Nursing behavior recognition and intervention system based on artificial intelligence

Abstract

The present invention relates to the field of artificial intelligence, and specifically refers to a caregiver behavior recognition and intervention system based on artificial intelligence, including integrated equipment, data acquisition and preprocessing modules, behavior recognition and analysis modules, intervention and response modules, user interface modules and security and privacy modules. The present invention proposes that in the behavior recognition and analysis module, a COA-CNN-LSTM model is used to analyze the preprocessed data, identify and classify the behaviors of caregivers and patients. The COA-CNN-LSTM model combines the characteristics of convolutional neural networks and long short-term memory networks, and improves the recognition accuracy of the behaviors of caregivers and patients. In the behavior recognition and analysis module, a VGG-ICNN model is used as a basic architecture to construct a behavior recognition model, thereby improving the feature representation capability of the behavior pattern, better understanding and predicting the behaviors of caregivers and patients, and improving the intelligence level and service quality of the system.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an artificial intelligence-based caregiver behavior recognition and intervention system. Background Art

[0002] As the aging of the population increases, the demand for care continues to increase, and the workload and work load of caregivers also increase accordingly. Therefore, it becomes increasingly important to monitor the behavior of caregivers. In general caregiver behavior recognition and intervention systems, the use of ordinary models for behavior analysis has problems such as low recognition accuracy, poor generalization ability and low efficiency. It cannot fully utilize the rich information in image and time series data, resulting in information loss and reduced accuracy of behavior recognition. It lacks the ability to adapt to data features in different environments and scenarios, and the ability to analyze and intervene in the personalized behavioral characteristics of different caregivers and patients. It is impossible to provide personalized nursing services, affecting the quality and efficiency of nursing. The general VGG network architecture has problems such as insufficient feature extraction ability, incomplete information representation and low computational efficiency. It cannot capture important features in the behavior pattern, thereby reducing the accuracy of behavior recognition. It does not introduce the model structure of attention mechanism, residual connection and dense connection, resulting in insufficient extraction of important information in the behavior pattern, affecting the depth and accuracy of behavior pattern analysis. Summary of the invention

[0003] In view of the above situation, the present invention provides a care worker behavior recognition and intervention system based on artificial intelligence. In general care worker behavior recognition and intervention systems, the use of ordinary models for behavior analysis has the problems of low recognition accuracy, poor generalization ability and low efficiency. In the behavior recognition and analysis module, the present invention uses the COA-CNN-LSTM model to analyze the preprocessed data, identify the behaviors of care workers and patients and classify them. The COA-CNN-LSTM model combines the characteristics of convolutional neural networks and long short-term memory networks, and can effectively capture the spatial and temporal features in image and time series data, thereby improving the recognition accuracy of care workers and patient behaviors. At the same time, the COA algorithm is introduced to optimize the COA-CNN-LSTM model parameters, so that the model can adapt to the data features in different environments and scenarios, improve the generalization ability, and Personalized intervention and response are carried out according to the behavioral characteristics of caregivers and patients, thereby improving the quality and efficiency of nursing care. In view of the problems of insufficient feature extraction capability, incomplete information representation and low computational efficiency in the general VGG network architecture, the present invention uses the VGG-ICNN model as the basic architecture in the behavior recognition and analysis module to construct a behavior recognition model. The deep convolutional network can effectively extract image features, and the introduction of the attention mechanism further strengthens the focus on important features, thereby improving the feature representation capability of the behavior pattern. The residual connection helps the model to better transmit gradient information, solving the gradient vanishing and gradient explosion problems in deep network training. The dense connection enables the model to perform feature fusion at different levels, effectively combines the features extracted at different levels, better understands and predicts the behavior of caregivers and patients, and improves the intelligence level and service quality of the system.

[0004] The artificial intelligence-based caregiver behavior recognition and intervention system provided by the present invention includes an integrated device, a data acquisition and preprocessing module, a behavior recognition and analysis module, an intervention and response module, a user interface module, and a security and privacy module, specifically including the following contents:

[0005] The integrated device includes a housing, a monitoring camera, a sound sensor, a temperature sensor, two motion sensors and a computing device. The housing is made of waterproof and dustproof polyethylene material and has a size of 5cm*5cm*5cm. The monitoring camera is located at the central outer bottom of the housing to detect visual information of the caregiver and the patient. The sound sensor is installed on the left side of the housing to detect sound information in the environment. The temperature sensor is installed on the right side of the housing to detect ambient temperature data. Two motion sensors are installed on the front side and the rear side of the housing to detect movement and posture change information of the caregiver. The monitoring camera, the sound sensor, the temperature sensor and the two motion sensors are connected to the computing device via USB interfaces respectively.

[0006] The data acquisition and preprocessing module receives visual information of the caregiver and the patient, sound information in the environment, ambient temperature data, and information on the movement and posture change of the caregiver detected by the integrated device, wherein the visual information of the caregiver and the patient includes the position, movement, expression, eye contact, and interactive movement of the caregiver and the patient, the sound information in the environment includes the conversation between the caregiver and the patient, the patient's breathing, the sound of moving objects, and the ambient noise, and the posture change information includes the posture type, posture duration, and posture switching frequency of the caregiver and the patient, and preprocesses the visual information of the caregiver and the patient, the sound information in the environment, the ambient temperature data, and the movement and posture change information of the caregiver to obtain preprocessed data;

[0007] The behavior recognition and analysis module uses the COA-CNN-LSTM model to analyze the preprocessed data, identifies and classifies the behaviors of caregivers and patients, including walking, standing, sitting and lying down, obtains the behavior pattern data of caregivers and patients, and uses the VGG-ICNN model to analyze the behavior pattern data of caregivers and patients to obtain behavior analysis results;

[0008] The intervention and response module receives the behavior pattern data of the caregiver and the patient obtained by the behavior recognition and analysis module and takes intervention measures;

[0009] The user interface module provides an interface used by users to monitor the operating status, view the behavior analysis results, and set the COA-CNN-LSTM model parameters;

[0010] The security and privacy module uses encryption algorithms to encrypt and store data, and uses identity authentication technologies, including passwords, fingerprint recognition, and face recognition.

[0011] Furthermore, in the behavior recognition and analysis module, the COA-CNN-LSTM model is used to analyze the preprocessed data, identify the behaviors of caregivers and patients, and classify them, which specifically includes the following steps:

[0012] Step S1: feature extraction, extracting features from the preprocessed data, extracting spatial features, motion features, appearance features and temporal features;

[0013] Step S2: Data serialization, serializing the extracted spatial features, motion features, appearance features and temporal features according to the frame sequence to obtain a feature sequence;

[0014] Step S3: Model construction, establish the COA-CNN-LSTM model, the architecture parameters include input layer, convolution layer 1, convolution layer 2, activation layer, maximum pooling layer, LSTM unit and fully connected layer, and use the COA algorithm to obtain the final parameter settings of the COA-CNN-LSTM model;

[0015] Step S4: model training, using the final parameter setting of the COA-CNN-LSTM model and training the COA-CNN-LSTM model using the feature sequence to obtain a trained COA-CNN-LSTM model;

[0016] Step S5: Behavior classification: Use the trained COA-CNN-LSTM model to classify behaviors and output corresponding behavior labels, including walking, standing, sitting, and lying down;

[0017] Further, step S3 uses the COA algorithm to obtain the final parameter settings of the COA-CNN-LSTM model, which specifically includes the following steps:

[0018] Step S31: parameter identification, setting the parameters that need to be improved in the COA-CNN-LSTM model, including architecture parameters, hyperparameters, and regularization parameters;

[0019] Step S32: setting a fitness function, using the cross entropy loss function as the fitness function, and calculating a fitness value;

[0020] Step S33: Initialize the parameter range, initialize the architecture parameters, hyperparameters and regularization parameters, the architecture parameters include the number of hidden layers and the number of neurons in each hidden layer, the hyperparameters include the learning rate, batch size and optimizer type, and the regularization parameters include the dropout rate and weight decay;

[0021] Step S34: COA algorithm, using the COA algorithm to iteratively optimize the COA-CNN-LSTM model parameters until the fitness value changes in the range of [0, 0.0001], stop iterating, and output the optimal solution;

[0022] Step S35: obtaining the best configuration, obtaining the best solution output by the COA algorithm and setting it as the best parameter configuration, and using the best parameter configuration as the final parameter setting of the COA-CNN-LSTM model;

[0023] Further, step S34 specifically includes the following steps:

[0024] Step S341: Initialize the population, randomly generate a set of initial solutions as individuals of the population, each individual represents a set of parameter settings of the COA-CNN-LSTM model;

[0025] Step S342: Evaluate the fitness, apply the COA-CNN-LSTM model to each individual, and calculate the corresponding fitness value as the performance indicator of the COA-CNN-LSTM model;

[0026] Step S343: Select excellent individuals, and select individuals with high fitness values ​​as excellent individuals;

[0027] Step S344: random perturbation, using crossover mutation to randomly perturb the individuals except the excellent individuals, simulating the behavior of cuckoos looking for new nests, and generating new solutions;

[0028] Step S345: Update the fitness, apply the COA-CNN-LSTM model to the new solution, and calculate the corresponding fitness value;

[0029] Step S346: Select update, compare the fitness value of the newly generated solution with the fitness value of the initial solution, and retain the solution with the higher fitness value;

[0030] Step S347: Termination condition, iterate step S341 to step S346 until the fitness value changes in the range of [0, 0.0001], stop iteration, and output the solution with a high fitness value as the optimal solution.

[0031] Furthermore, in the behavior recognition and analysis module, the VGG-ICNN model is used to analyze the behavior patterns of caregivers and patients to obtain behavior analysis results, which specifically includes the following steps:

[0032] Step K1: Build a behavior recognition model, using the VGG-ICNN model as the basic architecture to build a behavior recognition model;

[0033] Step K2: training the behavior recognition model, using the behavior pattern data to train the behavior recognition model, optimizing the parameters of the behavior recognition model through the back propagation algorithm, and obtaining the optimized behavior recognition model;

[0034] Step K3: Behavior analysis, using the optimized behavior recognition model to analyze, including actions, postures and positions, to obtain analysis content;

[0035] Step K4: Anomaly detection and analysis: perform anomaly detection on the analysis content, identify abnormal behavior and analyze it by comparing it with normal behavior pattern data to obtain analysis results;

[0036] Furthermore, step K1, the architecture of the VGG-ICNN model, specifically includes the following:

[0037] The VGG-ICNN model adds attention mechanism, residual connection and dense connection to the VGG network structure;

[0038] The attention mechanism dynamically adjusts the degree of attention paid to different positions when processing behavior pattern data;

[0039] The residual connection allows the behavior pattern data to be transferred across layers in the VGG network architecture, adding the output of a layer to the input of a previous layer, so that the VGG network learns the residual information;

[0040] The dense connections connect the output of each layer to the outputs of all previous layers.

[0041] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0042] (1) Ordinary models for behavioral analysis have the problems of low recognition accuracy, poor generalization ability and low efficiency. In the behavior recognition and analysis module, the present invention uses the COA-CNN-LSTM model to analyze the preprocessed data, identify the behaviors of caregivers and patients and classify them. The COA-CNN-LSTM model combines the characteristics of convolutional neural networks and long short-term memory networks, and can effectively capture the spatial and temporal features in image and time series data, thereby improving the recognition accuracy of caregivers and patients' behaviors. At the same time, the COA algorithm is introduced to optimize the COA-CNN-LSTM model parameters, so that the model can adapt to the data features in different environments and scenarios, improve the generalization ability, and perform personalized intervention and response according to the behavioral characteristics of different caregivers and patients, thereby improving the quality and efficiency of nursing care.

[0043] (2) In view of the problems of insufficient feature extraction capability, incomplete information representation and low computational efficiency existing in the general VGG network architecture, the present invention uses the VGG-ICNN model as the basic architecture in the behavior recognition and analysis module to construct a behavior recognition model. The deep convolutional network can effectively extract image features, and the introduction of the attention mechanism further strengthens the focus on important features, thereby improving the feature representation capability of the behavior pattern. The residual connection helps the model to better transmit gradient information, solving the gradient vanishing and gradient exploding problems in deep network training. The dense connection enables the model to perform feature fusion at different levels, effectively combining the features extracted at different levels, better understanding and predicting the behavior of caregivers and patients, and improving the intelligence and service quality of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of integrated equipment in the artificial intelligence-based caregiver behavior recognition and intervention system provided by the present invention;

[0045] Among them, 101 is a surveillance camera, 102 is a sound sensor, 103 is a temperature sensor, 104 is a motion sensor, 105 is a motion sensor 2, and 106 is a computing device.

[0046] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] In the description of the present invention, it is necessary to understand that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.

[0049] Example 1, see Figure 1 The artificial intelligence-based caregiver behavior recognition and intervention system provided by the present invention includes an integrated device, a data acquisition and preprocessing module, a behavior recognition and analysis module, an intervention and response module, a user interface module, and a security and privacy module, specifically including the following contents:

[0050] The integrated device includes a housing, a monitoring camera, a sound sensor, a temperature sensor, two motion sensors and a computing device. The housing is made of waterproof and dustproof polyethylene material and has a size of 5cm*5cm*5cm. The monitoring camera is located at the central outer bottom of the housing to detect visual information of the caregiver and the patient. The sound sensor is installed on the left side of the housing to detect sound information in the environment. The temperature sensor is installed on the right side of the housing to detect ambient temperature data. Two motion sensors are installed on the front side and the rear side of the housing to detect movement and posture change information of the caregiver. The monitoring camera, the sound sensor, the temperature sensor and the two motion sensors are connected to the computing device via USB interfaces respectively.

[0051] The data acquisition and preprocessing module receives visual information of the caregiver and the patient, sound information in the environment, ambient temperature data, and information on the movement and posture change of the caregiver detected by the integrated device, wherein the visual information of the caregiver and the patient includes the position, movement, expression, eye contact, and interactive movement of the caregiver and the patient, the sound information in the environment includes the conversation between the caregiver and the patient, the patient's breathing, the sound of moving objects, and the ambient noise, and the posture change information includes the posture type, posture duration, and posture switching frequency of the caregiver and the patient, and preprocesses the visual information of the caregiver and the patient, the sound information in the environment, the ambient temperature data, and the movement and posture change information of the caregiver to obtain preprocessed data;

[0052] The behavior recognition and analysis module uses the COA-CNN-LSTM model to analyze the preprocessed data, identifies and classifies the behaviors of caregivers and patients, including walking, standing, sitting and lying down, obtains the behavior pattern data of caregivers and patients, and uses the VGG-ICNN model to analyze the behavior pattern data of caregivers and patients to obtain behavior analysis results;

[0053] The intervention and response module receives the behavior pattern data of the caregiver and the patient obtained by the behavior recognition and analysis module and takes intervention measures;

[0054] The user interface module provides an interface used by users to monitor the operating status, view the behavior analysis results, and set the COA-CNN-LSTM model parameters;

[0055] The security and privacy module uses encryption algorithms to encrypt and store data, and uses identity authentication technologies, including passwords, fingerprint recognition, and face recognition.

[0056] Embodiment 2: This embodiment is based on the above embodiment. In the behavior recognition and analysis module, the COA-CNN-LSTM model is used to analyze the preprocessed data, identify the behaviors of caregivers and patients, and classify them. Specifically, the following steps are included:

[0057] Step S1: feature extraction, extracting features from the preprocessed data, extracting spatial features, motion features, appearance features and temporal features;

[0058] Step S2: Data serialization, serializing the extracted spatial features, motion features, appearance features and temporal features according to the frame sequence to obtain a feature sequence;

[0059] Step S3: Model construction, establish the COA-CNN-LSTM model, the architecture parameters include input layer, convolution layer 1, convolution layer 2, activation layer, maximum pooling layer, LSTM unit and fully connected layer, and use the COA algorithm to obtain the final parameter settings of the COA-CNN-LSTM model;

[0060] Step S4: model training, using the final parameter setting of the COA-CNN-LSTM model and training the COA-CNN-LSTM model using the feature sequence to obtain a trained COA-CNN-LSTM model;

[0061] Step S5: Behavior classification: Use the trained COA-CNN-LSTM model to perform behavior classification and output corresponding behavior labels, including walking, standing, sitting, and lying down.

[0062] By performing the above operations, the present invention provides a care worker behavior recognition and intervention system based on artificial intelligence. In general care worker behavior recognition and intervention systems, the use of ordinary models for behavior analysis has the problems of low recognition accuracy, poor generalization ability and low efficiency. In the behavior recognition and analysis module, the present invention uses the COA-CNN-LSTM model to analyze the preprocessed data, identify the behaviors of care workers and patients and classify them. The COA-CNN-LSTM model combines the characteristics of convolutional neural networks and long short-term memory networks, and can effectively capture spatial and temporal features in image and time series data, thereby improving the recognition accuracy of care workers and patient behaviors. At the same time, the COA algorithm is introduced to optimize the COA-CNN-LSTM model parameters, so that the model adapts to data features in different environments and scenarios, improves generalization ability, and performs personalized intervention and response according to the behavioral characteristics of different care workers and patients, thereby improving nursing quality and efficiency.

[0063] Embodiment 3, based on the above embodiment, step S3, using the COA algorithm to obtain the final parameter settings of the COA-CNN-LSTM model, specifically includes the following steps:

[0064] Step S31: parameter identification, setting the parameters that need to be improved in the COA-CNN-LSTM model, including architecture parameters, hyperparameters, and regularization parameters;

[0065] Step S32: setting a fitness function, using the cross entropy loss function as the fitness function, and calculating a fitness value;

[0066] Step S33: Initialize the parameter range, initialize the architecture parameters, hyperparameters and regularization parameters, the architecture parameters include the number of hidden layers and the number of neurons in each hidden layer, the hyperparameters include the learning rate, batch size and optimizer type, and the regularization parameters include the dropout rate and weight decay;

[0067] Step S34: COA algorithm, using the COA algorithm to iteratively optimize the COA-CNN-LSTM model parameters until the fitness value changes in the range of [0, 0.0001], stop iterating, and output the optimal solution;

[0068] Step S34 specifically includes the following steps:

[0069] Step S341: Initialize the population, randomly generate a set of initial solutions as individuals of the population, each individual represents a set of parameter settings of the COA-CNN-LSTM model;

[0070] Step S342: Evaluate the fitness, apply the COA-CNN-LSTM model to each individual, and calculate the corresponding fitness value as the performance indicator of the COA-CNN-LSTM model;

[0071] Step S343: Select excellent individuals, and select individuals with high fitness values ​​as excellent individuals;

[0072] Step S344: random perturbation, using crossover mutation to randomly perturb the individuals except the excellent individuals, simulating the behavior of cuckoos looking for new nests, and generating new solutions;

[0073] Step S345: Update the fitness, apply the COA-CNN-LSTM model to the new solution, and calculate the corresponding fitness value;

[0074] Step S346: Select update, compare the fitness value of the newly generated solution with the fitness value of the initial solution, and retain the solution with the higher fitness value;

[0075] Step S347: Termination condition, iterate step S341 to step S346 until the fitness value changes in the range of [0, 0.0001], stop iterating, and output the solution with the highest fitness value as the optimal solution;

[0076] Step S35: Obtain the optimal configuration, obtain the optimal solution output by the COA algorithm and set it as the optimal parameter configuration, and use the optimal parameter configuration as the final parameter setting of the COA-CNN-LSTM model.

[0077] Embodiment 4: Based on the above embodiment, in the behavior recognition and analysis module, the VGG-ICNN model is used to analyze the behavior patterns of the caregivers and patients to obtain the behavior analysis results, which specifically includes the following steps:

[0078] Step K1: Build a behavior recognition model, using the VGG-ICNN model as the basic architecture to build a behavior recognition model;

[0079] Step K1, the architecture of the VGG-ICNN model, specifically includes the following:

[0080] The VGG-ICNN model adds attention mechanism, residual connection and dense connection to the VGG network structure;

[0081] The attention mechanism dynamically adjusts the degree of attention paid to different positions when processing behavior pattern data;

[0082] The residual connection allows the behavior pattern data to be transferred across layers in the VGG network architecture, adding the output of a layer to the input of a previous layer, so that the VGG network learns the residual information;

[0083] The dense connection connects the output of each layer to the output of all previous layers;

[0084] Step K2: training the behavior recognition model, using the behavior pattern data to train the behavior recognition model, optimizing the parameters of the behavior recognition model through the back propagation algorithm, and obtaining the optimized behavior recognition model;

[0085] Step K3: Behavior analysis, using the optimized behavior recognition model to analyze, including actions, postures and positions, to obtain analysis content;

[0086] Step K4: Anomaly detection and analysis: perform anomaly detection on the analysis content, identify abnormal behavior and analyze it by comparing it with normal behavior pattern data to obtain analysis results.

[0087] By performing the above operations, in view of the problems of insufficient feature extraction capability, incomplete information representation and low computational efficiency existing in the general VGG network architecture, the present invention uses the VGG-ICNN model as the basic architecture in the behavior recognition and analysis module to construct a behavior recognition model. The deep convolutional network can effectively extract image features, and the introduction of the attention mechanism further strengthens the focus on important features, thereby improving the feature representation capability of the behavior pattern. The residual connection helps the model to better transmit gradient information, solving the gradient vanishing and gradient explosion problems in deep network training. The dense connection enables the model to perform feature fusion at different levels, effectively combining the features extracted at different levels, better understanding and predicting the behavior of caregivers and patients, and improving the intelligence and service quality of the system.

[0088] Embodiment 5: This embodiment is based on the above embodiment. In the integrated device, the shell is made of waterproof and dustproof polyethylene material and has a size of 5cm*5cm*5cm. The surveillance camera uses Logitech C920 HD Pro Webcam, the sound sensor uses Adafruit Electret Microphone Amplifier - MAX9814, the temperature sensor uses Dallas Semiconductor DS18B20 Programmable Resolution 1-Wire Digital Thermometer, the two motion sensors are both Bosch Sensortec BNO055 Intelligent 9-axis absolute orientation sensor, and the computing device uses Raspberry Pi 4 Model B with 4GB of memory.

[0089] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0090] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0091] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. The artificial intelligence-based nursing behavior recognition and intervention system is characterized by: It includes integrated equipment, data acquisition and preprocessing module, behavior recognition and analysis module, intervention and response module, user interface module and security and privacy module. The integrated device detects visual information of the caregiver and the patient, sound information in the environment, ambient temperature data, and movement and posture change information of the caregiver; The data acquisition and preprocessing module receives the visual information of the caregiver and the patient, the sound information in the environment, the ambient temperature data, and the movement and posture change information of the caregiver detected by the integrated device, and preprocesses the visual information of the caregiver and the patient, the sound information in the environment, the ambient temperature data, and the movement and posture change information of the caregiver to obtain the preprocessed data; The behavior recognition and analysis module uses the COA-CNN-LSTM model to analyze the preprocessed data, identifies and classifies the behaviors of the caregivers and patients, obtains the behavior pattern data of the caregivers and patients, and uses the VGG-ICNN model to analyze the behavior pattern data of the caregivers and patients to obtain the behavior analysis results; The intervention and response module receives the behavior pattern data of the caregiver and the patient obtained by the behavior recognition and analysis module and takes intervention measures; The user interface module provides an interface used by users to monitor the operating status, view the behavior analysis results, and set the COA-CNN-LSTM model parameters; The security and privacy module uses encryption algorithms to encrypt and store data, and uses identity authentication technologies, including passwords, fingerprint recognition, and face recognition; In the behavior recognition and analysis module, the COA-CNN-LSTM model is used to analyze the preprocessed data, identify the behaviors of caregivers and patients, and classify them. The specific steps include: Step S1: feature extraction, extracting features from the preprocessed data, extracting spatial features, motion features, appearance features and temporal features; Step S2: Data serialization, serializing the extracted spatial features, motion features, appearance features and temporal features according to the frame sequence to obtain a feature sequence; Step S3: Model construction, establish the COA-CNN-LSTM model, the architecture parameters include input layer, convolution layer 1, convolution layer 2, activation layer, maximum pooling layer, LSTM unit and fully connected layer, and use the COA algorithm to obtain the final parameter settings of the COA-CNN-LSTM model; Step S4: model training, using the final parameter setting of the COA-CNN-LSTM model and training the COA-CNN-LSTM model using the feature sequence to obtain a trained COA-CNN-LSTM model; Step S5: Behavior classification: Use the trained COA-CNN-LSTM model to classify behaviors and output corresponding behavior labels; Step S3 specifically includes the following steps: Step S31: parameter identification, setting the parameters that need to be improved in the COA-CNN-LSTM model, including architecture parameters, hyperparameters, and regularization parameters; Step S32: setting a fitness function, using the cross entropy loss function as the fitness function, and calculating a fitness value; Step S33: Initialize the parameter range, initialize the architecture parameters, hyperparameters and regularization parameters, the architecture parameters include the number of hidden layers and the number of neurons in each hidden layer, the hyperparameters include the learning rate, batch size and optimizer type, and the regularization parameters include the dropout rate and weight decay; Step S34: COA algorithm, using the COA algorithm to iteratively optimize the COA-CNN-LSTM model parameters until the fitness value changes in the range of [0, 0.0001], stop iterating, and output the optimal solution; Step S35: obtaining the best configuration, obtaining the best solution output by the COA algorithm and setting it as the best parameter configuration, and using the best parameter configuration as the final parameter setting of the COA-CNN-LSTM model; In the behavior recognition and analysis module, the VGG-ICNN model is used to analyze the behavior patterns of caregivers and patients to obtain the behavior analysis results, which specifically includes the following steps: Step K1: Build a behavior recognition model, using the VGG-ICNN model as the basic architecture to build a behavior recognition model; Step K2: training the behavior recognition model, using the behavior pattern data to train the behavior recognition model, optimizing the parameters of the behavior recognition model through the back propagation algorithm, and obtaining the optimized behavior recognition model; Step K3: Behavior analysis, using the optimized behavior recognition model to perform analysis and obtain analysis content; Step K4: Anomaly detection and analysis: perform anomaly detection on the analysis content, identify abnormal behavior and analyze it by comparing it with normal behavior pattern data to obtain analysis results; The architecture of the VGG-ICNN model includes the following: The VGG-ICNN model adds attention mechanism, residual connection and dense connection to the VGG network structure; The attention mechanism dynamically adjusts the degree of attention paid to different positions when processing behavior pattern data; The residual connection allows the behavior pattern data to be transferred across layers in the VGG network architecture, adding the output of a layer to the input of a previous layer, so that the VGG network learns the residual information; The dense connections connect the output of each layer to the outputs of all previous layers.

Images