Multifunctional intelligent old-age care service platform

Abstract

The invention discloses a multifunctional intelligent old-age care service platform which comprises a multi-modal data acquisition module, a data processing and intelligent analysis module, an intelligent service module and an emotional social module. The multi-modal data acquisition module is used for acquiring physiological signs, behaviors and living environment data of old people; the data processing and intelligent analysis module fuses and processes the data and predicts a behavior trend; the intelligent service module provides intelligent health management and personalized active services accordingly; the emotion social module provides an emotion recognition virtual assistant and a social interaction platform. Compared with the prior art, through multi-dimensional data acquisition and deep analysis, the problems of single data and service lagging of a traditional platform are solved, accurate service recommendation and cross-scene linkage are achieved through reinforcement learning, the emotion care function is enhanced, the blank of emotion accompanying is filled up, more comprehensive, intelligent and humanized old-age care service is provided for old people, and the service experience of the old people is improved. And the old-age service quality and the old-age life quality are improved.

Description

Technical Field

[0001] The present invention relates to the field of elderly care, and in particular to a multifunctional smart elderly care service platform. Background Art

[0002] Against the backdrop of an aging society, smart elderly care service platforms have gained widespread adoption as a crucial tool for improving the quality and efficiency of elderly care services. Existing smart elderly care service platforms typically use sensors and wearable devices to collect physiological data from the elderly, such as heart rate and step count, and analyze it using simple data processing algorithms. They also provide basic life services such as emergency calls and housekeeping appointments. Some platforms also experiment with integrating cameras and other devices to monitor elderly behavior or building simple social modules to promote interaction.

[0003] However, current smart elderly care service platforms still have many shortcomings. First, in terms of data collection and processing, they mostly rely on a single data collection method, which cannot fully obtain multi-dimensional information such as the elderly's physiological signs, behavioral habits, and living environment. In addition, data processing capabilities are limited, making it difficult to conduct in-depth analysis of the elderly's condition. Second, service accuracy is poor. Traditional platforms provide services based on static data, lacking personalization and dynamic adjustment capabilities, and cannot meet the diverse and real-time changing needs of the elderly. Third, there is a lack of emotional care. Most existing platforms focus on health monitoring and daily life services, ignoring the emotional needs of the elderly, which makes them prone to loneliness and is not conducive to their mental health. Therefore, there is an urgent need for a multifunctional smart elderly care service platform that can solve the above problems. Summary of the Invention

[0004] The main purpose of this invention is to provide a multifunctional smart elderly care service platform, which aims to achieve precise and proactive services through multimodal data collection and fusion analysis, and strengthen emotional care functions to solve problems such as single data, delayed services and lack of emotion in existing platforms.

[0005] The technical solutions of the present invention are as follows:

[0006] A multifunctional smart elderly care service platform, including:

[0007] A multimodal data acquisition module collects the elderly's physiological sign data, behavioral data, and living environment data; the physiological sign data is collected through wearable devices and contactless devices, the behavioral data is collected through cameras and sensors, and the living environment data is collected through environmental sensing devices;

[0008] A data processing and intelligent analysis module is connected to the multimodal data acquisition module to integrate and process the collected data, analyze the data, and predict behavioral trends;

[0009] An intelligent service module is connected to the data processing and intelligent analysis module, and provides intelligent health management services and personalized proactive services for the elderly according to the analysis results; the intelligent health management services include establishing a dynamic health record, disease early warning and intervention; the personalized proactive services adjust the recommendation strategy according to the service feedback through a reinforcement learning algorithm, and realize cross-scene linkage of smart home, health monitoring and life services.

[0010] An emotional social module provides a virtual assistant with emotion recognition capability, and builds an intergenerational social and community interaction platform.

[0011] In a possible implementation, in the multi-modal data acquisition module, the camera uses differential privacy technology to desensitize the video data, and only extracts skeleton key point information for fall detection and abnormal behavior early warning.

[0012] In a possible implementation, the data processing and intelligent analysis module uses a federated learning framework and a blockchain technology, each data holder locally trains a model and uploads a parameter update result to the cloud, and simultaneously uses the blockchain to encrypt and store data and trace the source; a spatio-temporal sequence deep learning model is used to analyze data and predict behavior trends; edge computing is used to deploy a lightweight detection model on an edge node to analyze data in real time.

[0013] In a possible implementation, the spatio-temporal sequence deep learning model uses an LSTM or a Transformer model, and combines the historical wake-up time and daily activity data of the elderly to determine whether the current behavior is abnormal.

[0014] In a possible implementation, the edge computing immediately triggers a local audible and visual alarm when an abnormal event is detected, and uploads key feature data to the cloud.

[0015] In a possible implementation, the dynamic health record of the intelligent service module integrates physiological sign data, behavior data, living environment data, basic medical history and physical examination data, and generates a health trend report and personalized health suggestions through an AI algorithm.

[0016] In a possible implementation, when the intelligent service module performs disease early warning and intervention, it links the intelligent medicine box and the voice assistant to remind the elderly to take medicine and adjust the diet, and arranges on-site medical services or remote consultations according to the situation.

[0017] In a possible implementation, the reinforcement learning algorithm of the intelligent service module adjusts the service recommendation priority according to the evaluation of the elderly on the nursing services, and recommends alternative service schemes based on the preferences and health status of the elderly.

[0018] In a possible implementation, the virtual assistant of the emotional social module judges the emotional state by analyzing the voice tone and chat content of the old people based on natural language processing and emotional computing technology, and initiates a chat, plays music or pushes information actively.

[0019] In a possible implementation, the intergenerational social and community interaction platform of the emotional social module introduces a gamification mechanism, including family task check-in and community activity point reward.

[0020] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0021] The technical scheme of the present application realizes comprehensive perception of the state of the old people by setting a multi-modal data acquisition module to acquire physiological sign data, behavior data and living environment data of the old people; realizes in-depth understanding and early prediction of the state of the old people by connecting the multi-modal data acquisition module with a data processing and intelligent analysis module to fuse and analyze the acquired data and predict behavior trends; realizes precision and initiative of services by connecting the data processing and intelligent analysis module with an intelligent service module to provide intelligent health management services and personalized proactive services according to the analysis results, establish a dynamic health record, conduct disease early warning and intervention, and adjust the recommendation strategy according to the service feedback through a reinforcement learning algorithm, realizing cross-scene linkage of smart home, health monitoring and life services; and realizes attention and satisfaction of emotional needs of the old people by providing a virtual assistant with emotion recognition capability through an emotional social module and building an intergenerational social and community interaction platform, thereby solving the problems of single data, service lag and emotional loss of existing platforms. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0023] Fig. 1 The structural block diagram of the present application is shown in the figure.

[0024] Fig. 2 The differential privacy video processing flowchart in the present application is shown in the figure. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0026] AsFigs. 1-2 The embodiment shown provides a multifunctional smart elderly care service platform, which comprises:

[0027] A multi-modal data acquisition module acquires physiological data, behavior data, and living environment data of the elderly. The physiological data is acquired through wearable devices and non-contact devices, the behavior data is acquired through cameras and sensors, and the living environment data is acquired through environmental perception devices.

[0028] The multi-modal data acquisition module acquires physiological data, behavior data, and living environment data of the elderly through wearable devices, non-contact devices, cameras, sensors, and environmental perception devices, and can comprehensively obtain the life and health status information of the elderly. This multi-dimensional data acquisition method changes the limitation of single data of traditional platforms, provides a rich and comprehensive data basis for subsequent data analysis and service, and enables the platform to more accurately grasp the actual needs and health status of the elderly, effectively avoids misjudgment or service omission caused by insufficient data, and improves the perception ability of the platform to the life scene of the elderly and the pertinence of service.

[0029] A data processing and intelligent analysis module is connected to the multi-modal data acquisition module, which processes the collected data and analyzes the data to predict behavior trends.

[0030] The data processing and intelligent analysis module processes the collected data and analyzes the data to predict behavior trends. This module uses data processing technology to integrate and analyze multi-source data, and excavates potential information and rules behind the data. By predicting behavior trends, it can early detect potential health problems or abnormal behaviors of the elderly, such as early warning of disease risk and identification of sudden conditions such as falls, to gain time for timely intervention and provide corresponding services, enhance the platform's ability to protect the health and safety of the elderly, and improve the timeliness and effectiveness of services.

[0031] An intelligent service module is connected to the data processing and intelligent analysis module, which provides intelligent health management services and personalized proactive services for the elderly according to the analysis results; the intelligent health management services include establishing a dynamic health record, disease warning and intervention; the personalized proactive services adjust the recommendation strategy according to the service feedback through reinforcement learning algorithm, and realize the cross-scene linkage of smart home, health monitoring, and life services.

[0032] The intelligent service module provides intelligent health management services and personalized proactive services for the elderly based on the analysis results. In terms of health management, a dynamic health record is established, and disease early warning and intervention are carried out. In terms of personalized services, the recommendation strategy is adjusted according to service feedback through reinforcement learning algorithm, realizing cross-scene linkage of smart home, health monitoring and life services. This module changes the platform from passive response to active service, provides accurate and thoughtful services according to the real-time state and personalized needs of the elderly, such as customizing diet and nursing plan according to health status, automatically linking equipment to improve living environment, improving the convenience and comfort of the elderly's life, meeting the diverse needs of the elderly, and improving the quality of elderly care services.

[0033] The emotional social module provides a virtual assistant with emotion recognition capability and builds a intergenerational social and community interaction platform.

[0034] The emotional social module provides a virtual assistant with emotion recognition capability and builds a intergenerational social and community interaction platform. This module focuses on the emotional needs of the elderly, and the virtual assistant actively communicates and interacts with the elderly by recognizing their emotions, providing emotional companionship. The intergenerational social and community interaction platform promotes communication between the elderly and their families and peers, breaking the narrow limits of the elderly's social circle. By enriching the spiritual life of the elderly and reducing their sense of loneliness, it helps the elderly maintain a positive and optimistic attitude, promotes their mental health, and fills the gap in emotional care of traditional elderly care platforms, making elderly care more comprehensive and humanized.

[0035] In this embodiment, in the multi-modal data acquisition module, the camera uses differential privacy technology to desensitize the video data, and only extracts the skeletal key point information for fall detection and abnormal behavior warning.

[0036] Specifically, the steps of desensitizing video data by differential privacy technology are as follows:

[0037] Data preprocessing: After the camera collects the original video data, the video is split by frame and converted into an image sequence. Each frame of image is processed by grayscale to reduce data complexity and reduce subsequent computational load.

[0038] Add noise: For each pixel value in the image, add noise according to the set privacy budget parameter using the Laplace mechanism.

[0039] Skeletal key point extraction: Use a human pose estimation model based on deep learning (such as OpenPose) to detect skeletal key points in the image with added noise. The model inputs the image with added noise and outputs the coordinate position information of each joint of the human body (such as shoulder joint, elbow joint, knee joint, etc.), only retaining these skeletal key point data, and discarding other pixel information of the image.

[0040] Data transmission and storage: After encoding and compressing the extracted skeletal key point data, it is transmitted to the data processing and intelligent analysis module for subsequent fall detection and abnormal behavior warning analysis. When storing, encrypted storage method can be used to further protect data security, and relevant privacy budget parameters and other information are recorded to facilitate subsequent audit and verification of privacy protection effect.

[0041] Dynamic adjustment of privacy protection strength: According to different scenes and user needs, dynamically adjust the privacy budget parameters. For example, in the private activity area of the elderly, appropriately reduce the privacy budget and enhance privacy protection; in the public activity area, the privacy budget can be appropriately increased to improve the accuracy of behavior monitoring under the premise of ensuring privacy.

[0042] Differential privacy technology is used to desensitize video data collected by the camera, only retaining skeletal key point information required for fall detection and abnormal behavior warning. Under the premise of ensuring the privacy and safety of the elderly, the behavior state of the elderly is effectively monitored. This technology avoids the privacy leakage risk that may be caused by traditional video monitoring, so the elderly do not need to worry about their personal privacy being infringed, and thus are more willing to accept and cooperate with the platform's monitoring service; at the same time, the accurately extracted skeletal key point information can meet the functional requirements of fall detection and abnormal behavior warning, ensuring that the platform does not reduce the monitoring ability of the elderly's safety while protecting privacy, achieving a balance between privacy protection and safety monitoring, and improving the security and user acceptance of the smart elderly care service platform.

[0043] In this embodiment, the data processing and intelligent analysis module uses a federated learning framework and blockchain technology. Each data holder trains the model locally and uploads the parameter update results to the cloud, while using blockchain to encrypt and store data and trace the source. A spatio-temporal sequence deep learning model is used to analyze data and predict behavior trends. Edge computing is used to deploy lightweight detection models on edge nodes for real-time data analysis.

[0044] The spatio-temporal sequence deep learning model uses LSTM or Transformer model to determine whether the current behavior is abnormal based on the elderly's historical wake-up time and daily activity data.

[0045] When edge computing detects an abnormal event, it immediately triggers a local audible and visual alarm and uploads key feature data to the cloud.

[0046] Specifically, the implementation steps of the federated learning framework are as follows:

[0047] System architecture building: build a federated learning network composed of a cloud server and multiple data holders (such as elderly care institutions, community health service centers, and hospitals). Each data holder deploys a local model training environment, and the cloud server is responsible for model parameter aggregation and distribution.

[0048] Model initialization: The cloud server selects appropriate base models (such as multi-layer perceptron, convolutional neural network) according to the needs of pension services (such as health risk prediction, behavior anomaly detection), and distributes the initial parameters of the model to each data holder.

[0049] Local training and parameter uploading: The data holder uses the locally stored health and behavior data of the elderly to train the model, and after the training is completed, the model parameter update result (such as the change of weight and bias) is extracted and uploaded to the cloud server through a secure encryption channel. For example, the pension institution trains the local model using the daily activity data and health monitoring data of the elderly in the institution, and uploads the parameter update value obtained after training.

[0050] Parameter aggregation and delivery: The cloud server receives the parameter update results of all data holders, updates the global model parameters using aggregation algorithms such as weighted average (assigning weights according to the data size of each data holder), and then delivers the updated model parameters to each data holder. Repeat the above process until the model converges.

[0051] Specifically, the application steps of blockchain technology are as follows:

[0052] Blockchain network construction: Build a consortium chain, and the participants include data holders, platform operators, regulatory authorities, etc. Set the permissions of each participant, such as data holders can only write local data hash values and read authorized shared data, and regulatory authorities can view data operation records, etc.

[0053] Data storage and encryption: The data holder performs a hash operation on the original data locally to obtain a data hash value. Package the data hash value, data meta information (such as data type, collection time), and model parameter update results into a block, encrypt the block content using asymmetric encryption algorithms (such as RSA), and add the encrypted block to the blockchain through a consensus mechanism (such as the Practical Byzantine Fault Tolerance algorithm PBFT).

[0054] Data traceability and verification: When verifying the authenticity of data or tracing the source of data, the chain structure of the blockchain can be used to trace back from the current block to the source block where the data was generated. Use hash value comparison and digital signature to verify whether the data has been tampered with, and ensure the credibility of the data.

[0055] Specifically, the application steps of the spatiotemporal sequence deep learning model are as follows:

[0056] Data preprocessing: The health data of the elderly (such as changes in heart rate and blood pressure over time) and behavior data (such as daily activity trajectory and sleep-wake time) obtained by the multi-modal data acquisition module are arranged in chronological order into a fixed format data set, and normalized to eliminate the influence of different data dimensions.

[0057] Model selection and building: Select deep learning models suitable for processing time series data such as LSTM or Transformer. Preferably, an LSTM model is used to build a network structure containing multiple LSTM units. The input layer receives pre-processed time series data, the hidden layer learns the time-dependent features of the data, and the output layer outputs the behavior trend prediction results (such as the change trend of health indicators in the future period, the probability of abnormal behavior).

[0058] Model training and optimization: Use the labeled historical data to train the model, use cross-entropy loss function or mean square error loss function to measure the difference between the predicted results and the actual results, update the model parameters through back propagation algorithm and stochastic gradient descent optimization algorithm, and continuously improve the prediction accuracy of the model.

[0059] Specifically, the implementation steps of edge computing are as follows:

[0060] Edge node deployment: Deploy edge computing devices in old people's living places (such as home gateways, local servers in nursing homes), and install lightweight detection models (such as fall detection models based on convolutional neural networks, heart rate anomaly detection models based on threshold judgment).

[0061] Real-time data analysis: The multi-modal data acquisition module transmits the collected data to the edge node first, and the edge node uses the locally deployed lightweight detection model to analyze the data in real time. For example, when the edge node receives the video frame data collected by the camera, it immediately uses the fall detection model to determine whether a fall event has occurred.

[0062] Abnormal event handling: If the edge node detects an abnormal event (such as falling, abnormal heart rate), it immediately triggers the local sound and light alarm device to remind the surrounding personnel, and uploads the key feature data of the abnormal event (such as the health indicator value at the abnormal time, the key frame of the behavior image) to the cloud server for further analysis and processing, reducing data transmission delay and improving emergency response efficiency.

[0063] The data processing and intelligent analysis module realizes collaborative analysis and safe sharing of multi-source data by constructing a data processing system combining federated learning and blockchain, avoids the privacy leakage risk caused by centralized storage of data in traditional mode, and ensures the credibility and traceability of data by using the non-tamperable characteristics of blockchain. The spatio-temporal sequence deep learning model can deeply mine the time and space characteristics of old people's health and behavior data, accurately predict behavior trends, and identify potential risks in advance. The edge computing technology pre-processes part of the data at the edge node, greatly shortens the response time of abnormal events, and reduces the network transmission pressure. The three work together to improve the efficiency and security of data processing, and enhance the real-time monitoring and early warning capability of the platform for the health and safety of the old people, providing solid data and technical support for smart elderly care services.

[0064] In this embodiment, the dynamic health record of the intelligent service module integrates physiological data, behavior data, living environment data, basic medical history and physical examination data, and generates a health trend report and personalized health advice through AI algorithms.

[0065] By comprehensively integrating multi-dimensional data to build a dynamic health record, the problem of single and outdated data in traditional health records has been changed. The deep analysis of massive data by AI algorithms can accurately capture the changing trend of old people's health indicators, identify potential health risks in advance, and generate personalized health advice based on individual circumstances. This enables the old people and their families to intuitively understand their health status, and medical staff can also develop more scientific health management plans based on detailed data, achieving dynamic tracking and precise intervention of the old people's health status, effectively improving the refinement and prevention capabilities of health management, reducing the risk of sudden illness, and improving the quality of life of the old people.

[0066] Specifically, the implementation steps are as follows:

[0067] Data integration and storage: The intelligent service module receives real-time physiological data (such as heart rate, blood pressure, blood sugar, etc.), behavior data (daily activity trajectory, exercise steps, etc.), and living environment data (temperature and humidity, air quality, etc.) transmitted by the multi-modal data acquisition module in real time, and retrieves stored basic medical history and physical examination data. These data are integrated according to a unified data format and stored in the dynamic health record database. For example, a dedicated data table is established for each old person, and different types of data are recorded in chronological order.

[0068] AI algorithm model construction: an integrated learning algorithm (such as random forest, gradient boosting tree) and deep learning algorithm (such as LSTM) are combined to construct a health analysis model. The integrated learning algorithm is used to process structured health indicator data and mine the correlation between different indicators; the LSTM algorithm is used to analyze time series data and learn the changing rules of health indicators over time. At the same time, natural language processing technology is used to extract and analyze the text information in the basic medical history, and convert it into structured data for integration into the model.

[0069] Health trend analysis: AI algorithms regularly analyze data in dynamic health records. For physiological data, the mean, standard deviation, and other statistical quantities of each indicator are calculated, and trend curves are drawn. By comparing historical data and preset thresholds, it is determined whether the indicators have abnormal fluctuations. For behavior data and living environment data, the potential relationship between them and health indicators is analyzed, such as determining whether long-term exposure to poor living conditions affects health. Finally, a visual health trend report is generated, showing the changing trend of the elderly's health status in the form of charts and text.

[0070] Personalized health advice generation: Based on the results of health trend analysis, combined with the individual information of the elderly, such as age, gender, and medical history, the system selects matching advice content from the preset health advice knowledge base. If the elderly's blood pressure continues to rise, the system extracts dietary adjustment suggestions (such as reducing salt intake), exercise suggestions (such as increasing walking time), and medical advice (such as regular check-ups) from the knowledge base that are suitable for the elderly's physical condition, and generates a personalized health advice list to push to the elderly, family members, and medical staff terminals.

[0071] In this embodiment, the intelligent service module, in the disease early warning and intervention, links the intelligent medicine box and the voice assistant to remind the elderly to take medicine and adjust the diet, and arranges on-site medical services or remote consultations according to the situation.

[0072] By constructing a disease early warning and multi-device linkage intervention system, timely response and active management of the elderly's disease risk are achieved. When the system detects health abnormalities, it immediately links the intelligent medicine box and the voice assistant to remind the elderly to take the correct measures in an intuitive and convenient way, effectively avoiding the aggravation of the disease due to forgetfulness or negligence; according to the severity of the disease, on-site medical services or remote consultations are arranged flexibly to provide graded and precise medical support for the elderly. This whole-process disease management mode not only improves the elderly's compliance behavior, but also controls the development of the disease in the early stage, reduces medical costs, and at the same time reduces the care pressure of family members, improves the professionalism and effectiveness of overall pension services.

[0073] Specifically, the implementation steps are as follows:

[0074] Disease early warning monitoring: The data processing and intelligent analysis module transmits the analyzed health data to the intelligent service module in real time. The disease early warning system in the intelligent service module evaluates the data based on the preset disease risk model (based on medical guidelines and big data analysis). For example, when the old person's blood glucose measurement value exceeds the diabetes warning threshold for multiple times in a row, the system determines that there is a risk of diabetes and triggers the disease early warning mechanism.

[0075] Intelligent reminder linkage: Once the disease early warning is triggered, the intelligent service module immediately sends instructions to the intelligent medicine box, which reminds the old person to take the relevant medicine on time through flashing lights, vibration, and voice prompts. At the same time, the voice assistant automatically starts to explain the current health status, matters needing attention, and dietary adjustment suggestions (such as "your blood sugar is high, eat less sweets today, and eat more vegetables") to the old person in a friendly voice. In addition, the system also pushes warning information to the family member's mobile phone APP, informing the old person's health status and the reminder measures taken.

[0076] Medical service grading arrangement: The intelligent service module processes the disease early warning according to the severity. For mild risk (such as mild elevation of blood pressure), the system prioritizes remote consultation services, automatically books an online doctor for the old person, generates a consultation sheet containing health data and symptom description, and reminds the old person to participate in remote consultation on time. If it is a moderate or severe risk (such as sudden chest pain, confusion), the system immediately contacts the cooperating medical institutions to arrange on-site medical services, and sends the old person's detailed health information, residence address, etc. to the medical staff to ensure timely and accurate rescue. After the on-site service is completed, the system records the medical process and follow-up medical advice, and updates the dynamic health record.

[0077] In this embodiment, the reinforcement learning algorithm of the intelligent service module adjusts the service recommendation priority according to the old person's evaluation of the nursing service, and recommends alternative service schemes based on the old person's preferences and health status.

[0078] By introducing the reinforcement learning algorithm, the platform is given the ability to optimize the service recommendation strategy, changing the rigidity of the traditional service recommendation mode. The algorithm dynamically adjusts the recommendation priority according to the old person's real feedback on the nursing service (such as satisfaction score, service usage frequency), and simultaneously considers the old person's individual preferences and health status, accurately matching alternative service schemes. This enables the platform to quickly adapt to changes in the old person's needs, providing high-quality services that are more individualized and intelligent, improving the old person's satisfaction and acceptance of the service, avoiding resource waste, promoting efficient use of elderly care services, enhancing the platform's user stickiness and competitiveness, and promoting the development of intelligent elderly care services in a more personalized and intelligent direction.

[0079] Specifically, the implementation steps are as follows:

[0080] Service evaluation data collection: After each nursing service is completed, the intelligent service module invites the elderly or their family members to evaluate the service through mobile phone APP, voice interaction, etc. The evaluation content includes service quality (such as whether the nursing method is professional), service attitude (such as whether the nursing staff is patient), service timeliness, etc. Star rating (1-5 stars) and text comments are combined to provide feedback. At the same time, the system automatically records the service usage frequency, service duration, etc. as auxiliary evaluation indicators of service effect.

[0081] Reinforcement learning model construction: A reinforcement learning model based on Q-learning algorithm is constructed, and the service recommendation process is defined as a series of state-action-reward decision-making processes. The state includes the health status of the elderly (such as the type of chronic disease, the rehabilitation stage), service history records (types of services received and evaluation), preference information (such as dietary preferences, nursing time preferences), etc.; the action refers to various nursing services (such as daily care, rehabilitation care, psychological care) that the platform can recommend; the reward is set according to the evaluation results of the elderly, such as high reward value for 5-star evaluation and low reward value for 1-star evaluation.

[0082] Service recommendation strategy optimization: The reinforcement learning model selects a service recommendation action (i.e. recommends a certain type of nursing service) according to the current state. After the elderly evaluate the service, the model updates the Q-value table (records the expected reward of each state-action pair) according to the reward value. Through continuous iterative learning, the optimal service recommendation strategy is gradually found. For example, if the elderly give high evaluation to rehabilitation care services many times, the model will increase the recommendation priority of rehabilitation care services; if the evaluation of a psychological care service is low, the model will reduce its recommendation weight, and according to the health status and preferences of the elderly, it will select appropriate alternative services (such as music therapy, art therapy, etc.) from the service plan library for recommendation.

[0083] Personalized service plan generation: In each service recommendation, the model will not only consider the recommendation strategy obtained by reinforcement learning, but also combine the current health data of the elderly (such as the latest medical report, disease changes) and personalized preference settings to customize and adjust the recommended service plan. For example, when recommending nursing services for an elderly with heart disease, the service items related to heart rehabilitation are preferentially selected, and according to the elderly's preference for quietness, the service is arranged at an appropriate time period to generate the final personalized service recommendation list and push it to the elderly.

[0084] In this embodiment, the virtual assistant of the emotional social module, based on natural language processing and emotional computing technology, analyzes the emotional state by analyzing the voice tone and chat content of the elderly, and initiates a chat, plays music or pushes information.

[0085] By fusing natural language processing and sentiment computing technologies, the virtual assistant is endowed with the ability to perceive the emotions of the elderly, changing the status quo of the lack of emotional interaction in traditional elderly care services. The virtual assistant can real-time perceive the emotional changes of the elderly and actively initiate interactions that match the emotional state of the elderly. It can chat with the elderly when they are lonely, play soothing music when they are in a low mood, and push information that matches their interests. This not only effectively alleviates the loneliness of the elderly, but also improves their mental health through positive emotional guidance, providing all-weather and personalized emotional care for the elderly, making smart elderly care services more humane, and filling the gap in emotional companionship in traditional elderly care services.

[0086] Specifically, the implementation steps are as follows:

[0087] Voice and text data collection: The virtual assistant collects the voice data of the elderly in real-time through the microphone of the intelligent terminal (such as smart speaker, mobile phone APP), and converts the voice into text. At the same time, it records the text content input by the elderly during the chat process, providing a data basis for subsequent emotional analysis. For example, when the elderly communicate with the virtual assistant through the smart speaker, the device converts the voice signal into digital audio, and then uses voice recognition technology (such as an end-to-end speech recognition model based on deep learning) to convert it into processable text information.

[0088] Emotion analysis model construction: A deep learning model that fuses multi-modal information (such as a dual-channel neural network that combines voice features and text semantics) is used for emotion analysis. For voice data, acoustic features such as pitch, speech rate, and intonation fluctuation are extracted; for text data, natural language processing techniques are used for word segmentation, part-of-speech tagging, and sentiment word extraction, and then a word vector model (such as Word2Vec, BERT) is used to convert the text into a semantic vector. The two types of features are input into the neural network, and the model is trained to identify different emotional states (such as happy, sad, lonely, anxious, etc.).

[0089] Active interaction strategy formulation: Based on the results of emotional analysis, the virtual assistant selects the appropriate interaction method from the pre-set interaction strategy library. If it judges that the elderly are in a low mood, it will prefer to play healing music (such as classical piano music), and at the same time ask the elderly if they want to talk in a gentle tone; if it detects that the elderly are in a state of loneliness, it will actively initiate a light topic (such as recalling past funny stories or discussing the recent weather). In addition, the virtual assistant will also push personalized information content based on the historical chat records and interest preferences of the elderly, such as classic opera performance videos for opera lovers and health and wellness articles for health enthusiasts.

[0090] Interaction effect feedback and optimization: After each interaction, the virtual assistant obtains feedback through simple questions (such as "Do you like the conversation just now?") or collects the subsequent interaction behavior of the elderly (such as whether to continue the conversation or whether to choose the pushed information). Feedback information is used to optimize the emotion analysis model and interaction strategy, for example, if there is a mood misjudgment for several times, adjust the model parameters; if a certain interaction method has good feedback, increase the priority of using this strategy in similar emotional scenarios.

[0091] In this embodiment, the intergenerational socialization and community interaction platform of the emotional social module introduces game mechanism, including family task check-in and community activity point reward.

[0092] By integrating game elements into intergenerational socialization and community interaction, the limitations of single form and low participation of traditional social platforms are broken. The family task check-in function promotes frequent interaction and emotional communication between the elderly and their children, enhancing family cohesion; the community activity point reward mechanism stimulates the enthusiasm of the elderly to participate in community activities, expands the social circle, and enhances the sense of social belonging of the elderly. This innovative social mode makes the elderly obtain a sense of achievement and pleasure in interesting interaction, effectively improves the quality of social life of the elderly, alleviates psychological problems caused by social scarcity, and also creates a more active and warm atmosphere for the elderly community, promoting the formation of a positive and healthy social ecology for the elderly.

[0093] Specifically, the implementation steps are as follows:

[0094] Family task check-in system construction: Set up a family task module in the intergenerational socialization and community interaction platform, and the platform automatically generates diversified family tasks every week, such as "share interesting things with children through video call this week", "complete online jigsaw puzzle game with family", "send handwritten greeting cards to each other", etc. The task difficulty is set at three levels (easy, medium, and difficult), and the estimated completion time and point reward amount are marked.

[0095] Task execution and record: The elderly and their children receive tasks through their respective platform accounts, and after completing the tasks, they submit proof through uploading photos, videos, and text records, etc. For example, after completing the video call task, upload the call screenshot; after completing the puzzle game, submit the game completion interface screenshot. The platform automatically identifies the submitted content, records the task completion after passing the audit, and accumulates the corresponding points for the participants.

[0096] Community activity point system design: The community administrator publishes various online and offline activities on the platform, such as "elderly calligraphy exhibition", "health knowledge competition", "community volunteer patrol" and the like. The old people can obtain point rewards by participating in the activities and meeting the corresponding requirements (such as submitting works, correctly answering a certain number of questions, completing the patrol time). Points can be used to exchange for physical prizes (such as daily necessities, health products for the elderly), virtual rights (such as platform membership privileges, priority to participate in activities).

[0097] Social interaction incentive and atmosphere creation: The platform sets up a ranking list function to show the ranking of family task completion points and community activity points, stimulating the participation enthusiasm of the old people and families. At the same time, participants are encouraged to share photos and experiences of task completion on the platform to form an interactive discussion area and enhance the social atmosphere. In addition, special activities for point exchange and social gatherings are held regularly to further promote communication among the old people and families and consolidate social relationships.

[0098] The embodiment of the application can acquire and process related data based on artificial intelligence technology. Artificial intelligence (AI) is the use of digital computers or computer-controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0099] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0100] Those skilled in the art can understand that all or part of the processes in the above embodiments can be completed by instructing related hardware through computer readable instructions, which can be stored in a computer readable storage medium. The program can include the processes of the above embodiments when executed, wherein the storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0101] The above is only a preferred embodiment of the present application and does not limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A multifunctional smart elderly care service platform, characterized by: include: Multimodal data collection module, which collects the elderly's physiological data, behavioral data, and living environment data; The physiological sign data is collected through wearable devices and contactless devices, the behavioral data is collected through cameras and sensors, and the living environment data is collected through environmental sensing devices; A data processing and intelligent analysis module is connected to the multimodal data acquisition module to integrate and process the collected data, analyze the data, and predict behavioral trends; An intelligent service module connects the data processing and intelligent analysis modules to provide the elderly with intelligent health management services and personalized proactive services based on the analysis results; The intelligent health management service includes the establishment of dynamic health records, disease warning and intervention; the personalized proactive service uses reinforcement learning algorithms to adjust recommendation strategies based on service feedback, and realizes cross-scenario linkage of smart home, health monitoring and life services; The emotional social module provides a virtual assistant with emotion recognition capabilities, and builds an intergenerational social and community interaction platform.

2. A multifunctional smart elderly care service platform according to claim 1, characterized in that: In the multimodal data acquisition module, the camera uses differential privacy technology to desensitize the video data and only extracts skeletal key point information for fall detection and abnormal behavior warning.

3. A multifunctional smart elderly care service platform according to claim 1, characterized in that: The data processing and intelligent analysis module uses a federated learning framework and blockchain technology. Each data holder trains the model locally and uploads the parameter update results to the cloud. Blockchain is also used to encrypt, store and trace the data. A spatiotemporal sequence deep learning model is used to analyze data and predict behavioral trends. Use edge computing to deploy lightweight detection models on edge nodes to analyze data in real time.

4. A multifunctional smart elderly care service platform according to claim 3, characterized in that: The spatiotemporal sequence deep learning model adopts an LSTM or Transformer model and combines the elderly person's historical waking time and daily activity data to determine whether the current behavior is abnormal.

5. A multifunctional smart elderly care service platform according to claim 4, characterized in that: When the edge computing detects an abnormal event, it immediately triggers a local sound and light alarm and uploads key feature data to the cloud.

6. A multifunctional smart elderly care service platform according to claim 1, characterized in that: The dynamic health profile of the intelligent service module integrates physiological sign data, behavioral data, living environment data, basic medical history and physical examination data, and generates health trend reports and personalized health recommendations through AI algorithms.

7. The multifunctional smart elderly care service platform according to claim 1, characterized in that: During disease warning and intervention, the intelligent service module links the smart medicine box and voice assistant to remind the elderly to take medicine and adjust their diet, and arranges home medical services or remote consultations according to the situation.

8. The multifunctional smart elderly care service platform according to claim 1, characterized in that: The reinforcement learning algorithm of the intelligent service module adjusts the service recommendation priority according to the elderly's evaluation of nursing services, and recommends alternative service plans based on the elderly's preferences and health status.

9. The multifunctional smart elderly care service platform according to claim 1, characterized in that: The virtual assistant of the emotional social module is based on natural language processing and emotional computing technology. It judges the emotional state of the elderly by analyzing their voice intonation and chat content, and actively initiates chats, plays music or pushes information.

10. The multifunctional smart elderly care service platform according to claim 1, characterized in that: The intergenerational social and community interaction platform of the emotional social module introduces a gamification mechanism, including family task check-ins and community activity point rewards.

Images