vector is input into the disease matching module. The disease matching module calculates the matching degree between the physiological feature vector and the disease feature based on cosine similarity, constructs a symptom correlation matrix to perform correlation analysis on the matching results, uses a deep neural network to combine features of the symptom correlation matrix, and captures the dependency relationship between symptoms through a multi-head self-attention mechanism to generate symptom combination features.

5. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The process involves generating a health risk matrix based on symptom correlation, inputting the health risk matrix into a prediction model to generate a health risk score, and performing time-series modeling on the health risk score to obtain the risk development trend, including: Based on the symptom combination features, the correlation strength between different symptoms is calculated, a symptom transition probability matrix is ​​constructed to characterize the symptom evolution pattern, the symptom transition probability matrix is ​​multiplied by the symptom risk weight to obtain an initial risk matrix, the initial risk matrix is ​​iteratively updated using a Markov chain model, the risk cumulative effect of different symptom combinations is calculated based on the steady-state distribution, the risk cumulative effect is corrected by combining an expert rule base to obtain a health risk matrix, the health risk matrix is ​​input into a deep neural network, the deep neural network extracts risk features through multiple convolutions, and the features are fused through a fully connected layer to generate a health risk score; A temporal analysis network is constructed, which performs sequence modeling of the health risk score based on long short-term memory units. It selectively retains historical state information through a gating mechanism and captures temporal dependencies by combining an attention mechanism. It predicts the changing trend of the health risk score and calculates trend features using a sliding time window. The trend features are then input into a conditional random field model, which infers the risk development trend based on sequence labeling methods.

6. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The implementation of a multi-level early warning response classifies abnormal states into emergency, severe, and general levels based on the health risk score and the risk development trend, including: A warning classification model is constructed, which maps the health risk score to a preset score range, sets a risk level threshold based on the experience of medical experts, inputs the risk development trend into a trend prediction network, and uses a recurrent neural network structure to dynamically update the trend features, calculates the risk evolution speed and acceleration, classifies the abnormal state based on the combination rules of risk score and evolution features, the combination rules give priority to emergency situations where the risk score exceeds the threshold, and assesses the persistence of risk in combination with evolution features, and divides the assessment results into three levels: emergency, severe, and general. A tiered response mechanism is adopted, triggering differentiated early warning processes for different levels of abnormal states. An emergency plan database is built for emergency-level abnormalities, which includes contact information and handling procedures. A guardian feedback channel is established for severe-level abnormalities, which supports two-way information interaction. A health management module is set up for general-level abnormalities, which tracks and records the abnormal state in real time.

7. The intelligent companion machine interaction method based on physiological monitoring according to claim 1, characterized in that, The system simultaneously sends early warning information to the emergency center, guardian, and community medical institutions for emergency-level anomalies and triggers a coordinated response mechanism. For severe-level anomalies, it pushes real-time status information to the guardian and schedules follow-up visits with community doctors. For general-level anomalies, it records symptom descriptions and periodically pushes health analysis reports to the guardian. The interactive data during the response process is stored in the user's health record, including: An early warning information distribution module is constructed to generate early warning information containing location information and symptom descriptions for emergency-level anomalies. Based on a message queue, the early warning information is pushed in parallel to the emergency center dispatch system, the guardian's mobile terminal, and the community medical management platform, and a linkage response mechanism is activated. The linkage response mechanism automatically allocates treatment tasks based on a preset process. For severe-level anomalies, a status monitoring engine is constructed. The status monitoring engine continuously collects physiological data and extracts key indicators. The abnormal fluctuation information of key indicators is pushed to the guardian, and a follow-up application is sent to the community medical management platform. For general-level anomalies, a health management task is activated. The health management task records the time of occurrence of the anomaly and the symptom manifestation. A data storage module is constructed to uniformly store the early warning information distribution records, the status monitoring data, and the health management records into an interactive data table. The data in the interactive data table is indexed according to the time and function dimensions. Invalid information is filtered based on data cleaning rules, and the cleaned data is written into the user's health record using an incremental update strategy. The user's health record adopts a distributed storage structure to achieve data backup and disaster recovery.

8. An intelligent companion robot interaction device based on physiological monitoring, characterized in that, The device includes: The physiological data acquisition module is used to collect physiological data of the elderly. It collects heart rate and body temperature data through a smart bracelet, heartbeat waveform data through millimeter-wave radar, and sleep state data based on a smart pillow. The physiological data is segmented by a sliding window, a dynamic threshold model is constructed to calculate the data reliability score, abnormal data is corrected based on the reliability score, historical monitoring data is obtained from the user's health record to establish personalized baseline parameters, and the corrected data is normalized based on the personalized baseline parameters to obtain physiological feature vectors. The symptom analysis module is used to construct a symptom association analysis model, build a chronic disease atlas based on medical expert knowledge, input the symptom nodes and relation edges in the chronic disease atlas into a graph convolutional network to extract disease features, perform symptom matching analysis on the physiological feature vectors to obtain symptom combination features, calculate and generate a health risk matrix based on symptom correlation, input the health risk matrix into a prediction model to generate a health risk score, and perform time series modeling on the health risk score to obtain the risk development trend. The multi-level response module is used to execute multi-level early warning responses. Based on the health risk score and the risk development trend, abnormal states are divided into emergency, severe, and general levels. For emergency-level abnormalities, early warning information is sent to the emergency center, guardian, and community medical institutions, and a linkage response mechanism is triggered. For severe-level abnormalities, real-time status information is pushed to the guardian and a follow-up appointment with a community doctor is scheduled. For general-level abnormalities, symptom description information is recorded and health analysis reports are pushed to the guardian regularly. The interactive data during the response process is stored in the user's health record.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the intelligent companion machine interaction method based on physiological monitoring as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the intelligent companion machine interaction method based on physiological monitoring as described in any one of claims 1 to 7.

Images