A Method for Constructing an Integrated Medical and Elderly Care Knowledge Service System for Elderly Care

Abstract

This invention relates to the fields of medicine and artificial intelligence, and discloses a method for constructing a medical and elderly care integrated knowledge service system, including the following steps: collecting emotional expression data of the elderly, applying nonlinear time series analysis and deep learning models to extract emotional dynamic features and identify early signals of declining emotional regulation ability; based on the acquired emotional dynamic features, using a memory-enhanced attention network combined with a psychological knowledge graph to construct an individualized emotional-cognitive interaction model for the elderly, revealing the dynamic interaction mechanism between emotional state and cognitive pattern, and identifying key intervention nodes; based on the emotional-cognitive interaction model and key node identification algorithm, this invention enables the system to accurately locate the key links in the interaction between emotion and cognition, and achieve targeted intervention; and the dual-channel collaborative intervention strategy, through multi-agent reinforcement learning, regulates emotional state and reconstructs cognitive pattern, breaking the vicious cycle of emotional cognition.

Description

Technical Field

[0001] This invention relates to the fields of medicine and artificial intelligence, and more specifically, to a method for constructing a knowledge service system that integrates medical and elderly care for the elderly. Background Technology

[0002] With the accelerating aging of the population, mental health issues among the elderly, especially those living alone, are becoming increasingly prominent. Mental illnesses such as depression and anxiety have become significant factors affecting the quality of life for the elderly. Currently, the management of mental health in the elderly faces multiple challenges: on the one hand, mental health problems in the elderly are often subtle, making early identification difficult using traditional screening methods; on the other hand, resources for mental health intervention are limited, resulting in low intervention efficiency.

[0003] The existing technologies have the following shortcomings: First, they mainly rely on cross-sectional assessments of the elderly’s immediate emotional state, ignoring the dynamic characteristics of emotional changes and making it difficult to distinguish between normal emotional fluctuations and pathological changes; Second, they are insufficient in recognizing the emotional expression patterns unique to the elderly and are unable to capture the implicit psychological needs; Third, they often focus solely on emotional symptoms or cognitive distortions, lacking a collaborative intervention mechanism and failing to effectively break the vicious cycle between emotion and cognition; Fourth, they lack personalized adaptability and are unable to meet the individualized needs of different elderly people.

[0004] Therefore, there is an urgent need for a knowledge service system that can achieve early and accurate identification and efficient intervention, and improve the efficiency and accuracy of elderly mental health management through multimodal data analysis, intelligent algorithms and comprehensive intervention strategies. Summary of the Invention

[0005] This invention provides a method for constructing a knowledge service system integrating medical and elderly care, thereby solving the technical problems in the construction of such a system in related technologies.

[0006] This invention provides a method for constructing a knowledge service system integrating medical and elderly care for the elderly, comprising the following steps:

[0007] Collect emotional expression data from the elderly, apply nonlinear time series analysis and deep learning models to extract emotional dynamic features, and identify early signs of declining emotional regulation ability;

[0008] Based on the acquired emotional dynamic features, a memory-enhanced attention network combined with a psychological knowledge graph is used to construct an individualized emotional cognitive interaction model for the elderly, revealing the dynamic interaction mechanism between emotional state and cognitive pattern, and identifying key intervention nodes.

[0009] Based on the constructed interaction model, a multi-agent reinforcement learning algorithm is applied to generate a dual-channel collaborative intervention strategy for emotion and cognition, thereby maximizing the effect of precise intervention.

[0010] Implement the generated intervention strategies, establish a closed-loop intervention implementation and feedback mechanism, realize dynamic evaluation of intervention effects and strategy optimization, and improve the accuracy and sustainability of mental health management for the elderly;

[0011] By integrating functional modules into the integrated medical and elderly care knowledge service system, collaborative work with other health management modules can be achieved, thus building a complete ecosystem for elderly mental health services.

[0012] As a further optimization of the present invention, the functional modules include an emotional dynamic analysis module, an emotional cognition interaction modeling module, a dual-channel collaborative intervention module, and an intervention execution module.

[0013] As a further optimization of the present invention, the extraction of emotional dynamic features specifically includes:

[0014] Multimodal data of older adults, including voice, facial expressions, activity patterns, and physiological signals, are collected through sensor networks and used as the basic input for emotional state analysis.

[0015] We apply a deep multimodal fusion algorithm to extract feature parameters of emotional expression, including emotional intensity, emotional diversity, emotional inertia, and emotional resilience, and construct a dynamic feature vector of emotion.

[0016] Nonlinear time series analysis is used to perform complexity analysis on the time series of emotional dynamic feature vectors, and to quantify the changes in the complexity of the emotional dynamic system.

[0017] We construct a hybrid architecture of Hidden Markov Model and Long Short-Term Memory Network to capture the long-term evolutionary patterns of emotional state transitions and identify anomalous transition patterns.

[0018] As a further optimization of the present invention, the construction of the emotion cognition interaction model specifically includes:

[0019] Based on emotional dynamics and cognitive assessment data, construct emotional state vectors and cognitive pattern vectors;

[0020] A memory-enhanced attention network is used to construct a bidirectional influence model of emotion and cognition, capturing the mutual influence between the two, and storing historical state information through an attention weight matrix;

[0021] Based on individual data of the elderly, an optimization algorithm combining gradient descent and variational inference is used to train an individualized emotional cognitive interaction model.

[0022] By applying causal inference methods and based on a trained interaction model, we can identify key nodes in an individual's specific emotional and cognitive interactions, which will serve as the key targets for the formulation of subsequent intervention strategies.

[0023] As a further optimization of the present invention, the generation of the dual-channel collaborative intervention strategy specifically includes:

[0024] Based on the emotion-cognition interaction model, the intervention optimization objective function and constraints are defined;

[0025] Construct a multi-agent system composed of emotional intervention agents and cognitive intervention agents, and maximize joint rewards by sharing state information and coordinating action choices through a collaborative mechanism;

[0026] We employ a deep deterministic strategy gradient algorithm to optimize intervention strategies and design a collaborative decision-making mechanism to coordinate the timing of emotional and cognitive interventions, thereby maximizing synergistic effects.

[0027] As a further optimization of the present invention, the intervention execution specifically includes:

[0028] Construct a multimodal intervention execution system that integrates multiple interaction channels such as virtual assistants, smart home environments, and wearable devices to achieve seamless execution of emotional and cognitive dual-channel intervention strategies;

[0029] Establish a multi-dimensional intervention effect evaluation mechanism, including immediate emotional response evaluation, cognitive change evaluation, and long-term health status evaluation;

[0030] By applying Bayesian adjustment methods and incremental learning models, intervention strategy parameters are optimized in real time based on intervention effectiveness and acceptance.

[0031] As a further optimization of the present invention, the knowledge service system integration and application specifically includes:

[0032] Construct a modular system architecture, encapsulate technical components into standardized service modules, and form a scalable integrated medical and elderly care knowledge service system;

[0033] To establish a collaborative working mechanism between mental health management and other health management modules, and to construct a comprehensive picture of the health status of the elderly;

[0034] Based on the elderly’s personal interests, life experiences and social preferences, a personalized interaction strategy generation algorithm is constructed.

[0035] We will build a dynamic update and optimization mechanism for the knowledge base, continuously accumulate experience from intervention practices, and improve the overall service quality.

[0036] As a further optimization of the present invention, the nonlinear time series analysis method includes calculating the entropy increase rate, Lyapunov exponent, and fractal dimension of the emotional trajectory to quantify the complexity changes of the emotional dynamic system.

[0037] As a further optimization of the present invention, the memory-enhanced attention network includes three parts: an encoder, a memory module, and a decoder. The memory module maintains a historical state buffer, and the multi-head attention mechanism is used to calculate the correlation between the current state and historical memory.

[0038] A system for constructing an integrated medical and elderly care knowledge service system for elderly care, used to execute the aforementioned method for constructing an integrated medical and elderly care knowledge service system for elderly care, includes:

[0039] The emotional dynamics analysis module is used to collect and analyze multimodal emotional data of the elderly, extract emotional dynamics features, identify early signs of declining emotional regulation ability, and provide basic data support for subsequent interventions.

[0040] The Emotional Cognition Interaction Modeling Module is used to construct individualized emotional cognition interaction models, reveal the dynamic relationship between emotional states and cognitive patterns, identify key intervention nodes, and provide a theoretical basis for formulating intervention strategies.

[0041] The dual-channel collaborative intervention module is used to generate emotion and cognition dual-channel collaborative intervention strategies based on multi-agent reinforcement learning, so as to achieve the optimal combination of emotion intervention and cognitive intervention and maximize the intervention effect.

[0042] The personalized intervention execution module is used to execute intervention strategies through multimodal interaction channels, evaluate intervention effects in real time, and dynamically adjust intervention parameters to ensure the accuracy and continuity of the intervention.

[0043] The knowledge service integration module is used to integrate various functional modules, manage knowledge base updates, coordinate interfaces with other health management systems, and provide a unified service interface.

[0044] Each module communicates through a standardized API, and loosely coupled integration is achieved using an event-driven model.

[0045] The beneficial effects of this invention are as follows: Based on the emotional cognition interaction model and key node identification algorithm, this invention enables the Zhengege system to accurately locate the key links in the interaction between emotion and cognition, and achieve targeted intervention; and the dual-channel collaborative intervention strategy, through multi-agent reinforcement learning, simultaneously regulates emotional state and reconstructs cognitive patterns, effectively breaking the vicious cycle of emotion and cognition, and significantly improving the effect output ratio of unit intervention resources. Attached Figure Description

[0046] Figure 1 This is a flowchart of a method for constructing a medical and elderly care integrated knowledge service system according to the present invention;

[0047] Figure 2 This is a detailed flowchart of the emotional dynamic feature extraction and analysis method of the present invention;

[0048] Figure 3 This is a detailed flowchart of the construction of the emotional cognition interaction model of the present invention;

[0049] Figure 4 This is a detailed flowchart of the dual-channel collaborative intervention strategy generated by the present invention;

[0050] Figure 5 This is a detailed flowchart of the personalized intervention implementation and feedback adjustment of this invention;

[0051] Figure 6 This is a detailed flowchart of the knowledge service system integration and application of the present invention. Detailed Implementation

[0052] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0053] At least one embodiment of the present invention discloses a method for constructing a knowledge service system integrating medical and elderly care for the elderly, such as... Figures 1 to 6 As shown, the following steps are included:

[0054] Step 1: Collect emotional expression data of the elderly, apply nonlinear time series analysis and deep learning models to extract emotional dynamic features and identify early signs of declining emotional regulation ability;

[0055] Specifically, the following steps are included:

[0056] Step 1.1: Multimodal sentiment data collection;

[0057] Multimodal data on elderly individuals, including voice, facial expressions, activity patterns, and physiological signals, are collected through a sensor network. This data serves as the foundational input for emotional state analysis. Specifically, the sensor network includes an audio acquisition module, a video acquisition module, a wearable physiological signal monitoring module, and an environmental interaction behavior recording module. Furthermore, the system performs preliminary cleaning and standardization of the collected raw data to create a uniformly formatted raw dataset of emotional expressions. .

[0058] Step 1.2: Construction of emotional dynamic feature vectors;

[0059] Based on the collected multimodal data, this application applies a deep multimodal fusion algorithm to extract feature parameters of emotional expression, including emotional intensity. Emotional diversity Emotional inertia and emotional resilience Constructing a dynamic feature vector of emotion:

[0060]

[0061] in, Represents the dynamic feature vector of emotions; Indicates the intensity of emotional expression; An index representing the diversity of emotion types; Indicates the duration of an emotional state; This indicates the rate at which one recovers from negative emotions to the baseline state. This represents the transpose of a vector.

[0062] The system employs a cross-modal self-attention mechanism to integrate emotional features from different modalities, ensuring the robustness of feature extraction and avoiding the impact of missing information from a single modality on feature extraction.

[0063] Step 1.3: Dynamic analysis of emotional complexity;

[0064] After obtaining the emotional dynamic feature vector, this application employs a nonlinear time series analysis method to analyze the emotional dynamic feature vector. A complexity analysis was performed on the time series data. Specifically, the entropy increase rate of the sentiment trajectory was calculated. Lyapunov index and fractal dimension Quantifying the changes in complexity of the emotional dynamic system:

[0065]

[0066] in, Indicates time The change in emotional complexity; Indicates time The complexity of emotions; Indicates time The complexity of emotions; Indicates the size of the time window. Indicates a time step.

[0067] The formula for calculating the complexity index is as follows:

[0068] Entropy increase rate: ;

[0069] in, Indicates the rate of entropy increase; Indicates when The limit as it approaches infinity; Represents the normalization coefficient; Represents the Shannon entropy function; , , They represent the first , , Emotional state at a specific point in time Indicates the length of the time series.

[0070] Lyapunov Index: ;

[0071] in, Lyapunov index; Represents a continuous-time variable; This indicates the deviation from the initial state; Indicates time Deviation at time; Indicates when The limit as it approaches infinity; This represents the limit as the initial deviation approaches 0; Represents the time normalization coefficient; It represents the natural logarithm.

[0072] fractal dimension The fractal properties of emotional trajectories were calculated using box counting. Therefore, the changing trends of these complexity indicators can effectively distinguish between normal emotional fluctuations and pathological changes, providing a quantitative basis for early warning.

[0073] Step 1.4: Identification of emotional state transition patterns;

[0074] Finally, this application constructs a hybrid architecture of Hidden Markov Models (HMMs) and Long Short-Term Memory (LSTM) networks to capture long-term evolutionary patterns in emotional state transitions. In this hybrid architecture, the HMM is used to characterize the transition probability matrix between discrete emotional states. ,in Indicates from emotional state Transition to state The probability of this shift is determined by the LSTM network. Correspondingly, the LSTM network learns sequence patterns, predicts the future trajectory of emotional state sequences, and identifies anomalous transition patterns. Therefore, the model outputs a warning signal indicating a decline in emotional regulation ability. When the warning index exceeds the threshold At that time, the system marks potential mental health risks.

[0075] Specifically, the hybrid architecture is implemented as follows: First, the emotional state space is discretized into a finite set of states:

[0076]

[0077] in, Represents the emotional state space; , , They represent the first , , An emotional state; This represents the total number of emotional states.

[0078] Each state represents a combination of emotional states (e.g., "mild anxiety," "moderate depression," etc.). Then, the Hidden Markov Model is based on the observed sequence of emotional features:

[0079]

[0080] in, This represents the observed sequence of emotional features; , , They represent the first , , One observed emotional characteristic; This represents the total number of observed emotional characteristics.

[0081] Estimate the state transition probability matrix and observed probability distribution Building upon this, the LSTM network further captures long-term dependencies, with its input being a sequence of sentiment feature vectors:

[0082]

[0083] in, , , They represent the first , , One sentiment feature vector; This represents the total number of sentiment feature vectors.

[0084] The output is the probability distribution of the predicted emotional state at the next moment.

[0085] The workflow of this hybrid architecture in specific application scenarios is as follows: When an elderly person is continuously monitored by the system in their daily life, the system extracts emotional features from multimodal data and generates a sequence of emotional feature vectors. This sequence is then input into a hybrid architecture of a Hidden Markov Model and an LSTM to calculate the probability distribution of emotional state transitions and the predicted value of future emotional states. If the system detects abnormal transition patterns, such as a sudden increase in the probability of transitioning from "mild anxiety" to "severe depression," or if the LSTM predicts a continuously deteriorating trend in future emotional states, an early warning signal is generated. For example, if a senior living alone shows a decrease in emotional intensity and diversity in their voice analysis over several consecutive days, and their activity patterns show reduced social interaction, the hybrid architecture can predict that the senior is at risk of developing depression, and the system will immediately trigger an early warning and intervene in advance.

[0086] Step 2: Based on the acquired emotional dynamic features, a memory-enhanced attention network combined with a psychological knowledge graph is used to construct an individualized emotional cognitive interaction model for the elderly, revealing the dynamic interaction mechanism between emotional state and cognitive pattern, and identifying key intervention nodes.

[0087] Specifically, the following steps are included:

[0088] Step 2.1: Representation of Emotional State and Cognitive Pattern;

[0089] Based on the emotional dynamic features and cognitive assessment data obtained in step 1, an emotional state vector is constructed. and cognitive pattern vector .

[0090] Emotional state vector: ;

[0091] in, Represents an emotional state vector; , , They represent the first , , The intensity of each emotional dimension; This represents the total number of emotional dimensions; This represents the transpose of a vector.

[0092] Cognitive pattern vector: ;

[0093] in, Represents a cognitive pattern vector; , , They represent the first , , The degree of each cognitive dimension; This represents the total number of cognitive dimensions; This represents the transpose of a vector.

[0094] Specifically, the emotional state vector contains multidimensional emotional features, while the cognitive pattern vector includes elements such as cognitive distortions, belief systems, and thinking patterns. It should be noted that the system extracts cognitive pattern features from the language expressions of the elderly through natural language processing technology, and combines this with psychological scale assessment data to form a comprehensive cognitive pattern representation.

[0095] Step 2.2: Modeling the two-way influence relationship;

[0096] This application uses a memory-enhanced attention network to construct a model of the bidirectional influence between emotion and cognition. This model captures the mutual influence between the two:

[0097]

[0098]

[0099] in, express The emotional state vector at any given moment; express The emotional state vector at any given moment; express Cognitive pattern vector at any given moment; The model parameters representing the emotion state transition function; A transition function representing an emotional state; express Cognitive pattern vector at any given moment; A transfer function representing a cognitive pattern; The model parameters represent the cognitive pattern transfer function; Indicates a time step.

[0100] It should be noted that the memory enhancement mechanism utilizes an attention weight matrix. Storing historical state information, formally represented as:

[0101]

[0102] in, Represents the attention weight matrix; Represents the query matrix; Represents the transpose of the key-value matrix; Indicates the hidden layer dimension; This represents the softmax activation function.

[0103] This model is able to learn long-term dependencies, reflecting the long-term cumulative mutual influence between emotional states and cognitive patterns.

[0104] The specific implementation of the memory-enhanced attention network is as follows: The network architecture consists of three parts: an encoder, a memory module, and a decoder. The encoder maps the emotional state vector and the cognitive pattern vector to a latent space representation. The memory module maintains a historical state buffer.

[0105]

[0106] in, Represents the memory module; , , They represent the first , , One memory unit; This indicates the total number of memory units.

[0107] Each memory unit stores past state information. Multi-head attention mechanisms are used to calculate the correlation between the current state and historical memories.

[0108]

[0109] in, Indicates the first The output of each attention head; This represents the attention calculation function; , and They represent the first The query, key, and value matrix of each attention head is obtained through a linear transformation of the current state and historical memory; Represents the transpose of the key matrix; Indicates the dimension of the key vector; This represents the softmax activation function.

[0110] The outputs of multi-head attention are integrated through concatenation and linear transformation:

[0111]

[0112] in, This represents the output of the multi-head attention mechanism. Represents the query matrix, Represents the key matrix, Represents a value matrix; , , They represent the first , , The output of each attention head; Indicates the total number of heads of attention; This represents a linear transformation matrix.

[0113] The decoder predicts the emotional state and cognitive patterns of the next moment based on the encoder output and attention-weighted historical memory.

[0114] In specific application scenarios, the model's workflow is as follows: When emotional fluctuations are detected in an elderly person, the system collects their recent emotional state data and cognitive assessment data, inputting them into a memory-enhanced attention network. For example, if an elderly person exhibits persistent grief after the death of their spouse, the system detects a continuous increase in the grief dimension of their emotional state vector, while simultaneously, "excessive negative thinking" and "decreased self-worth" in their cognitive pattern vector also gradually increase. Through historical data analysis, the memory-enhanced attention network identifies this interactive reinforcement process of emotional cognitive patterns and predicts that without intervention, the elderly person may develop depression. The system identifies "excessive negative thinking" as a key intervention point because attention weight analysis reveals that this cognitive pattern has the highest weight in influencing grief, and based on this, a targeted intervention strategy is developed.

[0115] Step 2.3: Individualized model training and optimization;

[0116] Based on individual data from elderly individuals, this application employs an optimization algorithm combining gradient descent and variational inference to train a personalized emotional cognitive interaction model. Specifically, the objective function is set to minimize the weighted sum of prediction error and model complexity:

[0117]

[0118]

[0119] in, Represent the objective function; , and These represent the weight coefficients of sentiment prediction error, cognitive prediction error, and regularization term, respectively. Represents the mean square error function; Indicates the predicted emotional state; To express a true emotional state; Cognitive patterns representing predictions; Represents the true cognitive pattern; Represents the regularization term; Indicates model parameters.

[0120] In addition, the system dynamically adjusts the weight coefficients to balance the model's fit and generalization ability, outputting individualized emotional cognitive interaction model parameters. .

[0121] Step 2.4: Identification of key intervention nodes;

[0122] Finally, this application applies a causal inference method to identify key nodes of individual-specific emotional and cognitive interactions based on a trained interaction model. It should be understood that a critical point is defined as the point at which the influence of emotion on cognition or cognition on emotion exceeds a threshold:

[0123]

[0124] in, Represents the set of key nodes; Indicates the first One emotional dimension; Indicates the first One cognitive dimension; This indicates the strength of the influence of cognitive dimensions on emotional state; This indicates the strength of the influence of the affective dimension on cognitive patterns; Indicates the threshold of emotional influence; Indicates the threshold of cognitive influence; Index representing the sentiment dimension, An index representing the cognitive dimension.

[0125] Therefore, the system quantifies the influence strength between each dimension by calculating the partial derivative matrix, and combines this with sensitivity analysis to determine the set of most influential key nodes. As a key objective in formulating subsequent intervention strategies, the specific formula for calculating the partial derivative matrix is ​​as follows:

[0126]

[0127]

[0128] in, A matrix representing the influence of emotion on cognition; A matrix representing the influence of cognition on emotion; This indicates the strength of the influence of cognitive dimensions on emotional state; This indicates the strength of the influence of the emotional dimension on cognitive patterns.

[0129] Step 3: Based on the constructed interaction model, apply a multi-agent reinforcement learning algorithm to generate a dual-channel collaborative intervention strategy for emotion and cognition, so as to maximize the effect of precise intervention.

[0130] Specifically, the following steps are included:

[0131] Step 3.1: Define the intervention objectives and constraints;

[0132] Based on the emotional cognitive interaction model constructed in step 2, this implementation method defines the intervention optimization objective function:

[0133]

[0134] in, Represent the objective function; Indicates the intervention strategy; Indicating in strategy The following expectations; Indicates a time step; Indicates from time step arrive The sum of discount rewards; Discount factor The power is used to balance the importance of current and future rewards; Indicates the total number of time steps; express The state at any given moment; express Intervention actions taken at any time; This represents the reward function.

[0135] Specifically, state Includes emotional state vector Cognitive pattern vector and environment context vector Formal representation:

[0136]

[0137] in, express The state vector at any given time; , , Let represent the emotional state vector, cognitive pattern vector, and environmental context vector at time t, respectively; This represents the transpose operator.

[0138] The intervention action space is divided into an emotional intervention action subspace and a cognitive intervention action subspace; its expression is:

[0139]

[0140] in, Indicates the scope of intervention actions; Represents the subspace for emotional intervention actions; This represents the cognitive intervention action subspace.

[0141] reward function The design was based on factors such as the degree of improvement in mental health, resource consumption, and acceptance among the elderly.

[0142] In addition, constraints include resource limitations. and psychological burden limitation ;

[0143] in, Represents the resource consumption function; Indicates the resource limit threshold; Represents the psychological burden function; This indicates the threshold for limiting psychological burden; Indicates state; This indicates intervention.

[0144] Step 3.2: Design of a multi-agent reinforcement learning framework;

[0145] This implementation method constructs an emotional intervention intelligent agent. and cognitive intervention intelligent agents A multi-agent system. Each agent is based on a Deep Q-Network (DQN) architecture and optimizes its own Q-value function.

[0146] The two agents share state information and coordinate action choices through a collaborative mechanism to maximize joint rewards. It should be noted that this multi-agent reinforcement learning framework can simultaneously optimize strategies for both emotional and cognitive interventions, resulting in synergistic intervention effects.

[0147] The specific implementation of the multi-agent reinforcement learning framework is as follows: The system adopts an attention-based multi-agent reinforcement learning (MARL) framework, which includes an emotion intervention agent. and cognitive intervention intelligent agents Two sub-agents, along with a central network of critics, are used to assess the value of the joint action;

[0148] Each agent's policy network structure consists of the following components:

[0149] 1. State encoder: converts states... Mapping to latent representation ;

[0150] 2. Action Generator: Generates action probability distributions based on state representations. ;

[0151] 3. Value function estimator: estimates the value of a state. ;

[0152] Cooperation between agents is achieved through an attention mechanism, where each agent pays attention to the state and actions of other agents:

[0153]

[0154] in, Represents intelligent agents Context vector; Indicates the absence of intelligent agents All other intelligent agents Summation; Represents intelligent agents For intelligent agents Attention weights; Represents intelligent agents The state representation.

[0155] The calculation formula is:

[0156]

[0157] in, Represents intelligent agents For intelligent agents Attention weights; Represents an exponential function; It is a similarity metric function used to calculate the correlation or similarity between the state representations of two agents; Represents intelligent agents State representation; Represents intelligent agents State representation; Indicates the absence of intelligent agents All other intelligent agents Summation; Represents intelligent agents The state representation.

[0158] The training process uses the Actor-Critic framework, and the policy gradient update for each agent is as follows:

[0159]

[0160] in, Indicates the parameter gradient operator; Represents intelligent agents The objective function; Indicates about state and action Expectations; This represents the natural logarithm function, used to compute the policy function. The log-likelihood; Represents intelligent agents The strategy function; Represents intelligent agents Actions; Indicates state; Represents intelligent agents The strategy network parameters; The Q-value function represents the central commentator network; This indicates an emotional intervention action; This indicates cognitive intervention actions; This represents the parameters of the Central Commentators Network.

[0161] To handle non-stationary training environments, the system implements an experience replay buffer. And the target network, and adopt a soft update mechanism:

[0162]

[0163] in, Indicates the target network parameters; Indicates the soft update coefficient; Indicates the current network parameters; Indicates the parameter update operator; This represents the complement of the soft update coefficient.

[0164] In practical applications, the workflow of this multi-agent reinforcement learning framework is as follows: the system continuously monitors the emotional state and cognitive patterns of the elderly, and initiates intervention when abnormal patterns are detected. For example, for an elderly person at risk of mild depression, and Analyzing their emotional state vector and cognitive pattern vector revealed that "social isolation" and "catastrophic thinking" are key issues. In the collaborative decision-making process, Choosing "increasing social participation activities" as the emotional intervention action, at the same time "Cognitive restructuring training" was selected as the cognitive intervention action. The system recommends nearby community activities to the elderly through a virtual assistant (emotional intervention) and provides targeted cognitive behavioral therapy exercises (cognitive intervention). The two intervention channels work together to improve the elderly's mental state. As the intervention progresses, the system continuously collects feedback data, and the two agents continuously adjust and optimize their respective intervention strategies to form a personalized intervention plan.

[0165] Step 3.3: Optimization of intervention strategies and collaborative decision-making;

[0166] This implementation employs the Deep Deterministic Policy Gradient (DDPG) algorithm to optimize the intervention strategy, which includes a policy network and a value network. Through experience replay and target network techniques, the impact of sample correlation and non-stationarity during training is reduced. Simultaneously, this implementation also designs a collaborative decision-making mechanism to coordinate the timing of emotional and cognitive interventions, based on the interaction effect of the two intervention pathways.

[0167]

[0168] in, This indicates the interaction effect of the two-channel intervention; This represents the function representing the intervention effect; This indicates an emotional intervention action; This indicates cognitive intervention actions; This indicates an air intervention operation.

[0169] In addition, the system determines the optimal intervention sequence based on the connection strength and key node positions in the emotional cognitive interaction model.

[0170] Step 4: Implement the generated intervention strategy, establish a closed-loop intervention implementation and feedback mechanism, realize dynamic evaluation of intervention effects and strategy optimization, and improve the accuracy and sustainability of mental health management for the elderly.

[0171] In another embodiment of this application, this step establishes a closed-loop intervention execution and feedback mechanism to achieve dynamic evaluation of intervention effects and strategy optimization, thereby improving the accuracy and sustainability of mental health management for the elderly. It should be understood that this step mainly includes the following sub-steps:

[0172] Step 4.1: Construction of a multimodal intervention execution system;

[0173] This implementation constructs a multimodal intervention execution system, integrating multiple interaction channels such as virtual assistants, smart home environments, and wearable devices to achieve seamless execution of emotion-cognitive dual-channel intervention strategies. Specifically, the system is based on a microservice architecture and includes the following core components: intervention strategy parsing service, multimodal interaction coordination service, response generation service, and status monitoring service. Intervention strategies are transformed into specific execution instruction sets through microservice APIs.

[0174]

[0175] in, , , They represent the first , , Each execution instruction is handed over to the corresponding execution terminal for completion; Indicates the total number of instructions executed.

[0176] The multimodal intervention execution system is implemented as follows: The system adopts a containerized microservice architecture to ensure the independence and scalability of each component. The core microservices include:

[0177] Strategy parsing service: This service breaks down advanced intervention strategies into a sequence of specific execution instructions and handles logical relationships such as execution order, priority, and preconditions. It employs a combination of decision trees and a rule engine to ensure the correctness and consistency of the instructions.

[0178] Interactive Coordination Service: Manages the collaboration and switching between various interaction channels (such as voice assistants, smart home devices, mobile applications, etc.). It employs a publish-subscribe pattern for event-driven communication, using message queues to ensure sequential processing of interactive events.

[0179] Response generation service: Generates adapted response content for different interaction channels. For voice assistants, it uses a Natural Language Generation (NLG) model to dynamically generate personalized dialogue content; for visual interfaces, it adopts a response generation mechanism that combines templates and dynamic content.

[0180] Status Monitoring Service: Monitors multimodal sensor data in real time, assesses the immediate effects of current interventions, and detects triggering conditions that may require policy adjustments. This service employs a stream processing architecture and supports Complex Event Processing (CEP) to identify important patterns.

[0181] In specific application scenarios, the system workflow is as follows: After the system generates a collaborative intervention strategy of "increasing social participation + cognitive restructuring" based on step 3, the strategy parsing service transforms it into a series of instructions:

[0182] (1) Push community activity suggestions through smart speakers;

[0183] (2) Arrange cognitive behavioral exercises on mobile applications;

[0184] (3) Adjust the intelligent lighting system to create a positive emotional environment.

[0185] Interactive coordination services ensure that instructions are executed at the appropriate time through the most suitable device. For example, when an elderly person wakes up in the morning, a smart speaker plays a greeting and reminds them of a gardening activity in the community that day, while the smart lighting system adjusts to bright, warm tones to improve their mood. When a senior citizen's mood is detected to be low, the response generation service dynamically creates dialogue content containing positive memories to guide the senior citizen in mood regulation. Throughout the process, the status monitoring service continuously evaluates the effectiveness of the intervention. If it finds that the social activity suggestions are not accepted, the system will adjust its strategy and switch to recommending small-scale social activities within the family.

[0186] Step 4.2: Real-time evaluation of intervention effectiveness;

[0187] This implementation establishes a multi-dimensional intervention effect evaluation mechanism, including immediate emotional response assessment, cognitive change assessment, and long-term health status assessment. Specifically, the system calculates emotional state change vectors and cognitive pattern change vectors in real time using multimodal sensor data, and comprehensively evaluates the intervention effect index.

[0188] Emotional state change vector: ;

[0189] in, Represents a vector of emotional state changes; Indicates the current emotional state; It indicates the emotional state at the previous moment.

[0190] Cognitive pattern change vector: ;

[0191] in, Represents a vector of cognitive pattern changes; Indicates the current cognitive pattern; It represents the cognitive pattern of the previous moment.

[0192] Intervention Effectiveness Index: ;

[0193] in, Indicates the intervention effectiveness index; and These represent the weighting coefficients for changes in emotional state and changes in cognitive pattern, respectively. and These are evaluation functions representing changes in emotional state and changes in cognitive pattern, respectively. Represents a vector of emotional state changes; This represents a vector representing changes in cognitive patterns.

[0194] In addition, the system combines subjective feedback and objective physiological indicators from the elderly to construct an intervention acceptance index:

[0195]

[0196] in, express Intervention acceptance index at any given time; This represents the comprehensive evaluation function; Represents the set of subjective feedback features; It represents the set of objective indicator features.

[0197] The real-time evaluation of intervention effects is implemented as follows: The evaluation system adopts a hierarchical time-series evaluation architecture, including three levels: micro (second-level) response evaluation, meso (daily) state change evaluation, and macro (weekly / monthly) trend evaluation.

[0198] Microscopic assessment primarily relies on real-time physiological signals and behavioral response data. A variational autoencoder (VAE) is used to extract a low-dimensional representation of emotional states and calculate state changes over continuous time periods. Emotional state changes are quantified using the following formula:

[0199]

[0200] in, It indicates changes in emotional state at the micro level; Indicates from arrive Summation operation; Indicates the length of the time window; This represents a time-weighted coefficient, assigning greater importance to recent changes. express The emotional state at any given moment; express The emotional state at any given moment.

[0201] The meso-level assessment focuses on changes in daily behavioral patterns and cognitive performance in older adults. The system applies a sequence pattern mining algorithm to extract behavioral pattern changes from sequences of daily activities and assesses changes in cognitive function through regular, lightweight cognitive tests. The calculation of meso-level cognitive pattern changes is as follows:

[0202]

[0203] in, This indicates a change in cognitive patterns at the meso-level. Indicates from arrive Summation operation; This represents the total number of cognitive dimensions; Indicates the first Weighting coefficients for each cognitive dimension; Indicates the current time. Scoring of each cognitive dimension; express Time before Scoring of each cognitive dimension; Indicates the evaluation time interval.

[0204] Macro-level assessments comprehensively analyze the trend changes of long-term health indicators, social activity participation, and mental health scale scores, using time series decomposition techniques to separate trend, seasonal, and random components, with a focus on the direction and slope of trend component changes.

[0205] In specific application scenarios, after the system implements a one-month intervention program for an elderly person, the real-time evaluation system can generate multi-level effect reports: at the micro level, it shows that the elderly person's average physiological stress indicators (such as heart rate variability) improved when receiving virtual natural environment intervention; at the meso level, it found that the frequency of their daily social interactions increased and their cognitive flexibility test score improved; at the macro level, it shows that the Geriatric Depression Scale (GDS) score decreased, indicating a significant relief of depressive symptoms. The system also identified that the elderly person had a low acceptance of cognitive restructuring exercises but a high level of participation in community activities, and adjusted the intervention program accordingly, increasing the proportion of social interventions.

[0206] Step 4.3, Bayesian Adjustment and Incremental Learning;

[0207] This implementation method applies a Bayesian adjustment method and an incremental learning model to optimize intervention strategy parameters in real time based on intervention effectiveness and acceptance. Specifically, the system maintains the posterior distribution of intervention parameters, continuously updates the distribution based on observation data, and selects the optimal parameters:

[0208]

[0209] in, Indicates the optimal intervention parameters; In parameter space Searching for utility functions The largest parameter; Indicating in the observation data Under the condition, utility function The expected value.

[0210] Furthermore, the intervention strategy network parameters are updated using a deep incremental learning method, formally represented as:

[0211]

[0212] in, Indicates the network parameters of the new intervention strategy; This represents the network parameters of the old intervention strategy; Indicates the learning rate; Performance index function For parameters The gradient.

[0213] Therefore, the system can continuously accumulate intervention experience, optimize personalized intervention strategies, and form adaptive personalized intervention plans.

[0214] The specific implementation of Bayesian adjustment and incremental learning is as follows: The system adopts a Bayesian optimization framework based on Gaussian processes, representing the intervention parameter space as a multidimensional continuous space. Bayesian optimization balances exploration and exploitation through the sampling strategy in the iterative update process, thereby efficiently finding the optimal parameters under limited sample data conditions.

[0215] The core of Bayesian optimization is constructing a surrogate model to estimate the true utility function. The system uses a Gaussian process with a radial basis function (RBF) kernel as the surrogate model.

[0216]

[0217] in, The kernel function representing a Gaussian process; This represents the vector of intervention parameters to be evaluated. Represents a vector of historical intervention parameters; This represents the magnitude parameter of the covariance; Represents the natural exponential function; Indicates the length scale parameter; This represents the squared Euclidean distance between two parameter vectors.

[0218] To determine the next set of parameters to be evaluated, the system uses the Upper Confidence Bound (UCB) function:

[0219]

[0220] in, Indicates the function for obtaining the upper confidence bound; Indicates Gaussian process in Mean prediction at [location]; This indicates controlled exploration using parameters of equilibrium. Indicates Gaussian process in Predict the standard deviation at a given location.

[0221] The incremental learning part employs the Elastic Weight Consolidation (EWC) algorithm, which adapts to new data while retaining previously learned knowledge. The network parameter update formula is:

[0222]

[0223] in, Represents the total loss function; This represents the loss function for the current batch; Represents the regularization coefficient; This indicates all parameters Summation; This represents the Fisher information matrix, used to measure the importance of parameters; Indicates the current parameter; This represents the previous optimal parameter value.

[0224] In practical applications, for an elderly user receiving system intervention for the first time, the system initially provides an initial intervention strategy based on a general model. As the intervention progresses, the Bayesian optimization framework continuously updates the posterior distribution of the parameter space. For example, the system might discover that the user responds best to cognitive exercises performed in the evening, and that the optimal intervention duration for music therapy is 15-20 minutes (where mood improvement is most significant). Simultaneously, the incremental learning module adjusts the intervention content generation model based on the user's specific feedback, such as gradually shifting the voice assistant's communication style from formal to humorous, as data shows the user responds more positively to humorous content. After 2-3 weeks of continuous optimization, the system develops a highly personalized intervention plan, with an improved intervention effectiveness index compared to the initial strategy, leading to increased user satisfaction.

[0225] Step 5: Integrate the functional modules into the integrated medical and elderly care knowledge service system to achieve collaborative work with other health management modules and build a complete ecosystem for elderly mental health services;

[0226] This step integrates the functional modules (technical components) from steps 1 to 4, including the emotional dynamic analysis module, emotional cognitive interaction modeling module, dual-channel collaborative intervention module, and personalized intervention execution module, into the integrated medical and elderly care knowledge service system. This enables collaborative work with other health management modules, constructing a complete ecosystem for elderly mental health services. The main steps include:

[0227] Step 5.1: Modular system architecture construction;

[0228] This application constructs a modular system architecture, encapsulating each technical component into standardized service modules to form a scalable integrated medical and elderly care knowledge service system. Specifically, the system adopts a microservice architecture, including the following core modules: emotional dynamic analysis service, emotional cognitive interaction modeling service, dual-channel collaborative intervention service, and personalized intervention execution service. It should be noted that the modules communicate through standardized APIs, employing an event-driven model to achieve loosely coupled integration. Furthermore, the system constructs a unified data storage layer, including a time-series database and a knowledge graph database, to achieve efficient data management and querying. Therefore, the data flow can be represented as:

[0229]

[0230] in, Represents the raw data stream; This indicates the emotional dynamics analysis module; This represents the emotion-cognition interaction modeling module; This indicates a dual-channel collaborative intervention module; This indicates the personalized intervention execution module; This represents the feedback data stream, used to continuously optimize the performance of each module.

[0231] The modular system architecture is implemented as follows: The system adopts a container orchestration-based cloud-native architecture to ensure the independent deployment, scaling, and maintenance of each service module. Core technical components include:

[0232] Service Mesh Layer: Employs service mesh technologies such as Istio to manage communication between microservices, enabling traffic control, fault recovery, and secure communication. Service discovery and registration are implemented through the Kubernetes service registry, ensuring dynamic connectivity between system components.

[0233] API Gateway Layer: Utilizes GraphQL as the API interface specification, providing a unified query language and flexible data retrieval capabilities. The API gateway implements functions such as request routing, load balancing, authentication and authorization, and rate limiting.

[0234] Data persistence layer: Employs a multi-model database architecture, including:

[0235] InfluxDB (Time Series Database): Stores time-series data on the emotional state and physiological indicators of the elderly.

[0236] Graph database (Neo4j): Stores knowledge graph structures to express complex relationships between health concepts;

[0237] Document database (MongoDB): Stores unstructured or semi-structured user profiles, intervention records, etc.

[0238] Distributed caching (Redis): Improves the performance of frequently accessed data reads.

[0239] Message Queue System: Utilizes Kafka for event-driven inter-module communication, supporting asynchronous data stream processing and system decoupling. Event topics include:

[0240] Emotional state change event (emotion.state.change);

[0241] Intervention strategy generates events (intervention.strategy.created);

[0242] Intervention execution event (intervention.execution.completed);

[0243] User feedback event (user.feedback.received).

[0244] Data Pipeline: Spark Streaming is used to process real-time data streams, and ETL tools are used to process batch data, ensuring efficient data flow and transformation between modules.

[0245] In a specific application scenario, the workflow of this modular architecture is as follows: When an elderly person begins using the system, they first register their personal information and authorize data access permissions through the API gateway. Subsequently, various sensing devices begin collecting data, which is transmitted to the system via the Internet of Things (IoT) gateway. The emotional dynamic analysis service processes this raw data, extracts emotional features, stores the results in a time-series database, and publishes emotional state change events via Kafka. The emotional cognitive interaction modeling service subscribes to these events, combines historical data to build a personalized interaction model, identifies key intervention nodes, and stores the results in a graph database. When the system detects a potential risk, the dual-channel collaborative intervention service generates an intervention strategy based on the current state and publishes an intervention strategy generation event. The personalized intervention execution service receives this event, coordinates multiple terminal devices to execute the intervention, and simultaneously collects feedback data to form a closed-loop optimization. For example, if the system detects that an elderly person living alone has experienced a decrease in emotional variability for three consecutive days, it will immediately initiate emotional dynamic analysis. After confirming the warning, the collaborative intervention service will generate a dual-channel intervention strategy of "enhancing social interaction + cognitive restructuring". The personalized execution service will recommend that the elderly person participate in community chess and card activities through a smart speaker based on the elderly person's interests (obtained from the user profile) and the current time (9:00 am), while pushing positive thinking exercises to their tablet device.

[0246] Step 5.2: Implementation of the collaborative work mechanism;

[0247] This application implements a collaborative working mechanism between mental health management and other health management modules. It should be noted that the system employs knowledge graph technology to construct a comprehensive picture of the health status of the elderly, including multi-dimensional information such as physical health, cognitive status, and mental health. Specifically, this is achieved through an ontology model. The relationships between the dimensions are defined as follows:

[0248]

[0249] in, Represents a health ontology model; Represents a set of health concepts; Represents a set of relationships between concepts; Represents a collection of attributes; Represents a collection of instances.

[0250] Accordingly, based on this knowledge graph, the system enables collaborative decision-making among modules such as mental health intervention, cognitive assessment, and chronic disease management, ensuring that intervention strategies are adapted to the overall health status of the elderly.

[0251] The specific implementation of the collaborative working mechanism is as follows: The system constructs a multi-layered health knowledge graph, supporting cross-domain health information integration and reasoning. The core components of the knowledge graph include:

[0252] Ontology Layer: Defines the core concepts and relationships in the medical and health care field. The ontology model includes the following main sets of concepts:

[0253] The concept of mental health includes: emotional state, cognitive patterns, mental illness, intervention methods, etc.

[0254] Physiological health concepts include: vital signs, disease state, drug treatment, and functional status.

[0255] Cognitive function concepts: memory ability, executive function, language ability, spatial perception, etc.

[0256] Social factors include: social networks, family support, community resources, and cultural background.

[0257] Relational model: Defines the relationships between concepts, including:

[0258] Causality: Represents a causal chain in which one state of health leads to another state;

[0259] Influence relationship: Indicates the degree of mutual influence between different factors;

[0260] Compositional relationships: representing the hierarchical structure of the whole and its parts;

[0261] Temporal relationship: Represents the evolution pattern of health status over time.

[0262] Inference Engine: Based on knowledge graphs, it performs health status inference and intervention decisions, including:

[0263] Rule-based reasoning: Reasoning is based on expert-defined IF-THEN rules;

[0264] Path analysis: Analyze paths in a knowledge graph to discover indirect connections;

[0265] Probabilistic reasoning: dealing with uncertainty and probabilistic relationships in health knowledge.

[0266] Conflict Detection and Reconciliation Mechanism: Identifying and resolving intervention conflicts between different health management modules, including:

[0267] Drug-psychological intervention conflict detection;

[0268] Cognitive training - prioritizing emotion management;

[0269] Coordinating the timing of physiological therapy and psychological intervention.

[0270] In specific application scenarios, the collaborative working mechanism operates as follows: When the system provides services to an elderly person suffering from both hypertension and mild depression, the knowledge graph captures the influence relationship between "hypertension medication" and "mood fluctuations." When the mental health module prepares to implement emotional intervention, the system uses the knowledge graph to find that the user is taking a beta-blocker, which may cause low mood as a side effect. The collaborative decision-making mechanism adjusts the emotional intervention strategy accordingly, increasing the proportion of cognitive restructuring, and suggests that the medical team consider adjusting the medication regimen. In another scenario, when the system detects a decline in a user's cognitive function test score, knowledge graph analysis shows that the user's recent mood fluctuations have intensified, and sleep quality has decreased. Through analysis of the association paths in the knowledge graph, the system infers that emotional problems may be upstream factors of cognitive changes, prioritizing emotional stabilization interventions before implementing cognitive training, forming a coordinated intervention plan.

[0271] Step 5.3: Generate personalized interaction strategies;

[0272] This application constructs a personalized interaction strategy generation algorithm based on the individual interests, life experiences, and social preferences of older adults. Specifically, the system maintains a user interest model and, through collaborative filtering and content analysis methods, predicts the older adult's preference for different interactive content, generating the most suitable psychological support interaction strategy. It is important to note that the interactive content library includes three categories: emotional companionship, cognitive stimulation, and social facilitation.

[0273] Emotional support content: Based on the current emotional state and personal preferences, provide music, stories, nostalgic activities, etc.

[0274] Cognitive stimulation content: Based on the results of cognitive ability assessment, adaptive cognitive training games and knowledge quizzes are provided;

[0275] Socially-facilitated content: Based on social network analysis, recommend social activities and group interaction opportunities.

[0276] In addition, the system optimizes content recommendation strategies through multi-armed gambling machine algorithms to maximize the participation and satisfaction of the elderly.

[0277] The specific implementation of personalized interaction strategy generation is as follows: The system constructs a multi-dimensional user profile of the elderly and a dynamic interaction strategy generation framework, which includes the following core components:

[0278] Multidimensional User Profile Model: Integrates multi-source data to build a comprehensive user profile, including:

[0279] Basic attribute dimensions: age, gender, education background, professional experience, etc.;

[0280] Dimensions of interests and preferences: cultural interests, activity preferences, topic interests, etc.;

[0281] Life experience dimension: important life events, memories of life stages, etc.;

[0282] Social network dimensions: family relationships, social circles, social habits, etc.

[0283] Interaction habits dimension: device preferences, interaction duration, feedback methods, etc.

[0284] Content Representation and Matching System: This system establishes semantic representations for interactive content and matches them with user profiles. Technical implementation includes:

[0285] Content vectorization: Using language models such as BERT to generate semantic vectors for text content;

[0286] Multimodal feature extraction: Extracting feature representations from content such as images and audio;

[0287] Similarity calculation: Methods such as cosine similarity are used to calculate the matching degree between user interests and content.

[0288] Hybrid recommendation algorithm: combines multiple recommendation methods to generate personalized interactive content.

[0289] Content-based recommendation: Recommendations based on semantic similarity of content;

[0290] Collaborative filtering: Recommendation based on the preferences of similar user groups;

[0291] Knowledge graph-enhanced recommendation: using domain knowledge to guide the recommendation process;

[0292] Context-aware recommendation: taking into account contextual factors such as time, place, and emotion.

[0293] Exploration-Exploitation Balanced Strategy: Employing the Contextual Multi-Armed Bandit (CMAB) algorithm to balance exploring new content with exploiting known preferences.

[0294] LinUCB algorithm implementation: linearly maps contextual features to reward prediction;

[0295] Thompson sampling: Handles uncertainty by sampling from the posterior distribution;

[0296] Layered exploration strategy: Explore at two levels: content category and specific content.

[0297] In a specific application scenario, the workflow of the personalized interaction strategy is as follows: Through initial surveys and continuous interaction, the system learns that a 78-year-old retired teacher has a particular interest in classical music, gardening, and history, habitually uses a tablet in the morning and evening, and has volunteered at a local history museum. When this user experiences low mood, instead of providing generic emotional support content, the system, based on their user profile, pushes music therapy content including Beethoven's Sixth Symphony ("Pastoral"), paired with a photo collection showcasing the historical changes in their hometown, and recommends volunteer opportunities at a nearby botanical garden. The system records the user's reaction to each recommendation (duration of stay, mood changes, direct feedback, etc.) and updates the recommendation model. Simultaneously, the system periodically explores new content types. For example, after detecting that the user has browsed several articles about astronomy, the system will try recommending activities at the local planetarium, observing feedback and adjusting the interest weights in the user profile. Through this continuous learning and adjustment process, the system gradually develops a highly personalized interaction strategy, improving the acceptance and effectiveness of psychological support.

[0298] Step 5.4: Dynamic updating and optimization of the knowledge base;

[0299] This application establishes a dynamic knowledge base update and optimization mechanism to continuously accumulate experience from intervention practices and improve overall service quality. Specifically, the system employs an incremental learning method, periodically updating the parameters of each model based on newly collected intervention effect data.

[0300]

[0301] in, This represents the updated model parameters; This represents the model parameters before the update; Indicates based on new data The calculated parameter increment.

[0302] It should be understood that the system also maintains a case library, recording typical intervention cases and their effects, supporting case-based reasoning in decision-making. Therefore, the knowledge base update process is monitored by the following quality control indicators: model performance indicators, intervention effect indicators, and user satisfaction indicators, ensuring that the update process continuously improves system performance.

[0303] The specific implementation method for dynamic updating and optimization of the knowledge base is as follows: The system establishes a multi-level knowledge accumulation and optimization framework to realize the continuous accumulation and application of experiential knowledge:

[0304] Distributed knowledge base architecture: Building a hierarchical, multi-perspective knowledge management system.

[0305] Core knowledge layer: Contains proven expertise and best practices;

[0306] Experience knowledge layer: empirical rules extracted from actual intervention cases;

[0307] Individual knowledge layer: Personalized knowledge tailored to specific user groups;

[0308] Temporary knowledge layer: Newly discovered knowledge and hypotheses to be verified.

[0309] Case library construction and retrieval mechanism: System collection, organization, and indexing of typical intervention cases:

[0310] Case representation: Each case is structured into a "context-intervention-outcome" triple;

[0311] Similarity calculation: Case similarity is calculated based on weighted feature matching;

[0312] High-efficiency retrieval: Employs techniques such as Locality Sensitive Hash (LSH) to achieve fast case retrieval;

[0313] Case adaptation: Adjusting intervention strategies for historical cases according to the current context.

[0314] Incremental learning and model update mechanism: The system continuously learns and improves the models of each component.

[0315] Incremental parameter calculation: ;

[0316] in, This represents the parameter increment calculated based on the new data; Indicates the learning rate; Indicates the parameter Find the gradient; Represents the loss function; Indicates new data; This represents the model parameters before the update;

[0317] Catastrophic forgetting prevention: Preventing new knowledge from overwriting old knowledge through experience replay and resilient weight merging;

[0318] Model version control: Maintains model version history, supports rollback and A / B testing;

[0319] Federated learning framework: Aggregating the learning outcomes of multiple instances while protecting privacy.

[0320] Quality Control and Knowledge Verification: Establishing a Multi-Dimensional Quality Assessment System

[0321] Model performance metrics Precision, recall, F1 score, etc.

[0322] Intervention effect indicators : Mood improvement rate, cognitive enhancement, etc.;

[0323] User satisfaction index Usage frequency, subjective ratings, dropout rate, etc.

[0324] Knowledge consistency check: Ensure that new knowledge is consistent with the existing knowledge system.

[0325] In specific application scenarios, the workflow for dynamic updating and optimization of the knowledge base is as follows: After serving hundreds of elderly users for several months, the system accumulated a large amount of intervention data. Through data analysis, the system discovered a significant pattern: for female users aged 70-75 with a teaching background, the intervention effect of "interest-based social activities" was on average 32% higher than that of "general social activities." This finding was recorded in the experiential knowledge layer and the intervention strategy generation model was updated through incremental learning. The system also identified a particularly successful case: an elderly person living alone joined a community choir through the system's recommendation, and their depressive symptoms significantly improved. This case was recorded in detail in the case library, including user characteristics, intervention process, and effect evaluation. When the system encounters new users with similar characteristics, it will adjust the intervention strategy based on this successful case. At the same time, the system continuously monitors the performance indicators of all models. When it finds that the accuracy of the emotional dynamic analysis model in identifying "hidden anxiety" is declining, it will trigger a targeted data collection and model optimization process. Through this cyclical and gradual knowledge accumulation and optimization process, the overall service quality of the system gradually improves, and both intervention effect indicators and user satisfaction indicators show a stable upward trend.

[0326] The implementation method provided in this application, through an innovative combination of key technologies such as emotional dynamic analysis, emotional cognitive interaction modeling, dual-channel collaborative intervention, and personalized intervention execution, realizes a medical and elderly care integrated knowledge service system, which has the following significant technical effects:

[0327] Improved accuracy of early warning: By extracting emotional dynamic features through nonlinear time series analysis and a hybrid neural network architecture, the system in this application can capture emotional change patterns that are difficult to detect using traditional methods. It should be understood that, compared to traditional methods based solely on cross-sectional assessments, this implementation can identify deep dynamic features such as emotional complexity and shift patterns, effectively identifying potential mental health risks in advance and providing a critical time window for early intervention.

[0328] Significantly improved intervention efficiency: Based on the emotional-cognitive interaction model and key node identification algorithm, the system in this application can accurately locate the key links in the interaction between emotion and cognition, enabling targeted intervention. Furthermore, the dual-channel collaborative intervention strategy, through multi-agent reinforcement learning, simultaneously regulates emotional states and reconstructs cognitive patterns, effectively breaking the vicious cycle of emotion and cognition and significantly improving the output ratio of unit intervention resources.

[0329] Comprehensive Enhancement of Personalization: This application's system utilizes multi-layered personalization technologies, including individualized interaction model training, adaptive intervention timing scheduling, and dynamic strategy optimization, to achieve end-to-end personalization from psychological state assessment to intervention strategy generation. Therefore, this precise "one-size-fits-all" service model can adapt to the psychological characteristics, lifestyles, and personal preferences of different elderly individuals, significantly improving the acceptance and adherence to intervention programs.

[0330] A qualitative improvement in user experience: Through implicit psychological assessment mechanisms and situational adaptive interventions, this application's system naturally integrates mental health management into the daily lives of the elderly, avoiding the labeling effect and sense of rejection caused by direct assessment and intervention in traditional methods. It should be noted that this "seamless" service approach significantly increases system stickiness and the frequency of continuous interaction, laying the foundation for long-term mental health management.

[0331] A virtuous cycle of knowledge accumulation: The modular architecture and dynamic update mechanism of the knowledge service system enable the system in this application to continuously accumulate experience from practice, optimize model parameters and intervention strategies, and form a virtuous cycle of experiential knowledge accumulation. Furthermore, this evolutionary system design ensures continuous improvement in service quality and adaptability to emerging problems.

[0332] Collaborative management of overall health: Through a collaborative working mechanism implemented using knowledge graph technology, the system in this application can organically combine mental health with other dimensions such as cognitive state and physical health, avoiding the limitations of single-dimensional management, providing comprehensive and integrated health management services for the elderly, and significantly improving their overall quality of life.

[0333] In summary, the implementation method of this application breaks through the technical barriers of traditional mental health management, and achieves a deep integration of emotional dynamic analysis, emotional cognitive interaction modeling and intelligent intervention. It provides a strong support capability for the mental health of the elderly for the integrated medical and elderly care knowledge service system, and effectively addresses the increasingly severe mental health challenges of the elderly in the context of population aging.

[0334] Application examples of this implementation method

[0335] To verify the feasibility and effectiveness of this implementation method, a six-month real-world application test was conducted at a municipal elderly care center. The center housed 120 elderly residents aged 65 and above, including 42 living alone, with an average age of 73.5 years. This section will detail the specific implementation process and effectiveness verification of the system in practical application.

[0336] 1. Application Scenarios and System Deployment:

[0337] The application scenario of the integrated medical and elderly care knowledge service system in elderly care centers, as described in this embodiment, mainly involves the early identification and continuous intervention of the mental health of elderly people living alone. The system's hardware includes: smart terminal devices (including cameras, microphones, and touchscreens) installed in each elderly person's home, wearable physiological monitoring bracelets, a smart home environment control system, and a central server cluster. The system deployment adopts a cloud-edge-device architecture, with core algorithms running on the central server, real-time interactive functions executed on edge devices, and data collection and preliminary processing completed on the terminal devices.

[0338] In the initial stage of system deployment, baseline data was collected from each elderly person for two weeks to establish a benchmark model of individualized emotional expression and cognitive patterns. With the consent of the elderly, the system collected multimodal data, including voice, facial expressions, daily activity habits, and physiological indicators, for subsequent extraction and analysis of emotional dynamic features.

[0339] 2. Example of Emotional Dynamic Feature Extraction and Analysis:

[0340] Taking a 73-year-old woman living alone (E037) as an example, the system collects her voice, facial expressions, and behavioral data daily through a smart terminal device. Using a deep multimodal fusion algorithm, the system calculates the elderly woman's emotional dynamic feature vector over 30 consecutive days, including emotional intensity. (Range 0-1) Emotional diversity (Range 0-5) Emotional inertia (Range 0-1) and emotional resilience (Range 0-1).

[0341] By performing nonlinear analysis on the collected emotional dynamic feature vector time series, the system calculates the entropy increase rate of the elderly person's emotional trajectory. Lyapunov index and fractal dimension Reference values ​​for healthy older adults ( ) , , In contrast, the elderly person's emotional complexity index showed a higher entropy increase rate and a lower fractal dimension, indicating reduced predictability and complexity of emotional changes.

[0342] Through analysis using a hybrid architecture of Hidden Markov Model and LSTM, the system discovered that the elderly person had recently exhibited an abnormal transition pattern from "mild anxiety" to "moderate depression," with the transition probability increasing from 0.12 at baseline to 0.37, exceeding the warning threshold (0.25). The system then generated a warning signal, marking the elderly person as having potential depressive risk, and initiated the subsequent process of constructing an emotional cognitive interaction model.

[0343] 3. Example of constructing an emotion-cognition interaction model:

[0344] In response to the warning signal E037 from the elderly user, the system further constructed an emotional cognitive interaction model. By analyzing the dialogue between the elderly user and the system using natural language processing technology, and combining this with a cognitive assessment questionnaire, the system constructed an emotional state vector. and cognitive pattern vector These correspond to different emotional dimensions (sadness, anxiety, fatigue, loneliness, helplessness) and cognitive dimensions (catastrophic thinking, overgeneralization, negative filtering, personalization, black-and-white thinking).

[0345] By analyzing the elderly person's emotional and cognitive interaction history data over the past three months, the memory-enhanced attention network found that when the "loneliness" emotional dimension increased, it was often accompanied by an enhancement of the "overgeneralization" cognitive pattern, and this change showed a clear temporal dependency. By calculating the partial derivative matrix, the system found that the influence weight of "loneliness → overgeneralization" (0.72) was much higher than the influence weight between other dimensions (average 0.31), and identified this node as a key intervention node.

[0346] Meanwhile, the system also discovered that the elderly person's "negative filtering" cognitive pattern had a strong feedback amplification effect on the "sadness" emotional dimension (influence weight 0.68), forming a vicious cycle of emotional cognition. Based on the identification of these key nodes, the system determined the key emotional and cognitive dimensions for intervention.

[0347] 4. Example of generating a dual-channel collaborative intervention strategy:

[0348] For the emotional cognitive interaction model of elderly person E037, the system applies a multi-agent reinforcement learning algorithm to generate personalized intervention strategies. Emotional intervention agent. and cognitive intervention intelligent agents Based on the elderly person's status information (including emotional state, cognitive patterns, and environmental factors), action suggestions for emotional intervention and cognitive intervention are generated respectively.

[0349] After multiple rounds of collaborative decision-making optimization, the system identified the following dual-channel intervention strategy: the emotional channel focuses on "increasing social interaction" (weight 0.65) and "music therapy" (weight 0.35); the cognitive channel focuses on "cognitive restructuring" (weight 0.70) and "mindfulness practice" (weight 0.30). System analysis revealed that the expected effect of this dual-channel collaborative intervention strategy (predicted value 0.78) was significantly higher than that of simple emotional intervention (predicted value 0.42) or simple cognitive intervention (predicted value 0.51).

[0350] In terms of timing, the system decided to first implement emotional intervention to alleviate the elderly person's loneliness, and then gradually introduce cognitive intervention to change their overgeneralization thinking patterns. This progressive intervention sequence design is based on the elderly person's acceptance model, which can maximize the overall effect of the intervention.

[0351] 5. Examples of Personalized Intervention Implementation and Feedback Adjustment:

[0352] In the actual intervention process, for elderly person E037, the system implemented the above intervention strategy through a multimodal intervention execution system. Specifically, the system first recommended that the elderly person participate in the choir activities organized by the care center (to address social loneliness) through a smart terminal device, and played her favorite classical music in the evening (based on her interest model) to regulate her mood.

[0353] After initial emotional intervention proved effective, the system began guiding the elderly woman through video dialogues to practice cognitive restructuring, challenging her overgeneralized thinking that "no one understands me," and helping her identify and adjust this irrational cognition. Simultaneously, the system also provided short mindfulness meditation audio clips to help the elderly woman focus on the present moment and reduce interference from negative thoughts.

[0354] The system monitored the intervention's effectiveness in real time and found that the elderly person had a high acceptance rate for social activities (participation rate of 85%), but a low acceptance rate for cognitive restructuring exercises (completion rate of only 40%). Based on this feedback, the system optimized the intervention parameters using a Bayesian adjustment method, changing the form of cognitive intervention from direct cognitive challenges to a more indirect story analogy approach. This adjustment increased the completion rate of cognitive exercises to 75%.

[0355] 6. Continuous monitoring and evaluation of intervention effectiveness:

[0356] The system continuously monitored and intervened with elderly patient E037 for three months, and regularly assessed the intervention's effectiveness. Micro-level assessments showed that the elderly patient's heart rate variability increased by an average of 25% during music therapy intervention, indicating an effective reduction in stress levels. Meso-level assessments found that the frequency of daily social interactions increased from 1-2 times per week to 4-5 times per week, and the "overgeneralization" score on the cognitive assessment test decreased from 7.8 to 4.2 (out of 10).

[0357] The macro-level assessment used standard scales. The Geriatric Depression Scale (GDS) score decreased from 14 points (mild depression range) before the intervention to 8 points (normal range), indicating a significant improvement in depressive symptoms. The quality of life assessment questionnaire score increased from 62 points before the intervention to 79 points (out of 100), showing a significant improvement in the elderly person's overall life satisfaction.

[0358] The system identified that social activities had the most significant effect on improving the elderly person's mood, storing this empirical knowledge in the individual's knowledge layer for future intervention strategy optimization. Simultaneously, the system determined the elderly person's optimal receptiveness to intervention measures (9-11 am and 4-6 pm), and adjusted the intervention schedule accordingly to further improve intervention efficiency.

[0359] 7. Verification of technical effectiveness:

[0360] In the application testing throughout the elderly care center, the medical and elderly care integrated knowledge service system of this implementation showed significant technical effects in terms of early warning accuracy and intervention efficiency.

[0361] The following is quantitative verification data on the effectiveness of the core technology:

[0362] Early warning accuracy verification results: The system's early identification results of depression and anxiety in the elderly were compared with the diagnoses of professional psychologists. Among 42 elderly people living alone, the system successfully issued early warnings for 15 elderly people with potential mental health risks. Of these, 13 were diagnosed with mild depression or anxiety disorder by professional doctors within 2-3 months, with an early warning accuracy rate of 86.7%. The system's average early warning time was 8.5 weeks earlier than traditional scale screening methods, providing a valuable time window for early intervention. Notably, among the 5 elderly people identified by the system who were in a subclinical state (not meeting the diagnostic criteria but having obvious symptoms), 4 avoided further deterioration of their symptoms through timely intervention, demonstrating a significant preventive effect.

[0363] Intervention Efficiency and Effectiveness Validation Results: Comparative studies showed a significant difference in the degree of mental health improvement between elderly individuals receiving the dual-channel synergistic intervention of this system (experimental group, n=20) and those receiving only traditional psychological counseling (control group, n=22). The experimental group experienced an average decrease of 5.8 points in their GDS scores, while the control group only saw a decrease of 2.3 points. The experimental group also saw an average increase of 15.7 points in their quality of life questionnaire scores, compared to an increase of 7.2 points in the control group. Furthermore, the average intervention time required to achieve the same therapeutic effect in the experimental group was 41% shorter than that in the control group, indicating a significant improvement in intervention efficiency. The advantages of the dual-channel synergistic intervention were particularly evident for elderly individuals with both emotional problems and cognitive distortions, validating the effectiveness of the emotional-cognitive interaction modeling and dual-channel synergistic intervention strategies implemented in this method.

[0364] The above real-world application examples fully demonstrate the technical feasibility and effectiveness of this implementation method in early warning and precise intervention for the mental health of the elderly. The system's modular design and adaptive optimization mechanism ensure its broad applicability to different elderly populations, providing strong technical support for addressing the increasingly serious mental health problems of the elderly in the context of an aging population.

[0365] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for constructing a knowledge service system integrating medical and elderly care for the elderly, characterized in that, Includes the following steps: Collect emotional expression data from the elderly, apply nonlinear time series analysis and deep learning models to extract emotional dynamic features, and identify early signs of declining emotional regulation ability; The extraction of emotional dynamic features specifically includes: Multimodal data of older adults, including voice, facial expressions, activity patterns, and physiological signals, are collected through sensor networks and used as the basic input for emotional state analysis. A deep multimodal fusion algorithm is applied to extract feature parameters of emotional expression, including emotional intensity, emotional diversity, emotional inertia, and emotional resilience, to construct a dynamic feature vector of emotion. in, Represents the dynamic feature vector of emotions. Indicates the intensity of emotional expression. An index representing the diversity of emotion types. Indicates the duration of an emotional state. This indicates the rate at which one recovers from negative emotions to the baseline state. Represents the transpose of a vector; A nonlinear time series analysis method is used to analyze the complexity of the time series of emotional dynamic feature vectors, quantifying the complexity changes of the emotional dynamic system. This nonlinear time series analysis method includes calculating the entropy increase rate, Lyapunov exponent, and fractal dimension of the emotional trajectory to quantify the complexity changes of the emotional dynamic system. The expression for the entropy increase rate is: in, Indicates the rate of entropy increase; Indicates when The limit as it approaches infinity; Represents the normalization coefficient; Represents the Shannon entropy function; , , They represent the first , , Emotional state at a specific point in time Indicates the length of the time series; The Lyapunov index: in, Lyapunov index; Represents a continuous-time variable; This indicates the deviation from the initial state; Indicates time Deviation at time; Indicates when The limit as it approaches infinity; This represents the limit as the initial deviation approaches 0; Represents the time normalization coefficient; Represents the natural logarithm; The deep learning model employs a hybrid architecture of Hidden Markov Model and Long Short-Term Memory Network to capture the long-term evolution patterns of emotional state transitions and identify anomalous transition patterns. Based on the acquired emotional dynamic features, a memory-enhanced attention network combined with a psychological knowledge graph is used to construct an individualized emotional cognitive interaction model for the elderly, revealing the dynamic interaction mechanism between emotional state and cognitive pattern, and identifying key intervention nodes. The construction of the emotional cognition interaction model specifically includes: Based on emotional dynamics and cognitive assessment data, construct emotional state vectors and cognitive pattern vectors; A memory-enhanced attention network is used to construct a bidirectional influence model of emotion and cognition, capturing the mutual influence between the two, and storing historical state information through an attention weight matrix; Based on individual data of the elderly, an optimization algorithm combining gradient descent and variational inference is used to train an individualized emotional cognitive interaction model. By applying causal inference methods and based on a trained emotional cognitive interaction model, we can identify key nodes in an individual's specific emotional cognitive interaction, which will serve as the key targets for the formulation of subsequent intervention strategies. The dual-channel collaborative intervention strategy specifically includes: Based on the emotion-cognition interaction model, the intervention optimization objective function and constraints are defined; Construct a multi-agent system composed of emotional intervention agents and cognitive intervention agents, and maximize joint rewards by sharing state information and coordinating action choices through a collaborative mechanism; The intervention strategy is optimized using a deep deterministic strategy gradient algorithm, and a collaborative decision-making mechanism is designed to coordinate the timing of emotional and cognitive interventions to maximize synergistic effects. Based on the constructed emotional cognition interaction model, a multi-agent reinforcement learning algorithm is applied to generate a dual-channel collaborative intervention strategy for emotional cognition, thereby maximizing the precise intervention effect. Implement the generated intervention strategies, establish a closed-loop intervention implementation and feedback mechanism, realize dynamic evaluation of intervention effects and strategy optimization, and improve the accuracy and sustainability of mental health management for the elderly; By integrating functional modules into the integrated medical and elderly care knowledge service system, it can achieve collaborative work with other health management modules and build a complete ecosystem for elderly mental health services. The functional modules include: emotional dynamic analysis module, emotional cognition interaction modeling module, dual-channel collaborative intervention module, and intervention execution module; The intervention specifically includes: Construct a multimodal intervention execution system, integrate multiple interaction channels, and achieve seamless execution of the dual-channel intervention strategy of emotion and cognition; Establish a multi-dimensional intervention effect evaluation mechanism, including immediate emotional response evaluation, cognitive change evaluation, and long-term health status evaluation; By applying Bayesian adjustment methods and incremental learning models, intervention strategy parameters are optimized in real time based on intervention effects and acceptance. Knowledge service system integration and application specifically include: Construct a modular system architecture, encapsulate technical components into standardized service modules, and form a scalable integrated medical and elderly care knowledge service system; To establish a collaborative working mechanism between mental health management and other health management modules, and to construct a comprehensive picture of the health status of the elderly; Based on the elderly’s personal interests, life experiences and social preferences, a personalized interaction strategy generation algorithm is constructed. Establish a dynamic update and optimization mechanism for the knowledge base, continuously accumulate experience from intervention practices, and improve the overall service quality; The memory-enhanced attention network consists of three parts: an encoder, a memory module, and a decoder. The memory module maintains a historical state buffer, and the multi-head attention mechanism is used to calculate the correlation between the current state and the historical memory. The sensor network includes: an audio acquisition module, a video acquisition module, a wearable physiological signal monitoring module, and an environmental interaction behavior recording module.

2. A system for constructing an integrated medical and elderly care knowledge service system for elderly care, used to execute the method for constructing an integrated medical and elderly care knowledge service system for elderly care as described in claim 1, characterized in that, include: The emotional dynamics analysis module is used to collect and analyze multimodal emotional data of the elderly, extract emotional dynamics features, identify early signs of declining emotional regulation ability, and provide basic data support for subsequent interventions. The Emotional Cognition Interaction Modeling Module is used to construct individualized emotional cognition interaction models, reveal the dynamic relationship between emotional states and cognitive patterns, identify key intervention nodes, and provide a theoretical basis for formulating intervention strategies. The dual-channel collaborative intervention module is used to generate emotion and cognition dual-channel collaborative intervention strategies based on multi-agent reinforcement learning, so as to achieve the optimal combination of emotion intervention and cognitive intervention and maximize the intervention effect. The personalized intervention execution module is used to execute intervention strategies through multimodal interaction channels, evaluate intervention effects in real time, and dynamically adjust intervention parameters to ensure the accuracy and continuity of the intervention. The knowledge service integration module is used to integrate various functional modules, manage knowledge base updates, coordinate interfaces with other health management systems, and provide a unified service interface. Each module communicates through a standardized API, and loosely coupled integration is achieved using an event-driven model.

Images