A home care system based on scene intelligent analysis

Abstract

The present invention provides a home care system based on scene intelligent analysis, which relates to the technical field of data processing. The method includes: collecting and preprocessing the behavior data and environmental data of the user to obtain daily data, then extracting features from the daily data, and calculating the weights of the behavior and the environment, aligning the behavior feature data and the environmental feature data in time to obtain a matching data set, and analyzing the correlation between the behavior and the environment according to the matching data set to generate behavior pattern data, predicting the behavior trend of the user under the real-time environmental data through the behavior pattern data to obtain predicted behavior data, and comparing the predicted behavior data with the real-time behavior data to obtain abnormal behavior data, judging whether the user is in potential danger according to the abnormal behavior data, and when the judgment result is yes, generating warning information, and sending the warning information to the guardian or a designated terminal. The present invention can judge whether the user is in danger by intelligently analyzing the scene.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a home care system based on scene intelligent analysis. Background Art

[0002] Existing home care systems monitor and care for users based on simple sensor technology. The system usually obtains and analyzes user behavior information through sensors installed in the home, such as infrared sensors, cameras, door magnetic sensors, etc. When the system detects abnormal user behavior or an emergency, such as a fall or not moving for a long time, it will automatically sound an alarm and notify the caregiver. This monitoring method is suitable for some simple scenarios, but it may not perform well in complex scenarios and it is difficult to accurately identify complex events.

[0003] For example, in the scenario of getting up at night, the system uses the sensor device to determine whether the user gets up and moves in the dark. However, if the user sits by the bed for a long time and is about to stand up, but does not move significantly, the system cannot accurately identify this behavior. The detection of the sensor device depends on the preset time threshold and movement amplitude. Once the user's behavior does not meet these preset standards, the system may not be able to determine whether there is a potential danger in time, resulting in the neglect of the risk of falling at night. Summary of the invention

[0004] The purpose of the present invention is to provide a home care system based on scene intelligent analysis, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] A home care system based on scene intelligent analysis, the system comprising:

[0007] The data collection module is used to collect the user's behavior data and environmental data, and pre-process the behavior data and environmental data to obtain daily behavior data and daily environmental data;

[0008] The feature extraction module is used to identify different behavior trajectories based on daily behavior data and extract behavior features from the behavior trajectories to obtain behavior feature data; based on daily environmental data, it collects spatial environmental features, extracts physical environmental features, and obtains environmental feature data;

[0009] The weight assignment module is used to analyze the impact of behavior on user safety based on behavior feature data to obtain behavior weights; and to analyze the impact of the environment on user safety based on environment feature data to obtain environment weights;

[0010] The behavior pattern module is used to time-align the behavior feature data and the environment feature data to obtain a matching data set, and analyze the correlation between the behavior and the environment based on the matching data set to generate behavior pattern data;

[0011] The behavior prediction module is used to obtain the user's real-time environment data and predict the user's behavior trend under the real-time environment data through the behavior pattern data to obtain the predicted behavior data;

[0012] The abnormal analysis module is used to obtain the real-time behavior data of users and compare the predicted behavior data with the real-time behavior data to obtain abnormal behavior data;

[0013] The danger judgment module is used to judge whether the user is in potential danger based on abnormal behavior data. If the judgment result is yes, a warning message is generated. If the judgment result is no, the user's behavior continues to be monitored;

[0014] The alarm prompt module is used to send warning information to the guardian or designated terminal.

[0015] Furthermore, the feature extraction module includes:

[0016] The feature extraction module comprises:

[0017] A behavior feature extraction unit is used to identify the amplitude, frequency and duration of user behavior based on daily behavior data, obtain a behavior trajectory, and extract features from the behavior trajectory to obtain behavior feature data;

[0018] The environmental feature extraction unit is used to analyze the distance between the object and the user based on the daily environmental data to obtain the spatial environmental features, and collect the temperature and light at different time points from the daily environmental data to obtain the physical environmental features, and obtain the environmental feature data based on the physical environmental features and the physical environmental features.

[0019] Furthermore, the weighting module includes:

[0020] The behavior weight unit is used to analyze the impact of different behaviors on user security based on the behavior feature data, calculate the weight value of the behavior, and obtain the behavior weight. The calculation formula of the behavior weight value is:

[0021] ,

[0022] in, For the The weight of each behavior, is the index of the behavior, For the behavior, For the The number of dangerous historical events for each behavior, is the total number of historical events, For the The frequency of the behavior, For the The duration of the behavior, For the The magnitude of the behavior, For the Behavior and other behaviors The covariance of and Respectively Behavior and other behaviors The standard deviation of For the The adjustment coefficient of each behavior, is the normalization coefficient of the behavior weight.

[0023] Furthermore, the weighting module further includes:

[0024] The environment weight unit is used to analyze the impact of different environments on user security based on the environment feature data, calculate the weight value of the environment, and obtain the environment weight. The calculation formula of the environment weight value is:

[0025] ,

[0026] in, For the The weight of the environment, is the index of the environment, For the environment, For the Number of dangerous historical events in each environment, is the total number of historical events, For the The obstacle distance of the environment, For the The light intensity of the environment, For the The temperature of the environment, For the environment and other environments The covariance of and Respectively environment and other environments The standard deviation of For the The adjustment factor of the environment, is the normalization coefficient of the environmental weight.

[0027] Furthermore, the behavior mode module includes:

[0028] A time alignment unit, used to arrange the behavior feature data and the environment feature data in chronological order, and to align the behavior feature data and the environment feature data at the same time point to generate a matching data set;

[0029] A data partitioning unit is used to perform similarity calculation on the environmental features in the matching data set, divide the different environments, and divide the matching data set according to the similarities to generate matching data subsets;

[0030] The correlation analysis unit is used to calculate the correlation between the behavior characteristics and the environment characteristics in the matching data subset to obtain the behavior environment correlation. The calculation formula of the behavior environment correlation is:

[0031] ,

[0032] in, For the Behavior and The behavior-environment correlation of each environment, For the The first behavioral characteristics, For the The first environmental characteristics, For the The behavioral characteristics and The sum of the standard deviations of the environmental characteristics of the environments, and are the total number of behavioral features and environmental features, is the cross entropy function, ;

[0033] The behavior pattern determination unit is used to screen and compare the behavior feature data and the environment feature data according to the behavior environment correlation and a preset correlation threshold to determine the behavior pattern.

[0034] Further, the behavior prediction module includes:

[0035] A real-time environment unit is used to obtain the user's real-time environment data and extract environment features from the real-time environment data to obtain real-time environment features;

[0036] An environment comparison unit is used to calculate the similarity between the real-time environment characteristics and the comparison environment characteristics, select the behavior pattern, and obtain the reference behavior pattern;

[0037] The behavior prediction unit is used to extract comparative behavior feature data according to the reference behavior pattern to obtain behavior prediction data.

[0038] Furthermore, the abnormality analysis module includes:

[0039] A behavior comparison unit is used to obtain the user's real-time behavior data and compare it with the behavior prediction data to obtain the behavior similarity;

[0040] The abnormal behavior unit is used to mark the real-time behavior data as abnormal when the behavior similarity is lower than a preset behavior similarity threshold, and obtain abnormal behavior data; when the behavior similarity is not lower than the preset behavior similarity threshold, the real-time behavior data is marked as normal.

[0041] Furthermore, the calculation formula for behavior similarity is:

[0042] ,

[0043] in, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, is the total number of behavioral features, For the The adjustment coefficient of the behavioral characteristics, For the The weighting coefficient of each behavioral feature, is a constant.

[0044] Furthermore, the risk judgment module includes:

[0045] A risk calculation unit is used to quantify the impact of abnormal behavior on user safety based on abnormal behavior data and calculate the risk matching degree;

[0046] The danger judgment unit is used to judge that the user is in potential danger when the danger matching degree is greater than the preset danger matching degree threshold, turn on the warning state and generate warning information; when the potential danger matching degree is not greater than the preset danger matching degree threshold, judge that the user is not in potential danger, do not turn on the warning state and continue to monitor the user's behavior.

[0047] Furthermore, the calculation formula of the risk matching degree is:

[0048] ,

[0049] in, For the The risk matching degree of each behavior, and Respectively Behavior and The weight of the environment, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, Real-time behavioral data Middle The maximum value of the behavior characteristics, is the total number of behavioral features, For the The risk factor of each behavioral characteristic is is the adjustment coefficient.

[0050] The above solution of the present invention includes at least the following beneficial effects:

[0051] By combining the dual dimensions of behavioral data and environmental data, the present invention enables the system to not only identify behavioral characteristics, but also obtain physical environmental characteristics under different time and space, providing data support for analyzing the relationship between behavior and environment; through dual-dimensional data acquisition, the system can capture changes in user behavior patterns under different environmental and time conditions. Traditional home monitoring systems usually rely on behavioral data and may ignore the impact of environmental changes on behavior. The system enriches the data source by mining data on both behavior and environment. This multi-dimensional feature extraction makes the system's monitoring more accurate, and can more comprehensively analyze the user's status and the risk factors of his or her surrounding environment, providing an accurate data basis for subsequent behavior prediction and abnormality sharing.

[0052] The present invention forms a risk analysis mechanism by assigning weight values ​​to behavioral and environmental characteristics respectively, so that the system can adjust the focus of risk identification according to actual conditions and adapt to dynamic changes in users and environments; the weight assignment function enables the system to have a highly personalized risk analysis capability. By assigning different weights to behavioral and environmental characteristics, the system can flexibly identify high-risk behaviors and environments. Compared with the fixed thresholds of existing home systems, the present system can adjust the analysis focus in real time, which is particularly suitable for elderly users or users who require special care, greatly improving the system's risk identification efficiency and accuracy, and avoiding underreporting or false alarms caused by behavioral or environmental changes.

[0053] The present invention generates the user's behavior pattern in a specific environment through time alignment and association analysis, so that the system can effectively identify the relationship between behavior and environment, establish multi-dimensional associations for behavior patterns under different environmental characteristics, and provide a dynamic behavior-environment association model; existing systems generally rely on single behavior analysis at a fixed time, while the present system synchronizes behavior with environmental characteristics, so that when the system detects low light at night, it can establish an association between the nighttime waking behavior pattern and the light conditions. This analysis method not only enables the system to capture changes in daily behavior, but also to identify potential abnormal patterns in a timely manner, providing more accurate basic data for behavior prediction and abnormality analysis. This design ensures that the system's monitoring results are more reliable and helps to comprehensively analyze user behavior and its risks.

[0054] The present invention generates user behavior prediction data by matching real-time environmental data with behavior pattern data, thereby evaluating the potential impact of the current environment on user behavior in real time; according to changes in the user's real-time environment, potential high-risk behaviors can be predicted in advance, with high predictability; the system can predict the user's possible next behavior based on the behavior records in similar environments in the behavior pattern. This prediction capability enables the system to intervene in potential risks in advance, helps to take measures before an accident occurs, and effectively improves the system's monitoring effect and user safety.

[0055] The present invention combines behavior prediction with comparison of real-time data to accurately identify abnormal behavior of users without relying on a single behavioral feature. Existing systems usually detect anomalies based on a single behavioral threshold, while the present system makes the detection of abnormal behavior more flexible and accurate by comparing predicted behavior with real-time behavior. The system can identify abnormal behavior through the behavior comparison unit. Such a design can effectively reduce false alarms and ensure that the system has a keener recognition ability for abnormal behavior, making the system's monitoring more reliable, greatly improving the system's risk monitoring effect, and ensuring that the system can respond quickly when abnormal behavior occurs.

[0056] The present invention provides a scientific basis for the system's early warning by accurately quantifying the risk level of abnormal behavior and dynamically adjusting the judgment standard according to the matching degree. The system refines the risk assessment of abnormal behavior through the dangerous matching degree of abnormal behavior, thereby avoiding misjudgment, calculates the risk level according to the dangerous matching degree of behavioral characteristics, and makes a more reliable early warning judgment after comparing it with the threshold, which greatly improves the system's early warning accuracy, avoids false alarms and missed alarms, ensures that guardians can be notified in time when there is a real risk, and improves the response timeliness and reliability of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1It is a flowchart of a home care system based on scene intelligent analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0059] like Figure 1 As shown, an embodiment of the present invention provides a home care system based on scene intelligent analysis, and the system includes:

[0060] The data collection module is used to collect the user's behavior data and environmental data, and pre-process the behavior data and environmental data to obtain daily behavior data and daily environmental data;

[0061] The feature extraction module is used to identify different behavior trajectories based on daily behavior data and extract behavior features from the behavior trajectories to obtain behavior feature data; based on daily environmental data, it collects spatial environmental features, extracts physical environmental features, and obtains environmental feature data;

[0062] The weight assignment module is used to analyze the impact of behavior on user safety based on behavior feature data to obtain behavior weights; and to analyze the impact of the environment on user safety based on environment feature data to obtain environment weights;

[0063] The behavior pattern module is used to time-align the behavior feature data and the environment feature data to obtain a matching data set, and analyze the correlation between the behavior and the environment based on the matching data set to generate behavior pattern data;

[0064] The behavior prediction module is used to obtain the user's real-time environment data and predict the user's behavior trend under the real-time environment data through the behavior pattern data to obtain the predicted behavior data;

[0065] The abnormal analysis module is used to obtain the real-time behavior data of users and compare the predicted behavior data with the real-time behavior data to obtain abnormal behavior data;

[0066] The danger judgment module is used to judge whether the user is in potential danger based on abnormal behavior data. If the judgment result is yes, a warning message is generated. If the judgment result is no, the user's behavior continues to be monitored;

[0067] The alarm prompt module is used to send warning information to the guardian or designated terminal.

[0068] In an embodiment of the present invention, the system continuously obtains the user's behavior data and environmental data through the data acquisition module. These data will serve as an important reference for analyzing the user's behavior pattern. These data are preprocessed, and the processed data include daily behavior data and environmental data; the feature extraction module extracts behavior features and environmental features based on the user's daily behavior data and environmental data, so that the system can comprehensively and accurately characterize the user's behavior pattern and environmental impact, laying a data foundation for subsequent analysis; the weight assignment module analyzes the risk level of different user behaviors and the impact of the environment on the user's behavior based on the behavior feature data and environmental feature data, and assigns them appropriate weights; the behavior pattern module time-aligns the behavior feature data and the environmental feature data to generate a matching data set, and analyzes the correlation between behavior and environment based on the matching data set, thereby generating behavior pattern data, recording the common behavior patterns of users in different environments, and providing important reference information for predicting the user's future behavior.

[0069] Then, the behavior prediction module matches the user's current environmental data with the environmental features in the behavior pattern data, and finds the most similar historical environmental features. Based on this, it predicts the user's behavior trend in the environment and provides a standard for the subsequent abnormal analysis. The abnormal analysis module compares the user's real-time behavior data with the predicted behavior data. When there is a significant deviation between the real-time behavior and the predicted behavior, the system will mark the behavior as abnormal behavior and generate abnormal behavior data to further determine whether the user is in potential danger. The danger judgment module calculates the user's potential danger matching degree based on the abnormal data. When the matching degree exceeds the preset danger threshold, the system determines that the user is in potential danger, generates warning information, and switches the system state to the warning state. Finally, the alarm prompt module sends the warning information to the guardian or designated terminal device according to the warning state.

[0070] Among them, in the data acquisition module, the system captures the user's movement trajectory, position change, acceleration and other information through sensors, records the user's activity status, and the collected environmental data includes environmental characteristics such as temperature, humidity, light intensity and sound. These data will serve as an important reference for analyzing user behavior patterns; next, the system cleans, normalizes and denoises the raw data to remove outliers and invalid data that may interfere with the analysis. The processed data includes daily behavior data and environmental data, which provides high-quality input for subsequent feature extraction. The system can effectively improve the accuracy and consistency of the data and avoid the impact of redundant data on the analysis results.

[0071] Among them, the alarm prompt module sends warning information to the guardian or designated terminal device so that relevant personnel can quickly learn and take appropriate measures. This module is not limited to sending warning information, but can also prompt or remind users through the system to help users avoid potential dangerous situations.

[0072] In a preferred embodiment of the present invention, the feature extraction module includes:

[0073] A behavior feature extraction unit is used to identify the amplitude, frequency and duration of user behavior based on daily behavior data, obtain a behavior trajectory, and extract features from the behavior trajectory to obtain behavior feature data;

[0074] The environmental feature extraction unit is used to analyze the distance between the object and the user based on the daily environmental data to obtain the spatial environmental features, and collect the temperature and light at different time points from the daily environmental data to obtain the physical environmental features, and obtain the environmental feature data based on the physical environmental features and the physical environmental features.

[0075] In an embodiment of the present invention, a behavior feature extraction unit collects the duration, frequency and amplitude of user behavior from daily behavior data, and uses time series analysis technology to identify behavior trajectories from continuous mobile data and extract behavior features therefrom. These behaviors are encoded into feature vectors for further analysis. At the same time, an environmental feature extraction unit analyzes environmental features in a living environment, which includes collecting data from environmental sensors, such as indoor temperature, humidity and light intensity, and analyzing the distance between an object and the user to determine whether an obstacle is formed, and extracting environmental features from these data to ensure that the impact of the environment on user behavior can be accurately reflected.

[0076] Among them, according to the daily behavior data, the amplitude, frequency and duration of the user's behavior are identified to obtain the behavior trajectory, and the characteristics of the behavior trajectory are extracted to obtain the behavior characteristic data, which specifically includes:

[0077] Based on daily behavior data, we can analyze behavior waveforms and measure instantaneous amplitude of movements to identify the amplitude of user behavior, that is, judge the movement intensity of the behavior; calculate the frequency of user behavior and record the degree of repetition of user behavior; monitor the duration of each behavior and identify whether the user is in a resting state or a continuously moving state by analyzing the length of time the user remains still or the continuity of the action; combine the amplitude, frequency and duration data to generate the user's behavior trajectory, perform feature extraction on the trajectory data, extract behavior features such as speed, acceleration, direction change, path length, etc. from the trajectory, and obtain behavior feature data.

[0078] Among them, according to the daily environmental data, the distance between the object and the user is analyzed to obtain the spatial environmental characteristics, and the temperature and light at different time points are collected from the daily environmental data to obtain the physical environmental characteristics. According to the physical environmental characteristics and the physical environmental characteristics, the environmental characteristic data is obtained, which specifically includes:

[0079] Based on daily environmental data, the distance between the objects around the user and the user can be analyzed, the relative position of the user and fixed objects such as furniture and electrical appliances can be measured to determine the spatial layout characteristics; the temperature data in the environment can be collected in real time, and the temperature change trend can be recorded; the light intensity in the environment can be collected, and the light changes can be recorded; the above-mentioned spatial layout characteristics and temperature and light data can be comprehensively analyzed to obtain complete environmental characteristic data.

[0080] In a preferred embodiment of the present invention, the weight assignment module includes:

[0081] The behavior weight unit is used to analyze the impact of different behaviors on user security based on the behavior feature data, calculate the weight value of the behavior, and obtain the behavior weight. The calculation formula of the behavior weight value is:

[0082] ,

[0083] in, For the The weight of each behavior, is the index of the behavior, For the behavior, For the The number of dangerous historical events for each behavior, is the total number of historical events, For the The frequency of the behavior, For the The duration of the behavior, For the The magnitude of the behavior, For the Behavior and other behaviors The covariance of and Respectively Behavior and other behaviors The standard deviation of For the The adjustment coefficient of each behavior, is the normalization coefficient of the behavior weight.

[0084] In an embodiment of the present invention, the behavior weight unit analyzes the impact of the behavior on the user's safety through the behavior characteristics, takes into account whether the behavior is dangerous and the strength of the association between the behaviors, and designs a behavior weight calculation formula, which ensures the rationality of the behavior weight through adjustment coefficients and normalization processing.

[0085] in, For the The risk score of a behavior indicates the number of times the behavior is associated with dangerous events in historical data. For example, frequent nighttime activities may be assigned a higher risk score because they increase the risk of falls and injuries. This score is obtained through historical data analysis to ensure the accuracy and real-time nature of the score.

[0086] in, For the The system further analyzes the correlation between different behavioral features. For example, if a certain behavioral feature often appears with other high-risk behaviors, the weight of the behavioral feature will be increased accordingly. The correlation analysis helps the module understand the complex interaction of behavioral patterns and increase the predictive ability and accuracy of the model.

[0087] In a preferred embodiment of the present invention, the weight assigning module further includes:

[0088] The environment weight unit is used to analyze the impact of different environments on user security based on the environment feature data, calculate the weight value of the environment, and obtain the environment weight. The calculation formula of the environment weight value is:

[0089] ,

[0090] in, For the The weight of the environment, is the index of the environment, For the environment, For the Number of dangerous historical events in each environment, is the total number of historical events, For the The obstacle distance of the environment, For the The light intensity of the environment, For the The temperature of the environment, For the environment and other environments The covariance of and Respectively environment and other environments The standard deviation of For the The adjustment factor of the environment, is the normalization coefficient of the environmental weight.

[0091] In an embodiment of the present invention, the environmental weight unit analyzes the impact of the environment on user safety through environmental characteristics, takes into account the risk level and synergy coefficient of the environment, integrates these factors together, and designs an environmental weight calculation formula, which ensures the rationality of the environmental weight through adjustment coefficients and normalization processing.

[0092] in, For the The risk level of an environment indicates the number of times the environment has been associated with dangerous events in historical data. For example, too low light may increase the risk of falling, and too high or too low temperature may affect the health of users. These risk assessments are determined through historical data statistics and health and safety standards to ensure the accuracy and relevance of the assessment.

[0093] in, For the The synergy coefficient between an environment and other environments is calculated, and the synergy relationship between different environmental characteristics is systematically analyzed. For example, the combined effect of temperature and humidity may have a greater impact on user comfort and health. Through this analysis, the module can more accurately evaluate the overall impact of complex environmental factors on user safety.

[0094] In a preferred embodiment of the present invention, the behavior pattern module includes:

[0095] A time alignment unit, used to arrange the behavior feature data and the environment feature data in chronological order, and to align the behavior feature data and the environment feature data at the same time point to generate a matching data set;

[0096] A data partitioning unit is used to perform similarity calculation on the environmental features in the matching data set, divide the different environments, and divide the matching data set according to the similarities to generate matching data subsets;

[0097] The correlation analysis unit is used to calculate the correlation between the behavior characteristics and the environment characteristics in the matching data subset to obtain the behavior environment correlation. The calculation formula of the behavior environment correlation is:

[0098] ,

[0099] in, For the Behavior and The behavior-environment correlation of each environment, For the The first behavioral characteristics, For the The first environmental characteristics, For the The behavioral characteristics and The sum of the standard deviations of the environmental characteristics of the environments, and are the total number of behavioral features and environmental features, is the cross entropy function, ;

[0100] The behavior pattern determination unit is used to screen and compare the behavior feature data and the environment feature data according to the behavior environment correlation and a preset correlation threshold to determine the behavior pattern.

[0101] In an embodiment of the present invention, the time alignment unit sorts the behavior feature data and the environment feature data in chronological order to ensure the consistency of the behavior data and the environment data on the time axis. Through timestamp alignment, the module merges the two types of data into a time series data set, so that each behavior data point corresponds to the corresponding environment data point, and generates a matching data set, thereby providing an accurate basis for analyzing the interaction between behavior and environment; the data partitioning unit divides the matching data set into different matching data subsets by calculating the similarity of environmental features in the matching data set, and each matching data subset is a different environment, so as to count the number of behavior patterns; then the association analysis unit module uses the association calculation formula to analyze the correlation between the behavior features and the environment features in the matching data subset to identify the strong correlation between the behavior and the specific environmental conditions; the behavior pattern determination unit identifies the behavior environment pattern with statistical significance according to the behavior environment correlation and the preset correlation threshold. When the behavior environment correlation meets the preset correlation threshold, these behavior patterns are considered to be important behavior patterns.

[0102] The behavior feature data and the environment feature data are arranged in chronological order, and the behavior feature data and the environment feature data at the same time point are aligned to generate a matching data set, which specifically includes:

[0103] Sort the behavioral feature data and environmental feature data in the order of their occurrence to make the time sequence of the data clear and facilitate subsequent processing; align the data with the same time point in the behavioral feature data and environmental feature data, that is, find the behavioral features and environmental features at the same time point, so that there is a temporal correspondence between the behavioral feature data and the environmental feature data, and obtain a matching data set, in which the behavioral features and environmental features at each time point appear in correspondence.

[0104] The similarity of the environmental features in the matching data set is calculated to divide different environments, and the matching data set is divided according to the similarity to generate matching data subsets, which specifically include:

[0105] By calculating the similarity of the environmental features in the matching data set, we can determine which environmental features have a higher similarity and divide them into different environmental feature categories in order to perform more detailed environmental data grouping; based on the similarity calculation, the environmental features are classified into several subsets, each subset represents an environmental state, and each subset corresponds to a specific environmental category; for each environmental category divided, the corresponding behavioral feature data are aligned to generate a matching data subset, each matching data subset contains the behavioral features under the environmental state, thus forming a data set between this type of environmental features and behavioral features.

[0106] The similarity calculation is performed on the environmental features in the matching data set, and the calculation formula is consistent with the calculation formula of the behavior similarity below.

[0107] Among them, according to the behavior-environment correlation and the preset correlation threshold, the behavior feature data and the environment feature data are screened and compared to determine the behavior pattern, which specifically includes:

[0108] Based on the calculated behavior-environment correlation, the correlation value can be compared with the preset threshold, and behavioral features and environmental features with higher correlation can be screened out, and low-correlation data can be eliminated to ensure high correlation of behavioral patterns, so as to improve the accuracy and precision of predictions; based on the screened high-correlation behavioral features and environmental features, the behavioral patterns are determined, that is, the typical behavioral performance of users in specific environments is identified. These behavioral patterns will serve as the basis for subsequent behavior prediction modules to evaluate the behavioral trends of users in real-time environments.

[0109] Among them, the preset correlation threshold is obtained by analyzing a large amount of user behavior and environmental data, and is optimized using statistical methods to ensure that it can effectively distinguish between relevant and irrelevant behavior patterns.

[0110] In a preferred embodiment of the present invention, the behavior prediction module includes:

[0111] A real-time environment unit is used to obtain the user's real-time environment data and extract environment features from the real-time environment data to obtain real-time environment features;

[0112] An environment comparison unit is used to calculate the similarity between the real-time environment characteristics and the comparison environment characteristics, select the behavior pattern, and obtain the reference behavior pattern;

[0113] The behavior prediction unit is used to extract comparative behavior feature data according to the reference behavior pattern to obtain behavior prediction data.

[0114] In an embodiment of the present invention, the real-time environment unit receives the real-time environment data of the user, which data include but are not limited to indoor temperature, humidity, light intensity and object placement, extracts key features in the real-time environment data, and identifies the main variables in the real-time environment data; the environment comparison unit calculates the similarity between the real-time environment characteristics and the comparison environment characteristics through a similarity formula, and finds the historical environment most similar to the current environmental conditions according to the similarity, and determines the corresponding behavior pattern to obtain a reference behavior pattern; the behavior prediction unit extracts the correlation between the environment and the behavior according to the comparison behavior pattern, and extracts the comparison behavior data corresponding to the real-time environment to obtain behavior prediction data.

[0115] The real-time environment data of the user is obtained, and the environment features are extracted from the real-time environment data to obtain the real-time environment features, which specifically includes:

[0116] Collect data from the user's current environment. This data reflects the state of the user's environment, that is, it is obtained at the user's current location and time point, and can reflect the user's current environmental conditions; from the collected real-time environmental data, extract environmental features that can be used for analysis, such as time, location, temperature, humidity, noise, light intensity, weather conditions, etc., to obtain real-time environmental features.

[0117] Among them, calculating the similarity between the real-time environment characteristics and the comparison environment characteristics, selecting the behavior pattern, and obtaining the reference behavior pattern specifically includes:

[0118] The similarity between the two sets of features is calculated to quantify the closeness between the real-time environment and the comparison environment. The higher the similarity, the closer the current environment is to the comparison environment. After determining the similarity between the real-time environment and the comparison environment, a comparison environment with high similarity can be selected and the corresponding behavior pattern can be selected. According to the behavior pattern corresponding to the comparison environment similar to the current environment, a reference behavior pattern can be obtained.

[0119] Among them, the similarity between the real-time environment characteristics and the comparison environment characteristics is calculated, and the calculation formula is consistent with the calculation formula of the behavior similarity below.

[0120] In a preferred embodiment of the present invention, the abnormality analysis module includes:

[0121] A behavior comparison unit is used to obtain the user's real-time behavior data and compare it with the behavior prediction data to obtain the behavior similarity;

[0122] The abnormal behavior unit is used to mark the real-time behavior data as abnormal when the behavior similarity is lower than a preset behavior similarity threshold, and obtain abnormal behavior data; when the behavior similarity is not lower than the preset behavior similarity threshold, the real-time behavior data is marked as normal.

[0123] In an embodiment of the present invention, the behavior comparison unit compares the real-time behavior data with the predicted behavior data, and uses a similarity formula to calculate the similarity between the two, so as to accurately identify instances in which there are significant differences between the actual behavior and the predicted behavior; the behavior prediction unit uses the system's preset behavior similarity threshold to mark the behavior data whose behavior similarity is lower than the preset behavior similarity threshold as abnormal, and obtains abnormal behavior data; if the behavior similarity is not lower than the preset behavior similarity threshold, the behavior data is marked as normal, and the user behavior continues to be detected.

[0124] Among them, the preset behavior similarity threshold is determined based on historical data analysis to ensure sensitivity and accuracy to abnormal behavior. The setting of the threshold takes into account the variability of different types of behavior and common behavior patterns.

[0125] In a preferred embodiment of the present invention, the calculation formula of behavior similarity is:

[0126] ,

[0127] in, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, is the total number of behavioral features, For the The adjustment coefficient of the behavioral characteristics, For the The weighting coefficient of each behavioral feature, is a constant.

[0128] In the embodiment of the present invention, the similarity calculation formula considers the weighted sum of the difference between the behavior characteristics of the real-time behavior and the predicted behavior, reflects the difference between the real-time behavior characteristics and the predicted behavior characteristics, and normalizes them through the adjustment coefficient to standardize the similarity calculation.

[0129] in, is a minimum constant used to avoid the denominator being zero.

[0130] In a preferred embodiment of the present invention, the risk judgment module includes:

[0131] A risk calculation unit is used to quantify the impact of abnormal behavior on user safety based on abnormal behavior data and calculate the risk matching degree;

[0132] The danger judgment unit is used to judge that the user is in potential danger when the danger matching degree is greater than the preset danger matching degree threshold, turn on the warning state and generate warning information; when the potential danger matching degree is not greater than the preset danger matching degree threshold, judge that the user is not in potential danger, do not turn on the warning state and continue to monitor the user's behavior.

[0133] In an embodiment of the present invention, the danger calculation unit analyzes abnormal behavior data, which includes the user's real-time behavior characteristics, the corresponding behavior weights, and the environmental weights and the correlation information between the behavior and the environment, and calculates the degree of danger of the abnormal behavior to the user to obtain the danger matching degree; the danger judgment unit compares the calculated danger matching degree with the system's preset danger matching degree threshold. If the danger matching degree exceeds the preset danger matching degree threshold, the system determines that the user is in a potentially dangerous state, immediately turns on the warning state, and generates warning information. The warning information includes the detected abnormal behavior and its degree of danger. If the danger matching degree does not exceed the preset danger matching degree threshold, the system will continue to monitor the user's behavior. During this process, regardless of whether the system turns on the warning state, it will continue to collect and analyze the user's behavior and environmental data. This continuous monitoring ensures that the system can dynamically adapt to changes in user behavior and adjust the danger assessment in time when necessary to better protect the user's safety.

[0134] Among them, the preset danger matching threshold sets a matching upper limit through experiments and user safety demand analysis. It is usually calculated using a risk assessment model and dynamically optimized based on feedback from real life.

[0135] In a preferred embodiment of the present invention, the calculation formula of the risk matching degree is:

[0136] ,

[0137] in, For the The risk matching degree of each behavior, and Respectively Behavior and The weight of the environment, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, Real-time behavioral data Middle The maximum value of a behavior characteristic, is the total number of behavioral features, For the The risk factor of each behavioral characteristic is is the adjustment coefficient.

[0138] In an embodiment of the present invention, the system quantifies the safety impact of abnormal behavior on users through a risk matching calculation formula. The risk matching calculation formula combines behavioral weights and environmental weights, accumulates all relevant factors, and then applies a maximum correction item to enhance sensitivity to potential extreme behavioral characteristics. This correction method ensures that even if the risk of individual behavioral characteristics is particularly high, the system can identify them in time and give them sufficient attention. The risk matching calculation formula adjusts the calculation results through an adjustment coefficient so that the risk matching remains stable and reasonable within a certain range.

[0139] in, For the The risk coefficient of each behavioral characteristic is determined by analyzing historical data to determine which behavioral characteristics are most significant in dangerous events. The risk coefficient is quantified based on these data, and the frequency and consequences of dangerous events are monitored by simulating real scenarios to further adjust and optimize the risk coefficient.

[0140] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A home care system based on scene intelligent analysis, characterized in that: The system comprises: The data collection module is used to collect the user's behavior data and environmental data, and pre-process the behavior data and environmental data to obtain daily behavior data and daily environmental data; The feature extraction module is used to identify different behavior trajectories based on daily behavior data and extract behavior features from the behavior trajectories to obtain behavior feature data; based on daily environmental data, it collects spatial environmental features, extracts physical environmental features, and obtains environmental feature data; The weight assignment module is used to analyze the impact of behavior on user safety based on behavior feature data to obtain behavior weights; and to analyze the impact of the environment on user safety based on environment feature data to obtain environment weights; The behavior pattern module is used to time-align the behavior feature data and the environment feature data to obtain a matching data set, and analyze the correlation between the behavior and the environment based on the matching data set to generate behavior pattern data; The behavior prediction module is used to obtain the user's real-time environment data and predict the user's behavior trend under the real-time environment data through the behavior pattern data to obtain the predicted behavior data; The abnormal analysis module is used to obtain the real-time behavior data of users and compare the predicted behavior data with the real-time behavior data to obtain abnormal behavior data; The danger judgment module is used to judge whether the user is in potential danger based on abnormal behavior data. If the judgment result is yes, a warning message is generated. If the judgment result is no, the user's behavior continues to be monitored; Alarm prompt module, used to send warning information to the guardian or designated terminal; The feature extraction module comprises: A behavior feature extraction unit is used to identify the amplitude, frequency and duration of user behavior based on daily behavior data, obtain a behavior trajectory, and extract features from the behavior trajectory to obtain behavior feature data; The environmental feature extraction unit is used to analyze the distance between the object and the user according to the daily environmental data to obtain the spatial environmental features, and collect the temperature and light at different time points from the daily environmental data to obtain the physical environmental features, and obtain the environmental feature data according to the physical environmental features and the physical environmental features; The weighting module comprises: The behavior weight unit is used to analyze the impact of different behaviors on user security based on the behavior feature data, calculate the weight value of the behavior, and obtain the behavior weight. The calculation formula of the behavior weight value is: ， in, For the The weight of each behavior, is the index of the behavior, For the behavior, For the The number of dangerous historical events for each behavior, is the total number of historical events, For the The frequency of the behavior, For the The duration of a behavior, For the The magnitude of the behavior, For the Behavior and other behaviors The covariance of and Respectively Behavior and other behaviors The standard deviation of For the The adjustment coefficient of each behavior, is the normalization coefficient of behavior weight; The environment weight unit is used to analyze the impact of different environments on user security based on the environment feature data, calculate the weight value of the environment, and obtain the environment weight. The calculation formula of the environment weight value is: ， in, For the The weight of the environment, is the index of the environment, For the environment, For the Number of dangerous historical events in each environment, is the total number of historical events, For the The obstacle distance of the environment, For the The light intensity of the environment, For the The temperature of the environment, For the environment and other environments The covariance of and Respectively environment and other environments The standard deviation of For the The adjustment factor of the environment, is the normalized coefficient of environmental weight; The behavior mode module includes: A time alignment unit, used to arrange the behavior feature data and the environment feature data in chronological order, and to align the behavior feature data and the environment feature data at the same time point to generate a matching data set; A data partitioning unit is used to perform similarity calculation on the environmental features in the matching data set, divide the different environments, and divide the matching data set according to the similarities to generate matching data subsets; The correlation analysis unit is used to calculate the correlation between the behavior characteristics and the environment characteristics in the matching data subset to obtain the behavior environment correlation. The calculation formula of the behavior environment correlation is: ， in, For the Behavior and The behavior-environment correlation of each environment, For the The first behavioral characteristics, For the The first environmental characteristics, For the The behavioral characteristics and The sum of the standard deviations of the environmental characteristics of the environments, and are the total number of behavioral features and environmental features, is the cross entropy function, ; The behavior pattern determination unit is used to screen and compare the behavior feature data and the environment feature data according to the behavior environment correlation and a preset correlation threshold to determine the behavior pattern.

2. A home care system based on scene intelligent analysis according to claim 1, characterized in that: The behavior prediction module includes: A real-time environment unit is used to obtain the user's real-time environment data and extract environment features from the real-time environment data to obtain real-time environment features; An environment comparison unit is used to calculate the similarity between the real-time environment characteristics and the comparison environment characteristics, select the behavior pattern, and obtain the reference behavior pattern; The behavior prediction unit is used to extract comparative behavior feature data according to the reference behavior pattern to obtain behavior prediction data.

3. A home care system based on scene intelligent analysis according to claim 2, characterized in that: The abnormality analysis module comprises: A behavior comparison unit is used to obtain the user's real-time behavior data and compare it with the behavior prediction data to obtain the behavior similarity; The abnormal behavior unit is used to mark the real-time behavior data as abnormal when the behavior similarity is lower than a preset behavior similarity threshold, and obtain abnormal behavior data; when the behavior similarity is not lower than the preset behavior similarity threshold, the real-time behavior data is marked as normal.

4. The home care system based on scene intelligent analysis according to claim 3 is characterized in that: The calculation formula for behavior similarity is: ， in, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, is the total number of behavioral features, For the The adjustment coefficient of each behavioral characteristic, For the The weighting coefficient of each behavioral feature, is a constant.

5. The home care system based on scene intelligent analysis according to claim 4 is characterized in that: The danger judgment module comprises: A risk calculation unit is used to quantify the impact of abnormal behavior on user safety based on abnormal behavior data and calculate the risk matching degree; The danger judgment unit is used to judge that the user is in potential danger when the danger matching degree is greater than the preset danger matching degree threshold, turn on the warning state and generate warning information; when the potential danger matching degree is not greater than the preset danger matching degree threshold, judge that the user is not in potential danger, do not turn on the warning state and continue to monitor the user's behavior.

6. The home care system based on scene intelligent analysis according to claim 5, characterized in that: The calculation formula of risk matching is: ， in, For the The risk matching degree of each behavior, and Respectively Behavior and The weight of the environment, Real-time behavioral data and behavioral prediction data The similarity of Real-time behavioral data Middle The first behavioral characteristics, Predicting data for behavior Middle The first behavioral characteristics, Real-time behavioral data Middle The maximum value of the behavior characteristics, is the total number of behavioral features, For the The risk factor of each behavioral characteristic is is the adjustment coefficient.

Images