Home-based care intelligent monitoring system and method based on artificial intelligence

Abstract

The invention discloses a home-based care intelligent monitoring system and method based on artificial intelligence, and relates to the technical field of artificial intelligence, and the method comprises the following steps: monitoring the physiological and behavior data of a monitored person through a sensor; performing cleaning preprocessing, normalization processing and synchronous integration on the acquired data; analyzing behavior data by using a model to identify abnormity, constructing an annotation data set, constructing a regression model for judging the state of the current monitored person through a neural network, and performing confidence evaluation to obtain confidence; and calculating a quantitative index of the transmission credibility, and adjusting a data transmission interval according to the calculated confidence and the quantitative index of the transmission credibility.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and specifically to an intelligent home care system and method based on artificial intelligence. Background Art

[0002] With the continuous deepening of the aging of the population and the rapid development of technology, the home care model has gradually become an important way of providing for the elderly. The wide application of smart home technology, sensor technology, and artificial intelligence technology provides strong support for realizing high-quality intelligent home care. In the home care scenario, using various intelligent devices to comprehensively monitor the elderly under guardianship can not only reduce the care burden on family members, but also improve the quality of life of the elderly and ensure their life, health, and safety.

[0003] However, the existing home care monitoring means face many problems in practical applications: First, the monitoring of the physiological and behavioral problems of the ward is missing, and potential hazards that may exist in the elderly cannot be detected in a timely manner; Second, during the monitoring process, the monitoring personnel cannot always successfully receive the monitoring reports of the monitored personnel, and the stability and reliability of data transmission are insufficient, resulting in the timeliness and integrity of monitoring information not being guaranteed. Once a health problem occurs to the monitored person, the treatment opportunity may be delayed. Therefore, to solve the above problems, there is an urgent need for an intelligent home care system and method based on artificial intelligence to realize the comprehensive monitoring of the physiology and behavior of the monitored personnel and provide more reliable guarantee for home care. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent home care system and method based on artificial intelligence to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An intelligent home care method based on artificial intelligence, the method comprising the following steps: Monitoring the physiological and behavioral data of the monitored person through sensors; Performing cleaning preprocessing, normalization processing, and synchronous integration on the obtained data; Using a model to analyze the behavioral data to identify anomalies, constructing an annotated data set, constructing a regression model for judging the state of the currently monitored person through a neural network, and performing confidence evaluation to obtain a confidence level; Calculating a quantization index of transmission credibility, and adjusting the data transmission interval according to the calculated confidence level and the quantization index of transmission credibility.

[0006] Monitoring the physiological, behavioral, conversation content, and facial expression data of the monitored person through sensors, the specific steps comprising: Obtain the physiological, behavioral, conversation content, and facial expression data of the person under guardianship through sensors. Among them, the physiological data includes the heart rate, blood pressure, and body temperature data of the person under guardianship; the behavioral data includes the activity trajectory, body posture change of the person under guardianship, and the interaction information between the person under guardianship and the environment; Obtain the heart rate, blood pressure, and body temperature data of the person under guardianship through wearable devices and home medical monitoring devices equipped with sensors; Use indoor positioning sensors, inertial sensors, and camera devices to obtain the activity trajectory and body posture change of the person under guardianship; Collect the interaction information between the person under guardianship and the environment through smart home sensors and camera image recognition technology; Obtain the conversation content between the person under guardianship and the smart speaker, including daily conversation content, intonation, speaking duration, and conversation frequency data; Collect the facial expressions of the person under guardianship using a smart camera. The smart camera is installed in a public area. On the premise of respecting the privacy of the person under guardianship, identify the facial expressions of the person under guardianship through image recognition technology and deep learning algorithms; Perform synchronous acquisition of the physiological, behavioral, conversation content, and facial expression data of the person under guardianship at regular time intervals, sort the obtained data in chronological order [t1, t2,..., tw], and form a time series of the physiological, behavioral, conversation content, and facial expression data of the person under guardianship. t1, t2,..., tw respectively represent the data acquisition times of the 1st, 2nd,..., wth times.

[0007] Clean, preprocess, normalize, and synchronously integrate the collected physiological, behavioral, conversation content, and facial expression data of the person under guardianship. The specific steps include: Clean and preprocess the collected physiological, behavioral, conversation content, and facial expression data of the person under guardianship, remove noise data. When the data value at a certain moment is missing, use interpolation based on the time series to fill it according to the data values at the adjacent previous and next moments, and standardize the data format. Store all data according to a unified database table structure; Use the Min - Max normalization algorithm to process the physiological data of the person under guardianship with different magnitudes, and map the data uniformly to the [0, 1] interval; Synchronously integrate the physiological, behavioral, conversation content, and facial expression data of the person under guardianship in the order of acquisition time. Based on the synchronous acquisition time points [t1, t2,..., tw], associate and combine the heart rate, blood pressure, body temperature, activity trajectory, body posture change, interaction information between the person under guardianship and the environment, conversation content, and facial expression data at the same moment to form a complete time series data record.

[0008] Analyze behavioral data through a Transformer machine learning model to identify abnormal behaviors, construct an annotated dataset, adjust thresholds according to cultural habits and disease characteristics to complete personalized settings, construct a regression model for judging the status of the current person under guardianship, and send the status of the person under guardianship to the guardian. The specific steps include: Based on the preprocessed data, analyze and model the behavioral data of the person under guardianship through a Transformer machine learning model. Through feature extraction and pattern recognition, starting from the time series and action pattern dimensions of the behavioral data, identify the abnormal behaviors of the person under guardianship, and jointly form a set of abnormal behaviors with all the identified abnormal behaviors; Construct an annotated dataset for distinguishing specific behavior patterns and abnormal action representations; For behavior types corresponding to specific cultural backgrounds, social customs, or personal long-term formed behavior habits, extract and describe features from the aspects of the periodicity, duration, and amplitude of actions. Based on the analysis of normal samples, summarize the characteristic patterns of these behaviors in the normal state; For abnormal action types associated with diseases, characterize features from the dimensions of the frequency of action occurrence, the regularity of action presentation, and the limb parts involved. Through comparative analysis with normal behavioral actions, determine the abnormal characteristic patterns of these actions; Set personalized thresholds based on the user's historical data. For behavior types corresponding to specific cultural backgrounds, social customs, or personal long-term formed behavior habits, adjust the described feature thresholds to specific multiples of the initial values in the behavior pattern knowledge base, where the behavior pattern knowledge base is constructed based on historical behavioral data; For abnormal action types associated with diseases, adjust the described feature thresholds to specific multiples of the initial values in the behavior pattern knowledge base; Construct a regression model for judging the current status of the person under guardianship through a neural network framework that fuses multi-source features. Process the real-time obtained physiological, behavioral, conversation content, and facial expression data of the person under guardianship, as well as the set alarm threshold, using the embedded coding method, and input the encoded data into the constructed neural network regression model that integrates a fully connected layer and a convolutional layer. Utilize the convolutional layer to mine the associations in the local features of the data, use the fully connected layer to integrate the global features, and output the health index score of the current person under guardianship through learning and mapping the association patterns between the input data features, so as to judge the current status of the person under guardianship; Adopt a cross-validation method to evaluate the confidence of the health index score output by the regression model for judging the current status of the person under guardianship, and obtain the confidence Z; Send the health index score of the current person under guardianship to the guardian.

[0009] Calculate the quantitative index of transmission credibility, and adjust the data transmission interval according to the calculated confidence level and the quantitative index of transmission credibility. The specific steps include: Use a graph neural network (GNN) that fuses spatio-temporal features to predict the influencing factors during the transmission of the health index score of the person under guardianship to the guardian. Encode multi-source data related to transmission in the form of a graph structure and input it into the constructed GNN model. The model learns the spatial associations and temporal evolution patterns in the data and outputs the predicted values of the influencing factors during the transmission of the health index score of the person under guardianship. Use the predicted values of the influencing factors during the transmission of the health index score of the person under guardianship as the evaluation factor set U = {u1, u2,..., un}, and construct the comment set V = {v1, v2,..., vm} of transmission credibility. Here, u1, u2,..., un represent the predicted values of the 1st, 2nd,..., nth influencing factors during the transmission of the health index score of the person under guardianship, and v1, v2,..., vm represent the 1st, 2nd,..., mth level divisions of transmission credibility. Based on the statistical analysis of historical data, determine the membership degree relationship between factor ui and each comment in the comment set V, and establish a fuzzy relationship matrix R. Specifically, through the statistical analysis of historical data, the membership degrees of factor u1 to each comment are r11, r12,..., r1m respectively, the membership degrees of factor u2 to each comment are r21, r22,..., r2m respectively, and so on, to obtain the membership degrees of factors u3,..., un to each comment. Based on the membership degree relationship between the evaluation factor set and each comment in the comment set V, establish a fuzzy relationship matrix R. The fuzzy relationship matrix R is expressed as: ; Determine the relative importance of each influencing factor in evaluating transmission credibility through the Delphi method, that is, determine the factor weight vector, and denote the factor weight vector as A = (a1, a2,..., an), where a1, a2,..., an represent the influence weights of the 1st, 2nd,..., nth factors on transmission credibility. Perform a synthesis operation on the factor weight vector A and the fuzzy relationship matrix R through weighted operation to obtain the fuzzy comprehensive evaluation result vector B. The calculation formula is as follows: B = A * R = (b1, b2,..., bm); Among them, , j = 1, 2,..., m, and b1, b2,..., bm represent the degree values of the transmission credibility being judged as levels v1, v2,..., vm after performing a synthesis operation on the factor weight vector A and the fuzzy relationship matrix R through weighted operation. Assign corresponding scores to each comment in the comment set of transmission credibility, calculate the fuzzy comprehensive evaluation result vector B, and take the weighted average of the corresponding scores assigned to each comment in the comment set of transmission credibility as the quantization index Q of transmission credibility; According to the calculated confidence level Z and the quantization index Q of transmission credibility, readjust the time interval for transmitting the health index score of the person under guardianship to the guardian. The time interval adjustment formula is as follows: T = T0 + β * Z * Q; where β represents the adjustment coefficient and T0 represents the initially set time interval.

[0010] An intelligent home care system based on artificial intelligence, the system includes a data acquisition module, a data processing module, an anomaly analysis and status evaluation module, a transmission credibility evaluation module, and a transmission control module. The data acquisition module is used to monitor the physiological and behavioral data of the person under guardianship through sensors; the data processing module is used to perform cleaning preprocessing, normalization processing, and synchronization integration on the acquired data; the anomaly analysis and status evaluation module is used to analyze the behavioral data using a model to identify anomalies, construct an annotated data set, build a regression model for judging the current status of the person under guardianship through a neural network, and perform confidence evaluation to obtain the confidence level; the transmission credibility evaluation module is used to calculate the quantization index of transmission credibility; the transmission control module adjusts the data transmission interval according to the calculated confidence level and the quantization index of transmission credibility.

[0011] The data acquisition module includes a physiological data collection unit and a behavioral data collection unit. The physiological data collection unit is used to obtain the heart rate, blood pressure, and body temperature data of the person under guardianship through sensors. The behavioral data collection unit is used to obtain the activity trajectory, body posture change situation, and the interaction information between the person under guardianship and the environment through sensors.

[0012] The data processing module includes a preprocessing unit, a normalization processing unit, and a synchronization integration unit. The preprocessing unit is used to remove the noise data in the collected physiological and behavioral data, fill in the missing data using interpolation based on time series, standardize the data format, and store it according to the unified database table structure; the normalization processing unit is used to process the physiological data of the person under guardianship with different magnitudes using the Min - Max normalization algorithm and map the data uniformly to the [0, 1] interval; the synchronization integration unit is used to associate and combine the heart rate, blood pressure, body temperature, activity trajectory, body posture change situation, and interaction information with the environment at the same moment in the order of collection time, based on the synchronous collection time points, to form a complete time series data record.

[0013] The abnormal analysis and status evaluation module includes an abnormal behavior recognition unit, a regression model construction unit, and a confidence evaluation unit. The abnormal behavior recognition unit is used to extract features and identify patterns from the time series and action pattern dimensions of behavior data based on the preprocessed data using a Transformer machine learning model, find the abnormal behaviors of the monitored person, and form a set of abnormal behaviors. The regression model construction unit is used to perform embedding encoding processing on the physiological and behavioral data of the monitored person obtained in real time and the set alarm threshold through a neural network framework that fuses multi-source features, input them into a neural network regression model that combines a fully connected layer and a convolutional layer, use the convolutional layer to mine the local feature correlations of the data, and the fully connected layer to integrate the global features, learn the correlation patterns between the input data features, output the health index score of the monitored person, and judge their current status. The confidence evaluation unit is used to evaluate the confidence of the health index score output by the regression model using cross-validation to obtain the confidence level.

[0014] The transmission credibility evaluation module includes an influencing factor prediction unit, a fuzzy relationship construction unit, and a comprehensive evaluation unit. The influencing factor prediction unit is used to use a graph neural network GNN that fuses spatio-temporal features to encode multi-source data related to transmission in the form of a graph structure and input it into the model, learn the spatial correlations and temporal evolution patterns in the data, and output the predicted values of the influencing factors during the transmission of the health index score of the monitored person. The fuzzy relationship construction unit is used to use the predicted values of the influencing factors as the evaluation factor set, construct a comment set for the transmission credibility, determine the membership relationship between the factors and each comment in the comment set based on historical data statistical analysis, and establish a fuzzy relationship matrix. The comprehensive evaluation unit is used to perform a synthesis operation on the factor weight vector and the fuzzy relationship matrix through weighted operations to obtain a fuzzy comprehensive evaluation result vector, assign corresponding scores to each comment in the transmission credibility comment set, and calculate the weighted average of the fuzzy comprehensive evaluation result vector and the corresponding scores of the comment set as the quantitative index of the transmission credibility.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. Using the graph neural network GNN to predict the influencing factors of data transmission, combining the fuzzy comprehensive evaluation method to calculate the quantitative index of transmission credibility, and dynamically adjusting the data transmission interval according to the confidence level and this index to ensure the transmission of monitoring data to the caregiver; 2. Through artificial intelligence technology, comprehensively analyze the physiological and behavioral data of the monitored person, and combine historical data and personalized threshold setting to identify abnormal behaviors and potential health risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of an intelligent home care method based on artificial intelligence of the present invention; Figure 2Schematic diagram of a smart home care intelligent monitoring system based on artificial intelligence according to the present invention. Specific implementation manner

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] In the embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution, a smart home care intelligent monitoring method based on artificial intelligence, and the method includes the following steps: Monitoring the physiological and behavioral data of the monitored person through sensors; Performing cleaning preprocessing, normalization processing, and synchronization integration on the acquired data; Using a model to analyze the behavioral data to identify abnormalities, constructing an annotated data set, constructing a regression model for judging the state of the currently monitored person through a neural network, and performing confidence evaluation to obtain the confidence; Calculating a quantization index of transmission credibility, and adjusting the data transmission interval according to the calculated confidence and the quantization index of transmission credibility.

[0019] Monitoring the physiological, behavioral, conversation content, and facial expression data of the monitored person through sensors, and the specific steps include: Obtaining the physiological, behavioral, conversation content, and facial expression data of the monitored person through sensors, wherein the physiological data includes the heart rate, blood pressure, and body temperature data of the monitored person; the behavioral data includes the activity trajectory, body posture change situation of the monitored person, and the interaction information between the monitored person and the environment; Obtaining the heart rate, blood pressure, and body temperature data of the monitored person through wearable devices and household medical monitoring devices equipped with sensors; Using indoor positioning sensors, inertial sensors, and camera devices to obtain the activity trajectory and body posture change situation of the monitored person; Collecting the interaction information between the monitored person and the environment through smart home sensors and camera image recognition technology; Obtaining the conversation content between the monitored person and the smart speaker, including daily conversation content, voice intonation, speaking duration, and conversation frequency data; Collecting the facial expression of the monitored person by using a smart camera, the smart camera is installed in a public area, and on the premise of respecting the privacy of the monitored person, identifying the facial expression of the monitored person through image recognition technology and deep learning algorithms; Synchronously collect the physiological, behavioral, conversation content, and facial expression data of the monitored person at regular time intervals, sort the obtained data in chronological order [t1, t2,..., tw] to form a time series of the physiological, behavioral, conversation content, and facial expression data of the monitored person, where t1, t2,..., tw represent the data collection times of the 1st, 2nd,..., wth times respectively; Specifically, collect the physiological data of an elderly person through a smart bracelet and a home medical device of the sensor. During the period from 9 am to 10 am on a certain day, collect data every 5 minutes. At 9 am, the heart rate is 75 beats per minute, the blood pressure is 120 / 80 mmHg, and the body temperature is 36.5 °C. At the same time, use the positioning sensor, inertial sensor, and camera installed indoors to obtain behavioral data. At 9:10 am, the elderly person walks from the bedroom to the living room, with an upright walking body posture. The interaction information with the environment shows that the TV in the living room is turned on. Record the conversation content of the elderly person through a smart speaker. During this hour, the elderly person has 5 conversations with the smart speaker, with an average conversation duration of 2 minutes each, and the voice intonation is stable. After the facial expression is recognized by the smart camera, the inferred state is calm.

[0020] Clean, preprocess, normalize, and synchronously integrate the physiological, behavioral, conversation content, and facial expression data of the monitored person collected. The specific steps include: Clean and preprocess the physiological, behavioral, conversation content, and facial expression data of the monitored person collected, remove noise data. When the data value at a certain moment is missing, use the interpolation method based on the time series to fill it according to the data values at the adjacent previous and next moments, and standardize the data format. Store all the data according to the unified database table structure; Use the Min - Max normalization algorithm to process the physiological data of the monitored person with different magnitudes, and map the data uniformly to the interval [0, 1]; Synchronously integrate the physiological, behavioral, conversation content, and facial expression data of the monitored person in the order of collection time. Based on the synchronously collected time points [t1, t2,..., tw], associate and combine the heart rate, blood pressure, body temperature, activity trajectory, body posture change, interaction information between the monitored person and the environment, conversation content, and facial expression data at the same moment to form a complete time series data record.

[0021] Analyze the behavioral data through the Transformer machine learning model to identify abnormal behaviors, construct an annotated data set, complete personalized settings by adjusting the threshold according to cultural habits and disease characteristics, construct a regression model for judging the current state of the monitored person, and send the state of the monitored person to the caregiver. The specific steps include: Based on the preprocessed data, the behavior data of the person under guardianship is analyzed and modeled through a Transformer machine learning model. Through feature extraction and pattern recognition, starting from the time series and action pattern dimensions of the behavior data, the abnormal behaviors of the person under guardianship are identified, and all the identified abnormal behaviors together form an abnormal behavior set; Construct an annotation data set for distinguishing specific behavior patterns and abnormal action representations; For the behavior types corresponding to specific cultural backgrounds, social customs, or personal long-term formed behavior habits, feature extraction and description are carried out from the aspects of the periodicity, duration, and amplitude of actions. Based on the analysis of normal samples, the characteristic patterns of these behaviors in the normal state are summarized; For the abnormal action types associated with diseases, feature characterization is carried out from the dimensions of the frequency of action occurrence, the regularity of action presentation, and the body parts involved. Through comparative analysis with normal behavior actions, the abnormal characteristic patterns of these actions are determined; According to the user's historical data, personalized threshold settings are made. For the behavior types corresponding to specific cultural backgrounds, social customs, or personal long-term formed behavior habits, the described feature thresholds are adjusted to specific multiples of the initial values in the behavior pattern knowledge base, where the behavior pattern knowledge base is constructed based on historical behavior data; For the abnormal action types associated with diseases, the described feature thresholds are adjusted to specific multiples of the initial values in the behavior pattern knowledge base; A regression model for judging the current state of the person under guardianship is constructed through a neural network framework that fuses multi-source features. The physiological, behavioral, conversation content, and facial expression data of the person under guardianship obtained in real time, as well as the set alarm threshold, are processed using an embedded coding method. The encoded data is input into the constructed neural network regression model that integrates a fully connected layer and a convolutional layer. By means of the convolutional layer, the associations in the local features of the data are mined, and the global features are integrated using the fully connected layer. Through the learning and mapping of the association patterns between the input data features, the health index score of the current person under guardianship is output to judge the current state of the person under guardianship; Using the cross-validation method, the confidence level of the health index score output by the regression model for judging the current state of the person under guardianship is evaluated to obtain the confidence level Z; Send the health index score of the current person under guardianship to the guardian; Specifically, a Transformer machine learning model is used to analyze behavioral data. Based on the fact that the elderly usually engage in simple indoor activities, such as reading or watching TV, with regular movement patterns from 9 am to 10 am every day. At 9:40 am, the model recognizes that the elderly person remains abnormally stationary in the living room for a long time and does not conform to their daily behavior pattern. This behavior is determined to be an abnormal behavior, and a labeled data set is constructed. For the morning exercise behavior that the elderly are accustomed to doing at 9:30 am every day, features are extracted from the aspects of the periodicity, duration, and amplitude of the movements to determine the characteristic pattern in the normal state. A personalized threshold is set. According to the initial threshold of the morning exercise movement amplitude in the behavior pattern knowledge base being 5 (unit: degree), and based on the elderly person's historical data, it is adjusted to 8 degrees. A regression model is constructed through a neural network framework that fuses multi-source features. The real-time physiological, behavioral, conversation content, and facial expression data, as well as the set alarm threshold, are input into the model after being embedded and encoded. After being processed according to the input data, the health index score output by the model is 70 (out of 100). The confidence level of the health index score is evaluated using the cross-validation method. Suppose the obtained confidence level Z is 0.8.

[0022] Calculate the quantization index of transmission credibility, and adjust the data transmission interval according to the calculated confidence level and the quantization index of transmission credibility. The specific steps include: Predict the influencing factors during the transmission of the health index score of the person under guardianship to the guardian through a graph neural network GNN that fuses spatio-temporal features. The multi-source data related to transmission is encoded and input into the constructed GNN model in the form of a graph structure. The model outputs the predicted values of the influencing factors during the transmission of the health index score of the person under guardianship by learning the spatial association and temporal evolution patterns in the data. Based on the predicted values of the influencing factors during the transmission of the health index score of the person under guardianship as the evaluation factor set U = {u1, u2,..., un}, construct the comment set V = {v1, v2,..., vm} of transmission credibility. Among them, u1, u2,..., un represent the predicted values of the 1st, 2nd,..., nth influencing factors during the transmission of the health index score of the person under guardianship, and among them, v1, v2,..., vm represent the 1st, 2nd,..., mth level divisions of transmission credibility. According to the statistical analysis of historical data, determine the membership relationship between factor ui and each comment in the comment set V, and establish a fuzzy relationship matrix R. Specifically, through the statistical analysis of historical data, the membership degrees of factor u1 to each comment are respectively r11, r12,..., r1m, the membership degrees of factor u2 to each comment are respectively r21, r22,..., r2m, and so on, to obtain the membership degrees of factors u3,..., un to each comment. According to the membership degree relationship between the evaluation factor set and each comment in the comment set V, establish the fuzzy relation matrix R; The fuzzy relation matrix R is expressed as: ; Determine the relative importance of each influencing factor in evaluating the transmission credibility through the Delphi method, that is, determine the factor weight vector, and denote the factor weight vector as A=(a1,a2,...,an), where a1,a2,...,an represent the influence weights of the 1st, 2nd,..., nth factors on the transmission credibility; Perform a synthesis operation on the factor weight vector A and the fuzzy relation matrix R through weighted operation to obtain the fuzzy comprehensive evaluation result vector B. The calculation formula is as follows: B = A * R = (b1,b2,...,bm); Among them, , j = 1,2,...,m, and b1,b2,...,bm represent the degree values of the transmission credibility being judged as levels v1,v2,...,vm after performing a synthesis operation on the factor weight vector A and the fuzzy relation matrix R through weighted operation; Assign corresponding scores to each comment in the comment set of the transmission credibility, and calculate the weighted average of the fuzzy comprehensive evaluation result vector B and the corresponding scores assigned to each comment in the comment set of the transmission credibility as the quantization index Q of the transmission credibility; According to the calculated confidence level Z and the quantization index Q of the transmission credibility, readjust the time interval for transmitting the health index score of the person under guardianship to the guardian. The time interval adjustment formula is as follows: T = T0 + β * Z * Q; where β represents the adjustment coefficient, and T0 represents the initially set time interval; Specifically, predict the data transmission influencing factors through the graph neural network GNN that fuses spatio-temporal features. Assume that the influencing factors include network signal strength, device load, data traffic, etc. Take the predicted values of these factors as the evaluation factor set U, and construct the comment set V of the transmission credibility as {high credibility, medium credibility, low credibility}. According to the statistical analysis of historical data, determine the membership degree relationship between the factors and the comment set. The membership degrees of the network signal strength factor u1 to high credibility, medium credibility, and low credibility are 0.6, 0.3, and 0.1 respectively, and establish the fuzzy relation matrix R; Determine the factor weight vector A=(0.4,0.3,0.3) through the Delphi method, and perform a synthesis operation on the factor weight vector A and the fuzzy relation matrix R through weighted operation to obtain the fuzzy comprehensive evaluation result vector B: B = A * R = (0.52,0.35,0.13). Assign corresponding scores to each comment in the comment set of the transmission credibility, 90 for high credibility, 70 for medium credibility, and 50 for low credibility. Calculate the weighted average of the fuzzy comprehensive evaluation result vector B and the corresponding scores of the comment set as the quantization index Q of the transmission credibility. After calculation, Q = 78.4; Transmission control phase: According to the initially set time interval T0 of 10 minutes and the adjustment coefficient β of 0.5, using the time interval adjustment formula, the calculated time interval for transmitting the health index score of the person under guardianship to the guardian after adjustment is approximately 41 minutes.

[0023] An intelligent home care intelligent monitoring system based on artificial intelligence. The system includes a data acquisition module, a data processing module, an anomaly analysis and status evaluation module, a transmission credibility evaluation module, and a transmission control module. The data acquisition module is used to monitor the physiological and behavioral data of the person under guardianship through sensors; the data processing module is used to perform cleaning preprocessing, normalization processing, and synchronous integration on the acquired data; the anomaly analysis and status evaluation module is used to analyze the behavioral data using a model to identify anomalies, construct an annotated data set, build a regression model for judging the current status of the person under guardianship through a neural network, and perform confidence evaluation to obtain the confidence level; the transmission credibility evaluation module is used to calculate the quantitative indicators of transmission credibility; the transmission control module adjusts the data transmission interval according to the calculated confidence level and the quantitative indicators of transmission credibility.

[0024] The data acquisition module includes a physiological data acquisition unit and a behavioral data acquisition unit. The physiological data acquisition unit is used to obtain the heart rate, blood pressure, and body temperature data of the person under guardianship through sensors. The behavioral data acquisition unit is used to obtain the activity trajectory, body posture change situation, and the interaction information between the person under guardianship and the environment through sensors.

[0025] The data processing module includes a preprocessing unit, a normalization processing unit, and a synchronous integration unit. The preprocessing unit is used to remove the noise data in the collected physiological and behavioral data, fill in the missing data using the interpolation method based on time series, standardize the data format, and store it according to the unified database table structure; the normalization processing unit is used to process the physiological data of the person under guardianship with different magnitudes using the Min-Max normalization algorithm, and map the data to the [0,1] interval uniformly; the synchronous integration unit is used to associate and combine the heart rate, blood pressure, body temperature, activity trajectory, body posture change situation, and interaction information with the environment at the same moment according to the acquisition time sequence and based on the synchronous acquisition time point, to form a complete time series data record.

[0026] The abnormal analysis and status evaluation module includes an abnormal behavior recognition unit, a regression model construction unit, and a confidence evaluation unit. The abnormal behavior recognition unit is used to extract features and identify patterns from the time series and action pattern dimensions of behavioral data based on the preprocessed data using a Transformer machine learning model, find the abnormal behaviors of the monitored person, and form a set of abnormal behaviors. The regression model construction unit is used to perform embedding encoding processing on the physiological and behavioral data of the monitored person obtained in real time and the set alarm threshold through a neural network framework that fuses multi-source features, input them into a neural network regression model that integrates a fully connected layer and a convolutional layer, use the convolutional layer to mine the local feature correlations of the data, and the fully connected layer to integrate the global features, learn the correlation patterns between the input data features, output the health index score of the monitored person, and judge their current status. The confidence evaluation unit is used to evaluate the confidence of the health index score output by the regression model using cross-validation to obtain the confidence level.

[0027] The transmission credibility evaluation module includes an influencing factor prediction unit, a fuzzy relationship construction unit, and a comprehensive evaluation unit. The influencing factor prediction unit is used to apply a graph neural network GNN that fuses spatio-temporal features, encode multi-source data related to transmission in the form of a graph structure and input it into the model, learn the spatial correlations and temporal evolution patterns in the data, and output the predicted values of the influencing factors during the transmission of the health index score of the monitored person. The fuzzy relationship construction unit is used to use the predicted values of the influencing factors as the evaluation factor set, construct a comment set for transmission credibility, determine the membership degree relationship between the factors and each comment in the comment set through statistical analysis of historical data, and establish a fuzzy relationship matrix. The comprehensive evaluation unit is used to perform a synthesis operation on the factor weight vector and the fuzzy relationship matrix through weighted operations to obtain a fuzzy comprehensive evaluation result vector, assign corresponding scores to each comment in the transmission credibility comment set, and calculate the weighted average of the fuzzy comprehensive evaluation result vector and the corresponding scores of the comment set as the quantitative index of transmission credibility.

[0028] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. An artificial intelligence-based home-based elderly care intelligent monitoring method, characterized in that: The method comprises the following steps: Monitor the physiological and behavioral data of the monitored person through sensors; Clean, pre-process, normalize and synchronize the acquired data; Use the model to analyze behavioral data to identify anomalies, build a labeled data set, build a regression model for judging the current status of the supervised person through a neural network, and perform confidence assessment to obtain confidence; A quantitative index of transmission credibility is calculated, and a data transmission interval is adjusted according to the calculated confidence and the quantitative index of transmission credibility.

2. According to claim 1, an artificial intelligence-based home-based elderly care intelligent monitoring method is characterized by: The physiological and behavioral data of the monitored person are monitored through sensors. The specific steps include: The sensors are used to obtain the physiological and behavioral data of the monitored person, where the physiological data include the monitored person's heart rate, blood pressure and body temperature data; the behavioral data include the monitored person's activity trajectory, body posture changes, and the monitored person's interaction with the environment; The physiological and behavioral data of the monitored person are synchronously collected at certain time intervals, and the acquired data are sorted in time order [t1, t2, ..., tw] to form a time series of the physiological and behavioral data of the monitored person, where t1, t2, ..., tw represent the 1st, 2nd, ..., wth data collection moments respectively.

3. The method for intelligent home-based elderly care based on artificial intelligence according to claim 2 is characterized by: The collected physiological and behavioral data of the monitored persons are cleaned, pre-processed, normalized and synchronously integrated. The specific steps include: Clean and pre-process the collected physiological and behavioral data of the monitored persons to remove noise data. When the data value at a certain moment is missing, use the time series-based interpolation method to fill it in according to the data values ​​of the adjacent moments before and after. Standardize the data format and store all data in a unified database table structure. The Min-Max normalization algorithm is used to process the physiological data of the monitored persons of different magnitudes and uniformly map the data to the [0,1] interval; The physiological and behavioral data of the monitored person are synchronously integrated in the order of collection time. Based on the synchronously collected time points [t1, t2, ..., tw], the heart rate, blood pressure, body temperature, activity trajectory, body posture changes, and interaction information between the monitored person and the environment at the same time are associated and combined to form a complete time series data record.

4. The method for intelligent home-based elderly care based on artificial intelligence according to claim 3 is characterized by: The Transformer machine learning model is used to analyze behavioral data to identify abnormal behaviors, build a labeled data set, adjust the threshold according to cultural habits and disease characteristics to complete personalized settings, build a regression model for determining the current status of the person under guardianship, and send the status of the person under guardianship to the guardian. The specific steps include: Based on the preprocessed data, the behavioral data of the supervised persons are analyzed and modeled through the Transformer machine learning model. Through feature extraction and pattern recognition, the abnormal behaviors of the supervised persons are identified from the time series and action pattern dimensions of the behavioral data, and all the identified abnormal behaviors are combined into an abnormal behavior set. Build an annotated dataset for distinguishing specific behavior patterns and abnormal action representations; For the behavior types corresponding to specific cultural backgrounds, social customs or long-term personal behavior habits, feature extraction and description are performed in terms of the periodicity, duration and amplitude of the action. Based on the analysis of normal samples, the characteristic patterns of such behaviors under normal conditions are summarized; For abnormal movement types associated with diseases, characterize them from the perspective of the frequency of movement, regularity of movement presentation, and the dimensions of the limbs involved. By comparing and analyzing them with normal behavior movements, determine the abnormal characteristic patterns of such movements. Personalized threshold setting is performed based on the user's historical data. For the behavior types corresponding to specific cultural backgrounds, social customs, or long-term behavioral habits of individuals, the characteristic thresholds described are adjusted to specific multiples of the initial values ​​in the behavior pattern knowledge base, where the behavior pattern knowledge base is constructed based on historical behavior data; For abnormal action types associated with diseases, the characteristic thresholds describing them are adjusted to specific multiples of the initial values ​​in the behavior pattern knowledge base; A regression model for judging the current status of the monitored person is constructed by integrating a neural network framework with multi-source features. The physiological and behavioral data of the monitored person acquired in real time and the set alarm threshold are processed by embedded coding. The encoded data is input into the constructed neural network regression model integrating the fully connected layer and the convolutional layer. The association in the local features of the data is mined with the help of the convolutional layer, and the global features are integrated using the fully connected layer. The health index score of the current monitored person is output by learning and mapping the association pattern between the input data features. Using a cross-validation method, a confidence evaluation is performed on the health index score output by the regression model for determining the current state of the supervised person to obtain a confidence Z; The health index score of the current monitored person is sent to the monitoring person.

5. The method for intelligent home-based elderly care based on artificial intelligence according to claim 4 is characterized in that: Calculate the quantitative index of transmission credibility, and adjust the data transmission interval according to the calculated confidence and the quantitative index of transmission credibility. The specific steps include: The graph neural network (GNN) that integrates spatiotemporal features is used to predict the influencing factors in the process of sending the health index score of the monitored person to the monitor. The multi-source data related to the transmission is encoded in the form of a graph structure and input into the constructed GNN model. The model outputs the predicted values ​​of the influencing factors in the process of transmitting the health index score of the monitored person by learning the spatial association and time evolution pattern in the data. According to the predicted values ​​of the influencing factors in the process of transmitting the health index score of the monitored person as the evaluation factor set U={u1,u2,...,un}, construct a transmission credibility comment set V={v1,v2,...,vm}, wherein u1,u2,...,un represent the 1st,2nd,...,nth predicted values ​​of the influencing factors in the process of transmitting the health index score of the monitored person, wherein v1,v2,...,vm represent the 1st,2nd,...,mth level division of transmission credibility; According to the statistical analysis of historical data, the membership relationship between the factor ui and each comment in the comment set V is determined, and the fuzzy relationship matrix R is established; Specifically, through the statistical analysis of historical data, we can obtain that the membership of factor u1 to each comment is r11, r12, ..., r1m, and the membership of factor u2 to each comment is r21, r22, ..., r2m, and so on, we can obtain the membership of factor u3, ..., un to each comment; According to the membership relationship between the evaluation factor set and each comment in the comment set V, a fuzzy relationship matrix R is established; The fuzzy relationship matrix R is expressed as: ; The relative importance of each influencing factor in evaluating transmission credibility is determined by the Delphi method, that is, the factor weight vector is determined, and the factor weight vector is recorded as A=(a1,a2,...,an), where a1,a2,...,an represents the influence weight of the 1st, 2nd,...,nth factor on transmission credibility; The factor weight vector A and the fuzzy relationship matrix R are combined by weighted operation to obtain the fuzzy comprehensive evaluation result vector B. The calculation formula is as follows: B=A*R=(b1,b2,...,bm); in, , j=1,2,...,m, b1,b2,...,bm represent the degree values ​​of the transmission credibility determined as levels v1,v2,...,vm after the factor weight vector A and the fuzzy relationship matrix R are synthesized by weighted operation; Assign a corresponding score to each comment in the transmission credibility comment set, and calculate the weighted average of the fuzzy comprehensive evaluation result vector B and the corresponding score assigned to each comment in the transmission credibility comment set as the quantitative index Q of transmission credibility; The time interval for transmitting the health index score of the monitored person to the guardian is readjusted according to the calculated confidence Z and the quantitative index Q of transmission credibility. The time interval adjustment formula is as follows: T=T0+β*Z*Q; wherein β represents the adjustment coefficient and T0 represents the initially set time interval.

6. An artificial intelligence-based home-based elderly care intelligent monitoring system, characterized by: The system includes a data acquisition module, a data processing module, an abnormality analysis and status evaluation module, a transmission credibility evaluation module and a transmission control module. The data acquisition module is used to monitor the physiological and behavioral data of the monitored person through sensors; the data processing module is used to clean, pre-process, normalize and synchronously integrate the acquired data; the abnormality analysis and status evaluation module is used to use a model to analyze the behavioral data to identify abnormalities, build a labeled data set, build a regression model for judging the current status of the monitored person through a neural network, and perform confidence evaluation to obtain confidence; the transmission credibility evaluation module is used to calculate the quantitative index of transmission credibility; the transmission control module adjusts the data transmission interval according to the calculated confidence and the quantitative index of transmission credibility.

7. The artificial intelligence-based home-based elderly care intelligent monitoring system according to claim 6 is characterized by: The data acquisition module includes a physiological data acquisition unit and a behavioral data acquisition unit. The physiological data acquisition unit is used to acquire the heart rate, blood pressure and body temperature data of the monitored person through sensors. The behavioral data acquisition unit is used to acquire the activity trajectory, body posture changes, and interaction information between the monitored person and the environment through sensors.

8. The artificial intelligence-based home-based elderly care intelligent monitoring system according to claim 7 is characterized by: The data processing module includes a preprocessing unit, a normalization processing unit and a synchronization integration unit. The preprocessing unit is used to remove noise data from the collected physiological and behavioral data, fill in missing data with a time series-based interpolation method, standardize the data format and store it in a unified database table structure; the normalization processing unit is used to use a Min-Max normalization algorithm to process the physiological data of the monitored persons of different magnitudes, and uniformly map the data to the [0,1] interval; the synchronization integration unit is used to associate and combine the heart rate, blood pressure, body temperature, activity trajectory, body posture changes at the same time, and interaction information with the environment in accordance with the acquisition time sequence and based on the synchronous acquisition time point to form a complete time series data record.

9. The artificial intelligence-based home-based elderly care intelligent monitoring system according to claim 8 is characterized by: The abnormal analysis and status assessment module includes an abnormal behavior recognition unit, a regression model construction unit and a confidence assessment unit. The abnormal behavior recognition unit is used to extract features and recognize patterns from the time series and action pattern dimensions of the behavior data based on the preprocessed data using the Transformer machine learning model to find out the abnormal behaviors of the monitored person and form an abnormal behavior set. The regression model construction unit is used to embed and encode the physiological and behavioral data of the monitored person obtained in real time and the set alarm threshold through a neural network framework that integrates multi-source features, input the neural network regression model that integrates the fully connected layer and the convolutional layer, use the convolutional layer to mine the local feature association of the data, and the fully connected layer to integrate the global features, learn the association pattern between the input data features, output the health index score of the monitored person, and judge his current state; the confidence assessment unit is used to use a cross-validation method to perform confidence assessment on the health index score output by the regression model to obtain confidence.

10. The artificial intelligence-based home-based elderly care intelligent monitoring system according to claim 9, characterized in that: The transmission credibility assessment module includes an influencing factor prediction unit, a fuzzy relationship construction unit and a comprehensive evaluation unit. The influencing factor prediction unit is used to use a graph neural network GNN that integrates spatiotemporal features to encode multi-source data related to transmission in a graph structure form and input the model, learn the spatial association and time evolution pattern in the data, and output the predicted value of the influencing factor in the process of transmitting the health index score of the monitored person; The fuzzy relationship construction unit is used to use the predicted value of the influencing factor as the evaluation factor set, construct a comment set of transmission credibility, determine the membership relationship between the factors and each comment in the comment set based on the statistical analysis of historical data, and establish a fuzzy relationship matrix; the comprehensive evaluation unit is used to synthesize the factor weight vector and the fuzzy relationship matrix through weighted operation to obtain a fuzzy comprehensive evaluation result vector, assign a corresponding score to each comment in the transmission credibility comment set, and calculate the weighted average of the fuzzy comprehensive evaluation result vector and the corresponding score of the comment set as a quantitative indicator of transmission credibility.

Images