Safety monitoring and warning method and system for elderly people living alone based on data analysis

Abstract

The invention discloses a safety monitoring and warning method and system for elderly people living alone based on data analysis, and relates to the technical field of intelligent monitoring, and the method comprises the steps: collecting the image data of the daily life of the elderly people living alone in an indoor environment in real time; monitoring the physiological sign data of the elderly living alone in real time; performing feature extraction on the acquired daily life image data of the elderly living alone by using a convolutional neural network to obtain human body image features; analyzing the monitored physiological sign data of the elderly living alone by using a recurrent neural network to obtain abnormal state reference data; according to the human body image features and the abnormal state reference data, identifying various abnormal states of the elderly living alone in the home environment; under the condition that various abnormal states of the elderly living alone in the home environment are identified, sending an alarm message to a remote guardian terminal; and the remote guardian terminal receives the alarm message.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring technology, and more specifically, to a method and system for safety monitoring and alarming of elderly people living alone based on data analysis. Background Art

[0002] The issue of guardianship for elderly people living alone has gradually become a focus of social attention. Due to the lack of daily care and guardianship, elderly people living alone are often at higher risk when facing health problems or accidental injuries. Whether their abnormal conditions can be discovered in time is directly related to their life safety.

[0003] At present, the monitoring system for elderly people living alone mainly focuses on the elderly's physical parameters, such as heart rate, blood pressure, etc. If the elderly's heart rate or blood pressure is abnormal, it is considered that the elderly is in an abnormal state. However, this method is relatively simple in identifying the abnormal state of the elderly and only achieves basic health monitoring.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method and system for safety monitoring and alarming of elderly people living alone based on data analysis to solve the above-mentioned technical problems.

[0006] This application provides a safety monitoring and alarm method for elderly people living alone based on data analysis, including:

[0007] Real-time collection of image data of daily life of elderly people living alone in indoor environments;

[0008] Real-time monitoring of the physiological sign data of the elderly person living alone;

[0009] The convolutional neural network is used to extract features from the image data of the daily life of the elderly living alone, and the human image features are obtained; the recurrent neural network is used to analyze the physiological sign data of the elderly living alone, and the abnormal state reference data is obtained;

[0010] According to the human body image features and the abnormal state reference data, multiple abnormal states of the elderly living alone in the home environment are identified; when multiple abnormal states of the elderly living alone in the home environment are identified, an alarm message is sent to a remote guardian terminal;

[0011] The remote guardian terminal receives the alarm message.

[0012] The present application provides a safety monitoring and alarm system for elderly people living alone based on data analysis, including an image acquisition device, a physiological parameter monitoring device, an abnormal state recognition module for elderly people living alone, and a remote guardian terminal; wherein,

[0013] The image acquisition device is used to collect image data of the daily life of elderly people living alone in an indoor environment in real time;

[0014] The physiological parameter monitoring device is used to monitor the physiological sign data of the elderly person living alone in real time;

[0015] The abnormal state recognition module for the elderly living alone is connected to the image acquisition device and the physiological parameter monitoring device, and is used to perform the following steps: using a convolutional neural network to extract features from the image data of the daily life of the elderly living alone collected to obtain human image features; using a recurrent neural network to analyze the monitored physiological sign data of the elderly living alone to obtain abnormal state reference data; based on the human image features and the abnormal state reference data, identifying multiple abnormal states of the elderly living alone in the home environment; in the case of identifying multiple abnormal states of the elderly living alone in the home environment, sending an alarm message to the remote guardian terminal;

[0016] Based on the embodiment provided by the present application, the image data of the daily life of the elderly living alone in the indoor environment is collected in real time; the physiological sign data of the elderly living alone is monitored in real time; the image data of the daily life of the elderly living alone collected is extracted by using a convolutional neural network to obtain the human image features; the physiological sign data of the elderly living alone monitored is analyzed by using a recurrent neural network to obtain abnormal state reference data; according to the human image features and the abnormal state reference data, the various abnormal states of the elderly living alone in the home environment are identified; when the various abnormal states of the elderly living alone in the home environment are identified, an alarm message is sent to the remote guardian terminal; the remote guardian terminal receives the alarm message. Thus, it is realized that not only the physical sign parameters of the elderly are paid attention to, but also the monitoring of sudden abnormal conditions is increased, so that the identification of the abnormal state of the elderly living alone is more comprehensive, not limited to the monitoring of a single health parameter; through the real-time intelligent identification of the abnormal state of the elderly living alone, the system can issue an alarm in time, so that the elderly can get timely assistance when they are in an abnormal state, and reduce the serious consequences caused by no one noticing; the effective application of this system can greatly alleviate the pressure of government departments and families in the elderly care, and has important social value and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 This is a flow chart of an optional data analysis-based safety monitoring and alarm method for elderly people living alone according to an embodiment of the present application;

[0019] Figure 2 The present invention is a structural diagram of an optional data analysis-based safety monitoring and alarm system for elderly people living alone according to an embodiment of the present application.

[0020] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0022] Alternatively, if Figure 1 As shown, the present application provides a safety monitoring and alarm method for elderly people living alone based on data analysis, including:

[0023] S101, real-time collection of image data of daily life of elderly people living alone in indoor environments;

[0024] S102, real-time monitoring of physiological sign data of elderly people living alone;

[0025] S103, using a convolutional neural network to extract features from the collected image data of the daily life of the elderly living alone, to obtain human image features; using a recurrent neural network to analyze the monitored physiological sign data of the elderly living alone, to obtain abnormal state reference data;

[0026] S104, identifying various abnormal states of the elderly living alone in the home environment according to the human image features and the abnormal state reference data; when identifying various abnormal states of the elderly living alone in the home environment, sending an alarm message to the remote guardian terminal;

[0027] S105, the remote guardian terminal receives the alarm message.

[0028] Among them, the various abnormal conditions include not only abnormal physical sign parameters, but also falls, low back pain, coughing, headache, vomiting, staggering and abnormal wandering, etc., thus more comprehensively covering the various abnormal conditions that may be encountered by elderly people living alone.

[0029] Based on the embodiment provided by the present application, the image data of the daily life of the elderly living alone in the indoor environment is collected in real time; the physiological sign data of the elderly living alone is monitored in real time; the image data of the daily life of the elderly living alone collected is extracted by using a convolutional neural network to obtain the human image features; the physiological sign data of the elderly living alone monitored is analyzed by using a recurrent neural network to obtain abnormal state reference data; according to the human image features and the abnormal state reference data, the various abnormal states of the elderly living alone in the home environment are identified; when the various abnormal states of the elderly living alone in the home environment are identified, an alarm message is sent to the remote guardian terminal; the remote guardian terminal receives the alarm message. Thus, it is realized that not only the physical sign parameters of the elderly are paid attention to, but also the monitoring of sudden abnormal conditions is increased, so that the identification of the abnormal state of the elderly living alone is more comprehensive, not limited to the monitoring of a single health parameter; through the real-time intelligent identification of the abnormal state of the elderly living alone, the system can issue an alarm in time, so that the elderly can get timely assistance when they are in an abnormal state, and reduce the serious consequences caused by no one noticing; the effective application of this system can greatly alleviate the pressure of government departments and families in the elderly care, and has important social value and application prospects.

[0030] Furthermore, the physiological sign data includes pulse data, respiration data, blood pressure data, body temperature data and blood oxygen saturation. Pulse data includes heart rate and pulse rate; respiration data includes respiration rate, tidal volume and vital capacity; blood pressure data includes systolic pressure, diastolic pressure and pulse pressure; body temperature data includes core body temperature and skin temperature; blood oxygen saturation includes arterial blood oxygen saturation and functional oxygen saturation. Optionally, as Figure 2 As shown, the present application provides a safety monitoring and alarm system for elderly people living alone based on data analysis, including an image acquisition device 201, a physiological parameter monitoring device 202, an abnormal state recognition module for elderly people living alone 203, and a remote guardian terminal 204; wherein,

[0031] Image acquisition device 201, used to collect image data of daily life of elderly people living alone in indoor environment in real time;

[0032] A physiological parameter monitoring device 202 is used to monitor the physiological sign data of the elderly living alone in real time;

[0033] The abnormal state recognition module 203 for the elderly living alone is connected to the image acquisition device and the physiological parameter monitoring device. The abnormal state recognition module for the elderly living alone is used to perform the following steps: using a convolutional neural network to extract features from the image data of the daily life of the elderly living alone collected to obtain human image features; using a recurrent neural network to analyze the monitored physiological sign data of the elderly living alone to obtain abnormal state reference data; based on the human image features and the abnormal state reference data, identifying various abnormal states of the elderly living alone in the home environment; in the case of identifying various abnormal states of the elderly living alone in the home environment, sending an alarm message to the remote guardian terminal 204;

[0034] The remote guardian terminal 204 is used to receive the alarm message sent by the abnormal state identification module for the elderly living alone.

[0035] Furthermore, the abnormal state recognition module for elderly people living alone recognizes various abnormal states of elderly people living alone in their home environment based on human image features and abnormal state reference data, and is configured as follows:

[0036] Extracting local features of parts of human body images from human body image features;

[0037] The gated recurrent unit network is used to learn the temporal dependency of time series data from human image features and capture temporal dependency features;

[0038] The extracted local features of the part are deeply integrated with the time-dependent features to form comprehensive spatiotemporal features;

[0039] Initialize the network structure of the variational autoencoder; wherein the network structure of the variational autoencoder includes an encoder and a decoder; the encoder is used to map the comprehensive spatiotemporal features and abnormal state reference data to a latent space, and the decoder is used to reconstruct the comprehensive spatiotemporal features and abnormal state reference data from the latent space;

[0040] Initialize the network structure of the coupled generative adversarial network; wherein the network structure of the coupled generative adversarial network includes two generators and two discriminators, wherein the first generator is used to generate a corresponding abnormal state image in a latent space according to the output of the encoder, and the second generator is used to generate a corresponding human image feature in a latent space according to the abnormal state reference data; the first generator and the second generator share a feature representation of the latent space; the first discriminator network is used to evaluate the authenticity of the abnormal state image generated by the first generator, and the second discriminator network is used to evaluate the authenticity of the human image feature generated by the second generator;

[0041] By using the adversarial training mechanism of coupled generative adversarial networks and variational autoencoders, the network structures of variational autoencoders and coupled generative adversarial networks are optimized with the goal of minimizing reconstruction error and KL divergence.

[0042] Based on the multimodal adversarial loss function, the network structure of the coupled generative adversarial network and the network structure of the variational autoencoder are optimized twice; wherein the multimodal adversarial loss function is used to combine image data and physiological data to generate adversarial loss, and reconstruct adversarial loss;

[0043] According to the optimized variational autoencoder and coupled generative adversarial network, the following difference metrics are calculated: the difference metric between the comprehensive spatiotemporal features and the reconstructed comprehensive spatiotemporal features, called the first difference metric; the difference metric between the abnormal state reference data and the reconstructed abnormal state reference data, called the second difference metric;

[0044] Through a deep learning model enhanced by the attention mechanism, abnormal patterns are identified based on the life scenarios of elderly people living alone;

[0045] The abnormal behavior pattern mining network is used to mine potential abnormal patterns. The abnormal behavior pattern mining network uses a graph convolutional network to analyze the graph structure of the behavior patterns of elderly people living alone to mine potential abnormal patterns.

[0046] According to the identified abnormal patterns and potential abnormal patterns, a fusion algorithm based on evidence theory is used to comprehensively consider multi-source difference information to identify various abnormal states of elderly people living alone in their home environment.

[0047] Among them, the abnormal state recognition module for elderly people living alone can dynamically adjust the parameters of the coupled generative adversarial network based on the real-time updated comprehensive spatiotemporal characteristics and abnormal state reference data to adapt to the changes in the behavior patterns of elderly people living alone.

[0048] Based on the embodiments provided in the present application, by integrating the local features and time-dependent features of human body images, it is possible to more accurately identify the abnormal conditions of elderly people living alone, including falls, back pain, coughing, headaches, vomiting, staggering, and abnormal wandering, etc.; by using variational autoencoders and coupled generative adversarial networks, the system can update and adapt to changes in the behavior patterns of elderly people living alone in real time, providing continuous monitoring and timely abnormal state identification; combining image data and physiological data, the multimodal adversarial loss function can improve the system's ability to recognize abnormal states, especially in complex environments; through deep fusion technology, the system can form comprehensive spatiotemporal features, which helps to capture subtle changes in the behavior of elderly people living alone and improve the sensitivity of anomaly detection; the adversarial training mechanism of the coupled generative adversarial network can improve the generalization ability of the model, reduce overfitting, and enhance the model in practical applications. Stability and reliability; through the deep learning model enhanced by the attention mechanism, the system can more accurately identify and focus on abnormal patterns in the life scenes of elderly people living alone; by using the graph convolutional network to mine potential abnormal behavior patterns, the system can provide early warning of possible health risks and safety issues; using the fusion algorithm based on evidence theory, the system can comprehensively consider multi-source difference information and improve the accuracy and robustness of identifying multiple abnormal states of elderly people living alone; this patented technology can provide personalized health monitoring for elderly people living alone, and provide customized care and intervention by analyzing individual behavior patterns; by timely identifying the abnormal states of elderly people living alone, this technology helps to reduce the occurrence of emergency medical events and improve the quality of life and safety of the elderly; through preventive monitoring and intervention, this technology helps to reduce the high medical costs caused by emergency medical events.

[0049] Based on the above analysis, this application can effectively improve the home safety of elderly people living alone. By promptly identifying abnormal conditions and notifying guardians, it reduces the risk of accidents for elderly people living alone.

[0050] Furthermore, the network structure of the variational autoencoder is optimized based on the following formula:

[0051]

[0052] Where L is the loss function of the variational autoencoder; In the example, E is the abbreviation of the expected reconstruction error, which represents the weighted average of the function of the latent variable z; the weighting is based on the latent variable z under the given input X temporal , r ref The probability distribution q(z|X temporal , r ref ); the latent variable z is the hidden variable in the variational autoencoder, representing the input X temporal , r ref Representation in latent space; q(z|X temporal , rref ) represents the conditional probability distribution of the latent variable z output by the encoder; X temporal Represents the comprehensive spatiotemporal characteristics; r ref Indicates abnormal state reference data; p(X temporal , r ref |z) represents the reconstructed X after a given latent variable z is output by the decoder. temporal , r ref The conditional probability of ; p(z) represents the prior distribution of the latent variable z; || is used to represent the probability of z in X tempooral , r ref The probability distribution under the condition of ; KL is the Kullback-Leibler divergence, which is used to measure the difference between probability distributions. Furthermore, the abnormal pattern is a behavior pattern that is different from the normal behavior pattern and is identified from the deep learning model enhanced by the attention mechanism;

[0053] According to the identified abnormal patterns and potential abnormal patterns, a fusion algorithm based on evidence theory is used to comprehensively consider multi-source difference information to identify various abnormal states of elderly people living alone in their home environment. It is configured as follows:

[0054] Based on the following formula, the difference measure is converted into a basic probability distribution according to the identified abnormal pattern and potential abnormal pattern;

[0055]

[0056]

[0057] Among them, D i (A) represents the i-th difference D i Abnormal state A; m1(A) is a first basic probability distribution of abnormal state A based on the first difference D1; m2(A) is a second basic probability distribution of abnormal state A based on the second difference D2;

[0058] The first basic probability distribution and the second basic probability distribution are combined using the Dempster-Shafer combination rule to obtain a global basic probability distribution;

[0059] According to the global basic probability distribution, the confidence of various abnormal states of elderly people living alone in the home environment is calculated to identify various abnormal states of elderly people living alone in the home environment.

[0060] Based on the embodiments provided by the present application, by converting the difference metric into a basic probability distribution, it is possible to provide accurate basic probability distributions for the abnormal states of elderly people living alone, which reflect the degree of trust in abnormal patterns, thereby improving the accuracy of abnormal detection; the first basic probability distribution and the second basic probability distribution are combined using the Dempster-Shafer combination rule to obtain a global basic probability distribution. This method comprehensively considers multi-source difference information and improves the comprehensiveness and reliability of identifying various abnormal states in the home environment of elderly people living alone; according to the global basic probability distribution, the confidence of elderly people living alone in various abnormal states in the home environment is calculated. This calculation method can more accurately evaluate the safety status of the elderly and provide guardians with a more reliable basis for decision-making; through the deep learning model enhanced by the attention mechanism, this patent can identify abnormal patterns based on the life scenes of elderly people living alone. This pattern recognition method improves the system's understanding of the elderly's behavior and the ability to identify abnormal states; the fusion algorithm based on evidence theory optimizes the decision-making process in an uncertain environment. This method has broad application prospects in the fields of multi-source information fusion and uncertainty reasoning.

[0061] Furthermore, the physiological sign data includes pulse data, respiration data, blood pressure data, body temperature data and blood oxygen saturation.

[0062] Furthermore, the safety monitoring and alarm system for elderly people living alone based on data analysis also includes:

[0063] The adaptive threshold adjustment module is used to adjust the recognition criteria for abnormal patterns according to the living habits and medical history of elderly people living alone, so as to adapt to the individual differences in behavioral and physiological parameters of elderly people living alone.

[0064] Furthermore, the safety monitoring and alarm system for elderly people living alone based on data analysis also includes:

[0065] The data security module is used to encrypt the data transmitted and stored in the safety monitoring and alarm system for elderly people living alone based on data analysis.

[0066] It should be noted that in the present application, the embodiments implemented by the system for safety monitoring and alarming of elderly people living alone based on data analysis can be referenced with the embodiments implemented by the method for safety monitoring and alarming of elderly people living alone based on data analysis, and the present application will not elaborate on them one by one.

[0067] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A safety monitoring and alarm method for elderly people living alone based on data analysis, characterized in that: include: Real-time collection of image data of daily life of elderly people living alone in indoor environments; Real-time monitoring of the physiological sign data of the elderly person living alone; The convolutional neural network is used to extract features from the image data of the daily life of the elderly living alone, and the human image features are obtained; the recurrent neural network is used to analyze the physiological sign data of the elderly living alone, and the abnormal state reference data is obtained; According to the human body image features and the abnormal state reference data, multiple abnormal states of the elderly living alone in the home environment are identified; when multiple abnormal states of the elderly living alone in the home environment are identified, an alarm message is sent to a remote guardian terminal; The remote guardian terminal receives the alarm message.

2. The data analysis-based safety monitoring and warning method for elderly people living alone according to claim 1 is characterized in that: The physiological sign data include pulse data, respiration data, blood pressure data, body temperature data and blood oxygen saturation.

3. The data analysis-based safety monitoring and warning method for elderly people living alone according to claim 2 is characterized in that: The pulse data includes heart rate and pulse rate; the respiratory data includes respiratory rate, tidal volume, and vital capacity; the blood pressure data includes systolic pressure, diastolic pressure, and pulse pressure; the body temperature data includes core body temperature and skin temperature; and the blood oxygen saturation includes arterial oxygen saturation and functional oxygen saturation.

4. A data analysis-based safety monitoring and alarm system for elderly people living alone, the system implements the data analysis-based safety monitoring and alarm method for elderly people living alone as claimed in claim 1, characterized in that: It includes an image acquisition device, a physiological parameter monitoring device, an abnormal state recognition module for elderly people living alone, and a remote guardian terminal; among which, The image acquisition device is used to collect image data of the daily life of elderly people living alone in an indoor environment in real time; The physiological parameter monitoring device is used to monitor the physiological sign data of the elderly person living alone in real time; The abnormal state recognition module for the elderly living alone is connected to the image acquisition device and the physiological parameter monitoring device, and is used to perform the following steps: using a convolutional neural network to extract features from the image data of the daily life of the elderly living alone collected to obtain human image features; using a recurrent neural network to analyze the monitored physiological sign data of the elderly living alone to obtain abnormal state reference data; based on the human image features and the abnormal state reference data, identifying multiple abnormal states of the elderly living alone in the home environment; in the case of identifying multiple abnormal states of the elderly living alone in the home environment, sending an alarm message to the remote guardian terminal; The remote guardian terminal is used to receive the warning message sent by the abnormal state identification module for the elderly living alone.

5. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 4 is characterized in that: The abnormal state recognition module for the elderly living alone recognizes multiple abnormal states of the elderly living alone in the home environment according to the human image features and the abnormal state reference data, and is configured as follows: Extracting local features of parts of the human body image from the human body image features; Using a gated recurrent unit network to learn the temporal dependency of time series data from the human image features to capture temporal dependency features; Deeply fusing the extracted local features of the part with the time-dependent features to form comprehensive spatiotemporal features; Initializing a network structure of a variational autoencoder; wherein the network structure of the variational autoencoder comprises an encoder and a decoder; the encoder is used to map the comprehensive spatiotemporal features and the abnormal state reference data to a latent space, and the decoder is used to reconstruct the comprehensive spatiotemporal features and the abnormal state reference data from the latent space; Initialize a network structure of a coupled generative adversarial network; wherein the network structure of the coupled generative adversarial network includes two generators and two discriminators, wherein the first generator is used to generate a corresponding abnormal state image in the latent space according to the output of the encoder, and the second generator is used to generate a corresponding human image feature in the latent space according to the abnormal state reference data; the first generator and the second generator share a feature representation of the latent space; the first discriminator network is used to evaluate the authenticity of the abnormal state image generated by the first generator, and the second discriminator network is used to evaluate the authenticity of the human image feature generated by the second generator; Utilizing the adversarial training mechanism of the coupled generative adversarial network and the variational autoencoder, the network structure of the variational autoencoder and the network structure of the coupled generative adversarial network are optimized with the goal of minimizing reconstruction error and KL divergence; Based on a multimodal adversarial loss function, a network structure of the coupled generative adversarial network and a network structure of the variational autoencoder are secondary optimized; wherein the multimodal adversarial loss function is used to generate adversarial loss in combination with image data and physiological data, and to reconstruct the adversarial loss; According to the optimized variational autoencoder and coupled generative adversarial network, the following difference metrics are calculated: the difference metric between the comprehensive spatiotemporal features and the reconstructed comprehensive spatiotemporal features is called the first difference metric; the difference metric between the abnormal state reference data and the reconstructed abnormal state reference data is called the second difference metric; Through the deep learning model enhanced by the attention mechanism, abnormal patterns are identified based on the life scenes of the elderly living alone; Using an abnormal behavior pattern mining network to mine potential abnormal patterns; wherein the abnormal behavior pattern mining network uses a graph convolutional network to analyze the graph structure of the behavior pattern of the elderly living alone to mine the potential abnormal patterns; According to the identified abnormal patterns and potential abnormal patterns, a fusion algorithm based on evidence theory is adopted to comprehensively consider multi-source difference information to identify various abnormal states of the elderly living alone in their home environment; Among them, the abnormal state identification module for elderly people living alone can dynamically adjust the parameters of the coupled generative adversarial network according to the comprehensive spatiotemporal characteristics and abnormal state reference data updated in real time to adapt to the changes in the behavior patterns of elderly people living alone.

6. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 5 is characterized in that: The network structure of the variational autoencoder is optimized based on the following formula: Wherein, L is the loss function of the variational autoencoder; In the example, E is the abbreviation of the expected reconstruction error, which represents the weighted average of the function of the latent variable z; the weighting is based on the latent variable z under the given input X temporal ,r ref The probability distribution q(z|X temporal ,r ref ); the latent variable z is the hidden variable in the variational autoencoder, representing the input X temporal ,r ref Representation in latent space; q(z|X temporal ,r ref ) represents the conditional probability distribution of the latent variable z output by the encoder; X temporal Represents the comprehensive spatiotemporal characteristics; r ref represents the abnormal state reference data; p(X temporal ,r ref |z) represents the decoder output given the latent variable z, the reconstruction of X temporal ,r ref The conditional probability of ; p(z) represents the prior distribution of the latent variable z; || is used to represent the probability of z in X temporal ,r ref The probability distribution under the condition of ; KL is the Kullback-Leibler divergence, which is used to measure the difference between probability distributions.

7. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 5 is characterized in that: The abnormal pattern is a behavior pattern different from the normal behavior pattern identified from the deep learning model enhanced by the attention mechanism; the method adopts a fusion algorithm based on the evidence theory based on the identified abnormal pattern and the potential abnormal pattern, comprehensively considers multi-source difference information, and identifies multiple abnormal states of the elderly living alone in the home environment, and is configured as follows: Based on the following formula, the difference measure is converted into a basic probability distribution according to the identified abnormal pattern and potential abnormal pattern; Among them, D i (A) represents the i-th difference D i m1(A) is a first basic probability distribution of the abnormal state A based on the first difference D1; m2(A) is a second basic probability distribution of the abnormal state A based on the second difference D2; Using the Dempster-Shafer combination rule to combine the first basic probability distribution and the second basic probability distribution to obtain a global basic probability distribution; According to the global basic probability distribution, the confidence that the elderly person living alone is in various abnormal states in the home environment is calculated to identify the various abnormal states of the elderly person living alone in the home environment.

8. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 4 is characterized in that: The physiological sign data include pulse data, respiration data, blood pressure data, body temperature data and blood oxygen saturation.

9. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 5 is characterized in that: The safety monitoring and alarm system for elderly people living alone also includes: The adaptive threshold adjustment module is used to adjust the recognition criteria of abnormal patterns according to the living habits and medical history of the elderly living alone, so as to adapt to the individual differences in the behavior and physiological parameters of the elderly living alone.

10. The data analysis-based safety monitoring and warning system for elderly people living alone according to claim 4 is characterized in that: The safety monitoring and alarm system for elderly people living alone also includes: The data security module is used to encrypt the data transmitted and stored in the safety monitoring and alarm system for elderly people living alone based on data analysis.

Images