Home behavior recognition and abnormal event early warning system and method for old people

Abstract

The invention discloses an old people home behavior recognition and abnormal event early warning system and method, and relates to the technical field of abnormal recognition, and the system comprises a behavior monitoring module, a behavior analysis module, an abnormal event recognition module and an early warning feedback module. The behavior monitoring module is used for collecting behavior data of old people in a home environment in real time based on a sensor and performing data fusion; the behavior analysis module is used for analyzing behavior characteristics of the old people and predicting behavior action data of the old people at the next moment based on a time sequence analysis method.By means of multi-sensor data fusion and time sequence analysis, the behavior recognition accuracy is improved, and the behaviors of the old people are monitored and predicted in real time; according to the intelligent recognition and risk assessment based on machine learning, false alarm and missing alarm are reduced, the home safety is enhanced, and the life safety and comfort of old people are improved.

Description

Technical Field

[0001] The present invention relates to the field of abnormality recognition technology, and in particular to a system and method for identifying elderly people's home behavior and early warning of abnormal events. Background Art

[0002] With the rapid acceleration of population aging, the number of elderly people is showing a sustained and significant growth trend. Elderly health management and home safety have become key social concerns. In their homes, elderly people, particularly those with limited self-care abilities, face numerous potential risks due to their gradually declining physical functions. Falls are a common accident for the elderly, potentially leading to serious consequences such as fractures and head injuries, severely impacting their health and quality of life. Sudden illnesses such as heart attacks and strokes can be life-threatening if not detected and treated promptly. Abnormal behavior, such as prolonged immobility or unusual wandering, can also be early signs of health problems. Therefore, effectively monitoring elderly people's behavior at home, detecting abnormalities and providing early warnings, has become a major challenge in the smart home industry.

[0003] Traditional monitoring systems mostly rely on a single type of sensor to collect information. However, the information a single sensor can capture is very limited and cannot fully and accurately reflect the complex behavioral patterns of the elderly. This incomplete and one-sided information makes it difficult for the system to accurately identify the elderly's behavior, increasing the risk of missed and false alarms. Furthermore, traditional systems employ relatively simple monitoring methods and lack the ability to comprehensively analyze and process multiple data types. They often simply apply threshold judgments to the raw data collected by the sensors, triggering alarms when the data exceeds or falls below a preset threshold. This simplistic judgment method fails to account for individual differences and the diversity of elderly people's behavior patterns, and is unable to adapt to the different living habits and health conditions of the elderly. For example, walking speed and range of movement vary significantly among elderly people. Setting a uniform threshold value can further lead to false alarms of normal behavior and underreporting of abnormal behavior. To address this issue, a system and method for identifying elderly people's behavior at home and providing early warning of abnormal events is proposed. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems in the prior art and to propose a system and method for identifying the behavior of the elderly at home and warning of abnormal events.

[0005] A system and method for identifying elderly people's home behavior and warning of abnormal events, including a behavior monitoring module, a behavior analysis module, an abnormal event identification module and a warning feedback module;

[0006] The behavior monitoring module is used to collect the behavior data of the elderly in the home environment in real time based on sensors and perform data fusion;

[0007] The behavior analysis module is used to analyze the behavioral characteristics of the elderly, predict the elderly's behavioral action data at the next moment based on the time series analysis method, determine whether the elderly's behavioral action data at the next moment conforms to the normal activity pattern based on the elderly's behavioral action data at the next moment, and identify whether the elderly have abnormal behavior;

[0008] The abnormal event recognition module is used to identify abnormal events based on the behavioral data of the elderly, use machine learning methods to rate the abnormal events, and generate alarm signals based on the risk rating;

[0009] The early warning feedback module is used to automatically trigger early warning information based on the identified abnormal events and send alarms to family members and caregivers.

[0010] Preferably, the behavior monitoring module specifically includes:

[0011] Data collection unit: collects the behavioral data of the elderly in their home environment in real time through acceleration sensors, temperature and humidity sensors, door and window sensors, cameras, and infrared sensors, and standardizes and normalizes the behavioral data;

[0012] Data fusion unit: Based on the Kalman filter method, data from different sensors are fused to obtain comprehensive behavioral data;

[0013] Behavioral feature extraction unit: Based on the fused sensor data, it extracts the characteristics of the elderly's behavior, including activity frequency, gait changes, and standing and sitting status data.

[0014] Preferably, the behavior analysis module specifically includes:

[0015] Time series analysis unit: used to analyze the behavioral data of the elderly using time series analysis methods, establish an elderly behavior prediction model, use the real-time behavior characteristics of the elderly as input, and use the predicted behavior data of the elderly at the next moment as output to predict the elderly's behavior pattern at the next moment and capture behavioral changes;

[0016] Behavior pattern recognition unit: used to determine whether the elderly's historical behavior data and prediction results conform to normal activity patterns. It uses cosine similarity to measure the similarity between predicted and actual behaviors, determine the similarity between the two behavior patterns, compare the differences between the predicted values ​​and actual behaviors, and identify abnormal behavior patterns.

[0017] Among them, the cosine similarity calculation formula is:

[0018]

[0019] Where X is the vectorized predicted behavior data, R is the actual behavior data, |X| is the norm of X, |R| is the norm of R, and COS(X,R) is the similarity between the predicted behavior and the actual behavior;

[0020] Abnormal behavior detection unit: used to use support vector machine algorithm to identify whether the elderly have abnormal behavior, classify the abnormal behavior data based on the abnormal behavior data, and distinguish different types of abnormal events.

[0021] Preferably, the method for analyzing the behavior data of the elderly using a time series analysis method and establishing a behavior prediction model for the elderly specifically includes:

[0022] Among them, the formula for the elderly behavior prediction model is:

[0023] X t+1 =l1X t +l2X t-1 +…+l h X t-h +ω

[0024] Where, X t+1 is the behavioral data at time point t+1, X t is the behavior data at the current moment, X t-1 is the behavior data at time t-1 in the past, X t-h is the behavior data at the past th moment, l1 is the weight coefficient of the behavior data at the current moment, l2 is the weight coefficient of the behavior data at the past t-1 moment, l h is the weight coefficient of the behavior data at the past th moment, and ω is the error term.

[0025] Preferably, the support vector machine algorithm is used to identify whether the elderly have abnormal behavior, and based on the abnormal behavior data, the abnormal behavior data is classified into behavior categories to distinguish different types of abnormal events, specifically including:

[0026] Obtain the cosine similarity calculation results, where 1 indicates exact similarity, 0 indicates complete dissimilarity, and -1 indicates the opposite behavior pattern. Use the support vector machine algorithm to train the behavior data and obtain a classification model.

[0027] Based on the trained model, the behavioral data collected in real time is classified into abnormal behaviors to obtain abnormal behavior classification data.

[0028] Preferably, the abnormal event identification module specifically includes:

[0029] Based on abnormal behavior classification data, the risk level of each classification result is calculated through an algorithm to generate a risk level index;

[0030] Based on the generated risk level index, determine whether to activate the alarm mechanism and set the risk threshold;

[0031] Based on the risk rating given by the risk assessment unit, if the event is assessed as a high-risk level, the system automatically triggers an alarm signal to remind people to pay attention to the safety of the elderly.

[0032] Preferably, the early warning feedback module specifically includes:

[0033] Warning generation unit: generates corresponding warning information in real time based on the output of the abnormal event recognition module;

[0034] Notification sending unit: When the warning generation unit generates warning information, the notification sending unit is notified to send the information;

[0035] Response tracking unit: Receives feedback from family members and caregivers, obtains their handling status, and sends confirmation replies to the system to ensure that the crisis is handled.

[0036] A method for identifying elderly people's home behavior and early warning of abnormal events, which is implemented by a system for identifying elderly people's home behavior and early warning of abnormal events, includes the following steps:

[0037] S1. Obtain behavioral data of the elderly in their home environment through multiple sensors;

[0038] S2. Perform multi-sensor data fusion on the acquired behavioral data to obtain a unified behavioral data set;

[0039] S3. Analyze the fused behavioral data, extract behavioral features, predict the elderly’s behavioral action data at the next moment, compare it with the normal activity model, and determine whether the elderly are in a normal activity state;

[0040] S4. Detect whether the elderly have abnormal behaviors and classify abnormal events;

[0041] S5. Based on the identified abnormal events, trigger warning information to be fed back to the nursing staff

[0042] Preferably, the step S3 specifically includes:

[0043] S301. Based on the time series analysis method, establish an elderly behavior prediction model, use historical behavior data to predict the elderly's behavior at the next moment, and determine whether the prediction results conform to normal activity patterns;

[0044] S302 : Compare the predicted behavior data with the actually collected behavior data, and calculate the similarity between the behavior patterns based on cosine similarity. If the cosine similarity calculation result is between 0 and -1, it means that abnormal behavior has occurred.

[0045] Preferably, the step S4 specifically includes:

[0046] S401, dynamically monitor the elderly behavior data collected in real time to detect whether there are any abnormal behaviors that exceed the normal activity pattern;

[0047] S402. Using a support vector machine machine learning method, identify potential abnormal behaviors and automatically classify abnormal behaviors;

[0048] S403: The classification results are rated for risk, and the system determines whether to trigger an alarm mechanism based on the different risk levels. After abnormal behavior is identified, the system assigns a risk rating label to it and determines whether to trigger an alarm mechanism based on the different risk levels.

[0049] S404. When it is determined that an alarm is needed, the system automatically generates and sends an early warning signal to notify the nursing staff to take corresponding measures.

[0050] Compared with the existing technology, the advantages of the present invention are:

[0051] This invention proposes to improve the accuracy of behavior recognition through multi-sensor data fusion and time series analysis, monitor and predict the behavior of the elderly in real time, detect anomalies in time and trigger early warnings, reduce false alarms and missed alarms through intelligent identification and risk assessment based on machine learning, enhance home safety, and improve the safety and comfort of the elderly's lives. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a system framework diagram of the present invention.

[0053] Figure 2 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0054] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0055] Reference Figure 1 As shown, a system for identifying elderly people's home behavior and warning of abnormal events includes a behavior monitoring module, a behavior analysis module, an abnormal event identification module and a warning feedback module.

[0056] The behavior monitoring module is used to collect the behavior data of the elderly in the home environment in real time based on sensors and perform data fusion;

[0057] The behavior monitoring module specifically includes:

[0058] Data collection unit: collects the behavioral data of the elderly in their home environment in real time through acceleration sensors, temperature and humidity sensors, door and window sensors, cameras, and infrared sensors, and standardizes and normalizes the behavioral data;

[0059] Data fusion unit: Based on the Kalman filter method, data from different sensors are fused to obtain comprehensive behavioral data. The Kalman filter method can be used to fuse multiple sensor data, smoothing and denoising the sensor data in a dynamic environment to ensure the accuracy of the sensor data. In addition, through data standardization and normalization, the dimensional differences between different sensor data can be eliminated, thus improving the quality of data fusion.

[0060] Behavioral feature extraction unit: Based on the fused sensor data, it extracts the characteristics of the elderly's behavior, including activity frequency, gait changes, and standing and sitting status data;

[0061] The activity frequency calculation formula is:

[0062]

[0063] Where f is the activity frequency, T is the total time of observation, N is the number of data points, and a i is the acceleration data at the i-th time point, |a i | is the absolute value of acceleration at the i-th time point;

[0064] The calculation formula for gait change is:

[0065]

[0066] Where g is the gait change, N is the number of data points, and a i is the acceleration data at the i-th time point, is the average value of acceleration data.

[0067] The behavior analysis module is used to analyze the behavioral characteristics of the elderly, predict the elderly's behavioral action data at the next moment based on the time series analysis method, determine whether the elderly's behavioral action data at the next moment conforms to the normal activity pattern based on the elderly's behavioral action data at the next moment, and identify whether the elderly have abnormal behavior;

[0068] The behavior analysis module specifically includes:

[0069] Time series analysis unit: used to analyze the behavioral data of the elderly using time series analysis methods, establish an elderly behavior prediction model, use the real-time behavior characteristics of the elderly as input, and use the predicted behavior data of the elderly at the next moment as output to predict the elderly's behavior pattern at the next moment and capture behavioral changes;

[0070] Among them, the formula for the elderly behavior prediction model is:

[0071] X t+1 =l1X t +l2X t-1 +…+l h X t-h +ω

[0072] Where, X t+1 is the behavioral data at time point t+1, X t is the behavior data at the current moment, X t-1 is the behavior data at time t-1 in the past, X t-h is the behavior data at the past th moment, l1 is the weight coefficient of the behavior data at the current moment, l2 is the weight coefficient of the behavior data at the past t-1 moment, l h is the weight coefficient of the behavioral data at the past th moment, ω is the error term;

[0073] Behavior pattern recognition unit: used to determine whether the elderly's historical behavior data and prediction results conform to normal activity patterns. It uses cosine similarity to measure the similarity between predicted and actual behaviors, determine the similarity between the two behavior patterns, compare the differences between the predicted values ​​and actual behaviors, and identify abnormal behavior patterns.

[0074] Among them, the cosine similarity calculation formula is:

[0075]

[0076] Where X is the vectorized predicted behavior data, R is the actual behavior data, |X| is the norm of X, |R| is the norm of R, and COS(X,R) is the similarity between the predicted behavior and the actual behavior;

[0077] Abnormal behavior detection unit: used to identify whether the elderly have abnormal behavior using the support vector machine algorithm, classify the abnormal behavior data based on the abnormal behavior data, and distinguish different types of abnormal events;

[0078] Obtain the cosine similarity calculation results, where 1 indicates exact similarity, 0 indicates complete dissimilarity, and -1 indicates the opposite behavior pattern. Use the support vector machine algorithm to train the behavior data and obtain a classification model.

[0079] Based on the trained model, the real-time collected behavior data is classified into abnormal behavior categories to obtain abnormal behavior classification data;

[0080] The behavior analysis module uses time series analysis methods to predict the future behavior of the elderly. Combined with cosine similarity and support vector machine algorithms, it can efficiently detect abnormal behavior. By comparing historical behavior data with difference analysis of prediction results, it can quickly determine whether the elderly have abnormal behavior, so as to issue early warnings in time. By establishing an accurate behavior prediction model, it can combine the current moment and historical behavior data to predict the future behavior of the elderly. This prediction model based on historical data has a strong time series property and can more accurately reflect the behavioral patterns of the elderly, avoiding the limitations of relying solely on instantaneous data for judgment.

[0081] The abnormal event recognition module is used to identify abnormal events based on the behavioral data of the elderly, use machine learning methods to rate the abnormal events, and generate alarm signals based on the risk rating;

[0082] The abnormal event identification module specifically includes:

[0083] Based on abnormal behavior classification data, the risk level of each classification result is calculated through an algorithm to generate a risk level index;

[0084] Based on the generated risk level index, determine whether to activate the alarm mechanism and set the risk threshold;

[0085] Based on the risk rating given by the risk assessment unit, if the event is assessed as a high-risk level, the system automatically triggers an alarm signal to remind people to pay attention to the safety of the elderly;

[0086] Using the support vector machine algorithm to classify abnormal behaviors can learn the characteristics of different types of abnormal behaviors through the training process and classify them in real time. The high-precision characteristics of the support vector machine help the system extract effective information from complex data, accurately distinguish abnormal behaviors from normal activity patterns, and improve the system's ability to identify abnormal behaviors.

[0087] The early warning feedback module is used to automatically trigger early warning information based on the identified abnormal events and send alarms to family members and caregivers.

[0088] The early warning feedback module specifically includes:

[0089] Warning generation unit: generates corresponding warning information in real time based on the output of the abnormal event recognition module;

[0090] Notification sending unit: When the warning generation unit generates warning information, the notification sending unit is notified to send the information;

[0091] Response tracking unit: Receives feedback from family members and caregivers, obtains their handling status, and sends confirmation replies to the system to ensure that the crisis is handled.

[0092] The early warning feedback module can not only generate early warning information, but also has the functions of sending notifications and tracking responses, which can ensure that family members or caregivers receive early warnings in the first time and can provide feedback on the processing results. The innovation of this function lies in providing closed-loop feedback to ensure the processing effect of the event and improve the intelligence level of the system.

[0093] Reference Figure 2 As shown, a method for identifying elderly people's home behavior and warning of abnormal events includes the following steps:

[0094] S1. Obtain behavioral data of the elderly in their home environment through multiple sensors;

[0095] S2. Perform multi-sensor data fusion on the acquired behavioral data to obtain a unified behavioral data set;

[0096] S3. Analyze the fused behavioral data, extract behavioral features, predict the elderly’s behavioral action data at the next moment, compare it with the normal activity model, and determine whether the elderly are in a normal activity state;

[0097] S4. Detect whether the elderly have abnormal behaviors and classify abnormal events;

[0098] S5. Based on the identified abnormal events, trigger warning information and feedback to the nursing staff.

[0099] The step S3 specifically includes:

[0100] S301. Based on the time series analysis method, establish an elderly behavior prediction model, use historical behavior data to predict the elderly's behavior at the next moment, and determine whether the prediction results conform to normal activity patterns;

[0101] S302 : Compare the predicted behavior data with the actually collected behavior data, and calculate the similarity between the behavior patterns based on cosine similarity. If the cosine similarity calculation result is between 0 and -1, it means that abnormal behavior has occurred.

[0102] The step S4 specifically includes:

[0103] S401, dynamically monitor the elderly behavior data collected in real time to detect whether there are any abnormal behaviors that exceed the normal activity pattern;

[0104] S402. Using a support vector machine machine learning method, identify potential abnormal behaviors and automatically classify abnormal behaviors;

[0105] S403: The classification results are rated for risk, and the system determines whether to trigger an alarm mechanism based on the different risk levels. After abnormal behavior is identified, the system assigns a risk rating label to it and determines whether to trigger an alarm mechanism based on the different risk levels.

[0106] S404. When it is determined that an alarm is needed, the system automatically generates and sends an early warning signal to notify the nursing staff to take corresponding measures.

[0107] When monitoring the behavior of the elderly in real time, the system uses a combination of dynamic monitoring and machine learning to promptly detect changes in behavioral patterns and conduct risk assessments. Through automatic classification and risk assessment mechanisms, the system can take corresponding early warning measures based on different levels of risk, thereby improving the safety of the elderly's home environment.

[0108] In summary, the advantages of the present invention are:

[0109] Improve the accuracy of behavior recognition. Through multi-sensor data fusion technology, it can comprehensively consider the data of various sensors, reduce the errors and blind spots that may be caused by a single sensor, and significantly improve the accuracy of elderly behavior recognition;

[0110] The system can monitor the behavior patterns of the elderly in real time and predict their future behavior, promptly detect abnormal behavior, and trigger early warnings based on risk ratings, sending alerts to family members or caregivers to ensure that swift response measures can be taken;

[0111] By predicting the behavior of the elderly through time series analysis methods and intelligently classifying and assessing the risk of abnormal behaviors through support vector machine algorithms, efficient and accurate behavior recognition and abnormal event warning can be achieved;

[0112] The system can help family members and caregivers better monitor the elderly’s home environment, reduce health risks caused by emergencies, and improve the safety and comfort of the elderly’s lives.

[0113] By fusing data from multiple sensors and adopting advanced machine learning methods, the system can significantly reduce the probability of false alarms and missed alarms and improve system reliability compared to traditional monitoring methods.

[0114] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. A system for identifying elderly people's home behavior and warning of abnormal events, characterized by: It includes behavior monitoring module, behavior analysis module, abnormal event identification module and early warning feedback module; The behavior monitoring module is used to collect the behavior data of the elderly in the home environment in real time based on sensors and perform data fusion; The behavior analysis module is used to analyze the behavioral characteristics of the elderly, predict the elderly's behavioral action data at the next moment based on the time series analysis method, determine whether the elderly's behavioral action data at the next moment conforms to the normal activity pattern based on the elderly's behavioral action data at the next moment, and identify whether the elderly have abnormal behavior; The abnormal event recognition module is used to identify abnormal events based on the behavioral data of the elderly, use machine learning methods to rate the abnormal events, and generate alarm signals based on the risk rating; The early warning feedback module is used to automatically trigger early warning information based on the identified abnormal events and send alarms to family members and caregivers.

2. The elderly home behavior recognition and abnormal event warning system according to claim 1 is characterized by: The behavior monitoring module specifically includes: Data collection unit: collects the behavioral data of the elderly in their home environment in real time through acceleration sensors, temperature and humidity sensors, door and window sensors, cameras, and infrared sensors, and standardizes and normalizes the behavioral data; Data fusion unit: Based on the Kalman filter method, data from different sensors are fused to obtain comprehensive behavioral data; Behavioral feature extraction unit: Based on the fused sensor data, it extracts the characteristics of the elderly's behavior, including activity frequency, gait changes, and standing and sitting status data.

3. The elderly home behavior recognition and abnormal event warning system according to claim 1 is characterized by: The behavior analysis module specifically includes: Time series analysis unit: used to analyze the behavioral data of the elderly using time series analysis methods, establish an elderly behavior prediction model, use the real-time behavior characteristics of the elderly as input, and use the predicted behavior data of the elderly at the next moment as output to predict the elderly's behavior pattern at the next moment and capture behavioral changes; Behavior pattern recognition unit: used to determine whether the elderly person's historical behavior data and prediction results conform to normal activity patterns. It uses cosine similarity to measure the similarity between predicted and actual behaviors, determine the similarity between the two behavior patterns, compare the differences between the predicted values ​​and actual behaviors, and identify abnormal behavior patterns. Among them, the cosine similarity calculation formula is: Where X is the vectorized predicted behavior data, R is the actual behavior data, |X| is the norm of X, |R| is the norm of R, and COS(X,R) is the similarity between the predicted behavior and the actual behavior; Abnormal behavior detection unit: used to use support vector machine algorithm to identify whether the elderly have abnormal behavior, classify the abnormal behavior data based on the abnormal behavior data, and distinguish different types of abnormal events.

4. The elderly home behavior recognition and abnormal event warning system according to claim 3 is characterized by: The method is used to analyze the behavior data of the elderly using a time series analysis method and establish a behavior prediction model for the elderly. include: Among them, the formula for the elderly behavior prediction model is: X t+1 =l1X t +l2X t-1 +…+l h X t-h +ω Where, X t+1 is the behavioral data at time point t+1, X t is the behavior data at the current moment, X t-1 is the behavior data at time t-1 in the past, X t-h is the behavior data at the past th moment, l1 is the weight coefficient of the behavior data at the current moment, l2 is the weight coefficient of the behavior data at the past t-1 moment, l h is the weight coefficient of the behavior data at the past th moment, and ω is the error term.

5. The elderly home behavior recognition and abnormal event warning system according to claim 4 is characterized by: The support vector machine algorithm is used to identify whether the elderly have abnormal behavior, and based on the abnormal behavior data, the abnormal behavior data is classified into different types of abnormal events, specifically including: Obtain the cosine similarity calculation results, where 1 indicates exact similarity, 0 indicates complete dissimilarity, and -1 indicates the opposite behavior pattern. Use the support vector machine algorithm to train the behavior data and obtain a classification model. Based on the trained model, the behavioral data collected in real time is classified into abnormal behaviors to obtain abnormal behavior classification data.

6. The elderly home behavior recognition and abnormal event warning system according to claim 1 is characterized by: The abnormal event identification module specifically includes: Based on abnormal behavior classification data, the risk level of each classification result is calculated through an algorithm to generate a risk level index; Based on the generated risk level index, determine whether to activate the alarm mechanism and set the risk threshold; Based on the risk rating given by the risk assessment unit, if the event is assessed as a high-risk level, the system automatically triggers an alarm signal to remind people to pay attention to the safety of the elderly.

7. The elderly home behavior recognition and abnormal event warning system according to claim 1 is characterized by: The early warning feedback module specifically includes: Warning generation unit: generates corresponding warning information in real time based on the output of the abnormal event recognition module; Notification sending unit: When the warning generation unit generates warning information, the notification sending unit is notified to send the information; Response tracking unit: Receives feedback from family members and caregivers, obtains their handling status, and sends confirmation replies to the system to ensure that the crisis is handled.

8. A method for identifying elderly people's home behavior and early warning of abnormal events, the method being implemented by the elderly people's home behavior identification and early warning of abnormal events according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1. Obtain behavioral data of the elderly in their home environment through multiple sensors; S2. Perform multi-sensor data fusion on the acquired behavioral data to obtain a unified behavioral data set; S3. Analyze the fused behavioral data, extract behavioral features, predict the elderly’s behavioral action data at the next moment, compare it with the normal activity model, and determine whether the elderly are in a normal activity state; S4. Detect whether the elderly have abnormal behaviors and classify abnormal events; S5. Based on the identified abnormal events, trigger warning information and feedback to the nursing staff.

9. The method for identifying elderly people's home behavior and early warning of abnormal events according to claim 8, characterized in that: The step S3 specifically includes: S301. Based on the time series analysis method, establish an elderly behavior prediction model, use historical behavior data to predict the elderly's behavior at the next moment, and determine whether the prediction results conform to normal activity patterns; S302 : Compare the predicted behavior data with the actually collected behavior data, and calculate the similarity between the behavior patterns based on cosine similarity. If the cosine similarity calculation result is between 0 and -1, it means that abnormal behavior has occurred.

10. The method for identifying elderly people's home behavior and early warning of abnormal events according to claim 8, characterized in that: The step S4 specifically includes: S401, dynamically monitor the elderly behavior data collected in real time to detect whether there are any abnormal behaviors that exceed the normal activity pattern; S402. Using a support vector machine machine learning method, identify potential abnormal behaviors and automatically classify abnormal behaviors; S403: The classification results are rated for risk, and the system determines whether to trigger an alarm mechanism based on the different levels of risk. After abnormal behavior is identified, the system assigns a risk rating label to it and determines whether to trigger an alarm mechanism based on the different levels of risk. S404. When it is determined that an alarm is needed, the system automatically generates and sends an early warning signal to notify the nursing staff to take corresponding measures.

Images