Intelligent reminding method and system for assisting old people and related equipment

Abstract

The invention is suitable for the technical field of intelligent reminding, and provides an intelligent reminding method and system for assisting the elderly and related equipment, and the method comprises the steps: collecting the health state data and life behavior data information of the elderly in multiple dimensions based on wearable equipment and intelligent home equipment; after data preprocessing is conducted on the data information, heart rate variability and life rhythm characteristics, body activity characteristics, cognition and interaction characteristics and life habit characteristics of the old people are extracted; carrying out feature fusion on the heart rate variability and life rhythm features, the body activity features, the cognition and interaction features and the life habit features of the old people to obtain a target fusion feature vector; inputting the target fusion feature vector into a trained global Actor network, and outputting probability distribution of each possible reminding behavior by the trained global Actor network; and the optimal reminding behavior is determined as the target intelligent reminding decision according to the probability distribution, so that the individuation and accuracy of intelligent reminding are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent reminder technology, and in particular to an intelligent reminder method, system and related equipment for assisting the elderly. Background Art

[0002] At present, in view of the memory loss and lack of concentration of the elderly, products on the market can remind the elderly to take medicine on time, carry out daily activities, participate in social activities and complete other important tasks in a variety of ways. It can improve the regularity and quality of life of the elderly and ensure the health and safety of the elderly. In the existing technology, there are great individual differences between different elderly people, including physical condition, living habits, cognitive ability, hearing and vision. The system needs to constantly learn the behavior patterns and changes in needs of the elderly, but the behavior and needs of the elderly may change dynamically due to various factors, which makes it impossible to update the reminder strategy in a timely and accurate manner, which affects the elderly's acceptance and user experience of the system to a certain extent.

[0003] In view of this, there is a need for an intelligent reminder method, system and related equipment for assisting the elderly. Summary of the invention

[0004] The embodiments of the present application provide a smart reminder method, system and related equipment for assisting the elderly, which are used to solve the problem that the smart reminder update strategy is not timely and accurate enough.

[0005] The first aspect of the embodiment of the present application provides an intelligent reminder method for assisting the elderly, comprising:

[0006] Collect health status data and life behavior data of the elderly in multiple dimensions based on wearable devices and smart home devices;

[0007] After preprocessing the collected data information, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted;

[0008] Fusing the heart rate variability of the elderly with life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fused feature vector;

[0009] Inputting the target fused feature vector into a trained global Actor network, the trained global Actor network outputs a probability distribution of each possible reminder behavior;

[0010] An optimal reminder behavior is determined according to the probability distribution as a target intelligent reminder decision.

[0011] Furthermore, after the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including:

[0012] Heart rate variability and life rhythm characteristics:

[0013]

[0014] Where: M is the number of windows in a day, SDNN w is the SDNN standard deviation of the time interval between two adjacent heartbeats in the wth acquisition window, SDNN avg is the SDNN average of all windows in one day, σ SDNN is the change of SDNN between different sampling windows; RMSSD w is the root mean square of the difference between the time intervals of two adjacent heartbeats in the wth acquisition window, RMSSD avg is the average RMSSD of all windows in one day, σ RMSSD is the fluctuation of RMSSD in different periods; LF w and HF w are the sum of the power of the low-frequency component and the high-frequency component in the w-th acquisition window, LFHF w To measure the balance between low-frequency components and high-frequency components, LF avg HF avg and LFHF avg They are the average value of low-frequency components, the average value of high-frequency components, and the average value of the ratio of low-frequency components to high-frequency components;

[0015]

[0016] Among them: SDT d SDT is the total time that the elderly person is in deep sleep on the dth day out of n days. avg and σ SDT NW is the mean and standard deviation of deep sleep time of the elderly; d NW is the number of times the elderly wake up from sleep on day d. avg and σ NW is the mean and standard deviation of the number of awakenings at night; N is the total sleep cycle, SC D is the length of the D sleep cycle of the elderly, SC avg and σ SC are the mean duration and standard deviation of sleep cycles, respectively.

[0017] Furthermore, after the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including:

[0018] Physical activity characteristics:

[0019]

[0020]

[0021] Among them: SS i is the step length of the i-th step, SL i is the pace of the ith step, T i is the time period of the i-th step, N′ is the total number of steps, SL avg and σ SL are the mean and standard deviation of the step length, SS avg and σ SS are the mean and standard deviation of the pace, GC i is the gait cycle of the i-th step, GC avg and σ GC are the mean and standard deviation of the gait cycle, The difference in time between the left and right feet landing. and are the mean and standard deviation of the difference in landing time between the left and right feet, respectively.

[0022] Furthermore, after the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including:

[0023] Cognitive and interactive features:

[0024]

[0025] Among them: OPE count is the total number of operational errors within T′, OPE avg is the average number of operational errors per day during T′, FMR count The number of times the elderly forget to check or perform the set reminder tasks, FMR rate To remind the task forgetting rate, NT total is the total number of reminder tasks set within T′, AD ratio is the ratio of the number of different applications used by the elderly within T′ to the total number of applications on the device, X is the total number of applications installed on the device, and x is the number of applications suitable for the elderly within T′.

[0026] Furthermore, after the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including:

[0027] Lifestyle characteristics:

[0028]

[0029] in: is the breakfast time on the dth day within n days, and are the mean and standard deviation of breakfast time, is the lunch time on the dth day within n days, and are the mean and standard deviation of lunch time, is the dinner time on the dth day within n days, and are the mean and standard deviation of dinner time, respectively; and The food intake for breakfast, lunch and dinner are respectively is the total daily food intake, and are the mean and standard deviation of total food intake within m days; is the time when the elderly take the jth drug on day d, and are the mean and standard deviation of the duration of taking the jth drug, is the actual dosage of the jth drug, The dose prescribed for the jth drug, DDR j,d is the dosage deviation rate of the jth drug.

[0030] Furthermore, the heart rate variability of the elderly is fused with life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fused feature vector, including:

[0031] Use the time series algorithm to select the sliding window size and obtain the feature vector in each window;

[0032] In each window, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics and life habit characteristics of the elderly are spliced ​​into a comprehensive feature vector in a preset order;

[0033] The concatenated comprehensive feature vectors are weighted fused to obtain the target fused feature vector.

[0034] Furthermore, the target fusion feature vector is input into a trained global Actor network, and the trained global Actor network outputs a probability distribution of each possible reminder behavior, including:

[0035] Build an environmental model that simulates the elderly's life scenes and intelligent reminder interactions;

[0036] Initialize hyperparameters, including learning rate, discount factor, and exploration rate, and construct the network structures of the Actor network and the Critic network;

[0037] Create multiple working threads and conduct intra-thread interaction and learning;

[0038] After each worker thread independently performs a preset number of interactions and parameter updates, it synchronizes the updated parameters of the Actor and Critic networks to the global network parameters;

[0039] Continuously monitor the training process and set stop conditions to determine when the training is finished.

[0040] Furthermore, the creation of multiple working threads and the interaction and learning within the threads include:

[0041] Obtain the target fusion feature vector at the current moment in each thread;

[0042] Input the comprehensive feature vector of the current state into the Actor network, and obtain the probability distribution corresponding to each possible intelligent reminder behavior through the network's forward propagation;

[0043] Execute the selected intelligent reminder action in the environment. The environment updates the state of the elderly according to the preset rules, and feeds back a reward signal and transfers to the next state.

[0044] Use the Critic network to evaluate the value of the current state and the next state;

[0045] The parameters of the Actor network and the Critic network are updated using the gradient ascent and gradient descent algorithms based on the calculated advantage estimates and the probability of selecting the action.

[0046] A second aspect of the embodiment of the present application provides an intelligent reminder system for assisting the elderly, including:

[0047] A data information collection unit, used to collect health status data and life behavior data information of the elderly in multiple dimensions based on wearable devices and smart home devices;

[0048] The feature extraction unit for the elderly is used to extract the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly after data preprocessing the collected data information;

[0049] A target fusion feature vector determination unit is used to fuse the heart rate variability of the elderly with the life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fusion feature vector;

[0050] A global Actor network learning and training unit, used for inputting the target fusion feature vector into a trained global Actor network, wherein the trained global Actor network outputs a probability distribution of each possible reminder behavior;

[0051] The target intelligent reminder decision determining unit is used to determine the optimal reminder behavior as the target intelligent reminder decision according to the probability distribution.

[0052] A third aspect of an embodiment of the present application provides a computer device, including:

[0053] memories, transceivers, processors, and bus systems;

[0054] Wherein, the memory is used to store programs;

[0055] The processor is used to execute the program in the memory, including executing any one of the above-mentioned intelligent reminder methods for assisting the elderly;

[0056] The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other.

[0057] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0058] After collecting the health status data and life behavior data information of the elderly, the present invention extracts the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics and life habit characteristics of the elderly, and extracts the characteristics of the elderly from multiple dimensions, which can reflect the health status and life status of the elderly in many aspects, and provide an accurate data basis for making intelligent reminder decisions; the fused feature vector is input into the trained global Actor network, and the complex mapping relationship between the elderly status and the reminder behavior learned by the network is used to output the probability distribution of each possible reminder behavior, and finally the optimal reminder behavior is determined as the target intelligent reminder decision according to the probability distribution, realizing personalized and precise intelligent reminder services, which can provide intelligent reminders for the elderly in a timely and accurate manner, and effectively improve the elderly's user experience.

[0059] Other advantages, objectives, and features of the present invention will be set forth in part in the following description, and in part will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The present invention is a flowchart of an embodiment of an intelligent reminder method for assisting the elderly. DETAILED DESCRIPTION

[0061] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0062] Embodiment 1

[0063] The implementation method in this embodiment can be implemented in the system, can be implemented in the server, and can also be implemented in the terminal, and the specific details are not clearly limited. The following will introduce the intelligent reminder method used to assist the elderly in this application from the perspective of system implementation. Figure 1 , the method provided in the embodiment of the present application comprises the following steps:

[0064] S11. Collect health status data and life behavior data information of the elderly in multiple dimensions based on wearable devices and smart home devices;

[0065] In this embodiment, the wearable device includes a smart bracelet or a smart watch to obtain the elderly's heart rate data, sleep data and activity data; it can also include a smart clothing chest strap with a sensor or a smart insole shoe with a sensor to obtain the elderly's electrocardiogram data and gait data respectively.

[0066] Smart home devices include smart refrigerators, smart medicine boxes, smart cameras, smart voice assistant devices or mobile phones. Smart refrigerators obtain dietary information through built-in weight sensors and image recognition systems; smart medicine boxes are equipped with sensors to record the name of the medicine and the time it is placed in the medicine box, and the time and dosage of the medicine when the elderly take it out; smart cameras are installed in each room in the elderly’s living environment to record the elderly’s activity videos at a certain frame rate, and use video analysis technology to obtain the elderly’s activity time, activity trajectory and some simple behavioral actions in each room; smart voice assistant devices record the voice interaction content between the elderly and the voice assistant, including questions, instructions, conversations, etc., as well as the time, frequency and success rate of the interaction; mobile phones record the elderly’s operating behaviors through system logs, such as the time, location and type of operations such as opening, closing, clicking, sliding and entering applications.

[0067] S12. After preprocessing the collected data information, extract the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly;

[0068] 1. Heart rate variability and life rhythm feature extraction:

[0069] For heart rate data, filtering is performed to remove baseline drift and noise interference, and methods such as Butterworth filter can be used. For example, a suitable cutoff frequency can be set to remove high-frequency noise and low-frequency drift, making the heart rate data smoother and more stable. Sleep data is cleaned to correct sleep stage records that are obviously illogical and to supplement possible missing data segments.

[0070]

[0071] Where: M is the number of windows in a day, SDNN w is the SDNN standard deviation of the time interval between two adjacent heartbeats in the wth acquisition window, SDNN avg is the SDNN average of all windows in one day, σ SDNN is the change of SDNN between different sampling windows; RMSSD w is the root mean square of the difference between the time intervals of two adjacent heartbeats in the wth acquisition window, RMSSD avg is the average RMSSD of all windows in one day, σ RMSSD is the fluctuation of RMSSD in different periods; LF w and HF w are the sum of the power of the low-frequency component and the high-frequency component in the w-th acquisition window, LFHF w To measure the balance between low-frequency components and high-frequency components, LF avg HFavg and LFHF avg They are the average value of the low-frequency component, the average value of the high-frequency component, and the average value of the ratio of the low-frequency component to the high-frequency component.

[0072]

[0073] Among them: SDT d SDT is the total time that the elderly person is in deep sleep on the dth day out of n days. avg and σ SDT NW is the mean and standard deviation of deep sleep time of the elderly; d NW is the number of times the elderly wake up from sleep on day d. avg and σ NW is the mean and standard deviation of the number of awakenings at night; N is the total sleep cycle, SC D is the length of the D sleep cycle of the elderly, SC avg and σ SC are the mean duration and standard deviation of sleep cycles, respectively.

[0074] 2. Physical activity feature extraction:

[0075] Using the accelerometer and gyroscope in the wearable device, the body movement data of the elderly is collected within one week with a sampling interval of 1 minute to obtain the acceleration values ​​and angular velocity values ​​of the three axes. The gravity acceleration component in the acceleration data is removed, and a filtering algorithm based on the acceleration amplitude and direction is used to separate the gravity component so that the data can more accurately reflect the actual movement of the body. The acceleration and angular velocity data are normalized and the data values ​​are mapped to a reasonable range to facilitate subsequent feature calculation and analysis.

[0076] According to the preset acceleration amplitude threshold, the exercise data per minute is divided into sedentary, that is, the acceleration amplitude is less than 0.05g, light activity, that is, the acceleration amplitude is 0.05g-0.2g, moderate activity, that is, the acceleration amplitude is 0.2g-0.5g, and heavy activity, that is, the acceleration amplitude is greater than 0.5g. The time SED of the elderly in each activity intensity is counted every day d , LAT d , MAT d , HAT d , in minutes.

[0077] Calculate the percentage of sedentary time per day for a week Light activity time Moderate activity time Proportion of time spent on heavy activities Then calculate the mean and corresponding standard deviation of these proportions within a week.

[0078] Physical activity characteristics are calculated as follows:

[0079]

[0080] Among them: SS i is the step length of the i-th step, SL i is the pace of the ith step, T i is the time period of the i-th step, N′ is the total number of steps, SL avg and σ SL are the mean and standard deviation of the step length, SS avg and σ SS are the mean and standard deviation of the pace, GC i is the gait cycle of the i-th step, GC avg and σ GC are the mean and standard deviation of the gait cycle, The difference in time between the left and right feet landing. and are the mean and standard deviation of the difference in landing time between the left and right feet, respectively.

[0081] 3. Cognitive and interactive feature extraction:

[0082] The interaction data between the elderly and the device within one month is recorded through smart devices, including operation records, setting and execution of reminder tasks, and application usage. The interaction data is sorted and classified to remove irrelevant redundant information, such as device automatic update operation records, system notification click records, and other data that do not reflect the cognitive behavior of the elderly. The time data in the operation records is standardized and unified into the same time format to facilitate subsequent statistical analysis.

[0083] Operation error statistics: Analyze the operation errors of the elderly when using smart devices, such as repeatedly clicking the same button, getting lost in the operation process (such as returning to the previous menu multiple times or entering the wrong function interface, entering wrong information, etc. Count the total number of operation errors within T′ time OPE count , and calculate the average number of operational errors per day within T′:

[0084] Statistics of the number of times the reminder task is forgotten: record the number of times the elderly forget to check or execute the set reminder task, recorded as FMR count , calculate the forgetting rate of reminder tasks NT total The total number of reminder tasks set within T′.

[0085] Application usage diversity analysis: Calculate the ratio of the number of different applications used by the elderly within T′ to the total number of applications on the device: Among them, X is the total number of applications installed on the device, and x is the number of applications that the elderly use within T′. This ratio reflects the degree of exploration and utilization of the functions of smart devices by the elderly, and to a certain extent reflects their cognitive ability and acceptance of new things.

[0086] 4. Extraction of living habits features:

[0087] Food-related data is obtained through smart refrigerators. Specifically, smart refrigerators are equipped with image recognition technology to identify the types of food put in and taken out; at the same time, weight sensors are used to record changes in food weight to infer the amount of food. These data include the time the food was put into the refrigerator, the time it was taken out, the type and approximate quantity. For dining out, the data can be supplemented by the mobile payment records used by the elderly or the elderly’s own dietary information, including dining time, restaurant type and approximate food types. Smart medicine boxes record drug-related information. When the medicine is put into the medicine box, the name of the medicine and the time it is put in are recorded; when the elderly take out the medicine, the time of removal and the dosage are recorded. The sensors in the medicine box can accurately sense these operations and send the data to the storage device.

[0088] Check whether there are any anomalies in the data recorded by the smart refrigerator, and sort all dietary data by date and meal time. Remove false data caused by misoperation in the smart medicine box records, check the accuracy of the drug name and dosage, ensure that the drug name recorded in the medicine box is consistent with the actual drug put in, and the dosage record meets the conventional specifications of the drug. Sort the medication data by date and drug type to facilitate subsequent analysis of the time and dosage pattern of each drug.

[0089] The life habit characteristics are calculated as follows:

[0090]

[0091]

[0092] in: is the breakfast time on the dth day within n days, and are the mean and standard deviation of breakfast time, is the lunch time on the dth day within n days, and are the mean and standard deviation of lunch time, is the dinner time on the dth day within n days, and are the mean and standard deviation of dinner time, respectively; and The food intake for breakfast, lunch and dinner are respectively is the total daily food intake, and are the mean and standard deviation of total food intake within m days; is the time when the elderly take the jth drug on day d, and are the mean and standard deviation of the duration of taking the jth drug, is the actual dosage of the jth drug, The dose prescribed for the jth drug, DDR j,d is the dosage deviation rate of the jth drug.

[0093] The above steps extract heart rate variability and life rhythm characteristics, which can provide in-depth insights into the health status of the elderly's cardiovascular system and key health dimensions such as sleep quality. Heart rate variability-related indicators can reflect the cardiac autonomic nervous regulation function, while life rhythm characteristics are related to the recovery of the elderly's nervous system and the regulation of the overall physical state. These characteristics help to discover potential cardiovascular disease risks and sleep disorders in advance. The extraction of physical activity characteristics can accurately evaluate the elderly's physical movement ability and balance ability. For example, by analyzing the changes in parameters such as step length, pace, and gait cycle, timely detect the signs of decline in the elderly's physical function, which is of great significance for preventing accidents such as falls, and also provides a basis for formulating personalized rehabilitation training or activity recommendations. Cognitive and interactive features can mine the cognitive ability of the elderly from their interactive behaviors with smart devices, which helps to detect the decline of cognitive function early, so as to intervene in time and provide corresponding cognitive training support, and improve the elderly's ability to take care of themselves and their quality of life. The extraction of living habit characteristics can clearly present the daily life rules and self-health management of the elderly, discover bad living habits such as unbalanced diet and failure to take medicine on time, and then provide targeted references for adjusting lifestyles and optimizing health management strategies.

[0094] S13. Fusing the heart rate variability of the elderly with the life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fusion feature vector;

[0095] In this embodiment, step S13 further includes the following:

[0096] 1. Use the time series algorithm to select the sliding window size and obtain the feature vector in each window;

[0097] 2. In each window, the heart rate variability of the elderly is combined with the characteristics of life rhythm, physical activity, cognition and interaction, and life habits in a preset order to form a comprehensive feature vector;

[0098] Using sliding window technology, select an appropriate window size, such as a 7-day window, and slide the window on the time series to obtain the feature vector in each window. In each window, the four types of features are spliced ​​into a comprehensive feature vector in a certain order.

[0099] The characteristic vector of life rhythm and heart rate variability is:

[0100] LRHV=[SDNN std ,RMSSD std , LF std , HF std , LFHF std , STD std , N.W. std , S.C. std ];

[0101] The physical activity feature vector is:

[0102]

[0103] The cognitive and interactive feature vectors are:

[0104]

[0105] The living habit feature vector is:

[0106]

[0107] The concatenated comprehensive feature vector F = [LRHV, PA, CI, LH].

[0108] 3. Perform weighted fusion on the concatenated comprehensive feature vector to obtain the target fused feature variable.

[0109] The concatenated feature vectors are weighted and fused, and each feature is assigned a weight according to its importance to the elderly status assessment. The fused target feature vector F fusion :

[0110] F fusion =W LRHV ·LRHV+W PA PA+W CI ·CI+W LH ·LH

[0111] Where: W LRHV , W PA , WCI and W LH They are the weights corresponding to the life rhythm and heart rate variability feature vector, physical activity feature vector, cognition and interaction feature vector and life habit feature vector.

[0112] S14. Input the target fusion feature vector into the trained global Actor network, and the trained global Actor network outputs the probability distribution of each possible reminder behavior;

[0113] In this embodiment, step S14 includes the following:

[0114] 1. Build an environmental model that simulates the elderly’s life scenes and intelligent reminder interactions;

[0115] 2. Initialize hyperparameters, including learning rate, discount factor, and exploration rate, and build the network structure of the Actor network and the Critic network;

[0116] 3. Create multiple working threads and interact and learn within the threads;

[0117] Specifically, the target fusion feature vector of the current moment is obtained in each thread; for example, assuming that the comprehensive feature vector includes the result of integrating multiple feature data such as the elderly's heart rate variability, activity intensity, and dietary habits, the vector represents the overall state of the elderly at this moment.

[0118] The comprehensive feature vector of the current state is input into the Actor network, and the probability distribution corresponding to each possible intelligent reminder behavior is obtained through the forward propagation of the network; then according to this probability distribution, a specific intelligent reminder behavior is selected as the current action according to a certain strategy.

[0119] Execute the selected intelligent reminder action in the environment. The environment updates the state of the elderly according to the preset rules, and feeds back a reward signal and transfers to the next state.

[0120] The Critic network is used to evaluate the value of the current state and the next state; the comprehensive feature vector of the current state is input into the Critic network to obtain the current state value estimate, and the comprehensive feature vector of the next state is input into the Critic network to obtain the next state value estimate.

[0121] The parameters of the Actor network and the Critic network are updated using the gradient ascent and gradient descent algorithms based on the calculated advantage estimates and the probability of selecting the action.

[0122] 4. After each worker thread independently performs a preset number of interactions and parameter updates, it synchronizes the updated parameters of the Actor and Critic networks to the global network parameters;

[0123] 5. Continuously monitor the training process and set stop conditions to determine when the training ends. The stop conditions here include reaching the maximum number of training rounds, the loss function values ​​of the Actor and Critic networks converge to below a certain threshold, or the average reward on the validation set reaches the expected level. When the stop condition is met, the training process ends.

[0124] S15. Determine the optimal reminder behavior as the target intelligent reminder decision based on the probability distribution.

[0125] Through the forward propagation of the Actor network, the probability distribution corresponding to each intelligent reminder behavior in the current state is obtained, and the intelligent reminder behavior with the highest probability value is selected as the optimal intelligent reminder decision according to the maximum probability principle. For example, if the output probability distribution shows that "reminding the elderly to exercise appropriately" has the highest probability, then this reminder behavior is determined as the most suitable decision for the elderly at present, and the corresponding reminder is issued to the elderly or their caregivers to help the elderly maintain a healthy life and good living habits.

[0126] The above-mentioned A3C algorithm is used to determine the optimal intelligent reminder decision based on the comprehensive feature vector after time series fusion, so as to provide personalized and accurate intelligent reminder services according to the actual status of the elderly and improve the elderly's user experience.

[0127] Embodiment 2

[0128] The embodiment of the intelligent reminder system for assisting the elderly in the present invention comprises the following steps:

[0129] A data information collection unit, used to collect health status data and life behavior data information of the elderly in multiple dimensions based on wearable devices and smart home devices;

[0130] The feature extraction unit for the elderly is used to extract the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly after data preprocessing the collected data information;

[0131] A target fusion feature vector determination unit is used to fuse the heart rate variability of the elderly with the life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fusion feature vector;

[0132] A global actor network learning and training unit is used to input the target fusion feature vector into the trained global actor network, and the trained global actor network outputs the probability distribution of each possible reminder behavior;

[0133] The target intelligent reminder decision determination unit is used to determine the optimal reminder behavior as the target intelligent reminder decision according to the probability distribution.

[0134] Embodiment 3

[0135] The present invention provides a computer device, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the steps of the above method are implemented.

[0136] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0137] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored. In addition, each functional unit in each embodiment of the present invention can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0138] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0139] It can be understood that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. An intelligent reminder method for assisting the elderly, characterized in that: include: Collect health status data and life behavior data of the elderly in multiple dimensions based on wearable devices and smart home devices; After preprocessing the collected data information, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted; Fusing the heart rate variability of the elderly with life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fused feature vector; Inputting the target fused feature vector into a trained global Actor network, the trained global Actor network outputs a probability distribution of each possible reminder behavior; An optimal reminder behavior is determined according to the probability distribution as a target intelligent reminder decision.

2. The intelligent reminder method for assisting the elderly according to claim 1, characterized in that: After the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including: Heart rate variability and life rhythm characteristics: Where: M is the number of windows in a day, SDNN w is the SDNN standard deviation of the time interval between two adjacent heartbeats in the wth acquisition window, SDNN avg is the SDNN average of all windows in one day, σ SDNN is the change of SDNN between different sampling windows; RMSSD w is the root mean square of the difference between the time intervals of two adjacent heartbeats in the wth acquisition window, RMSSD avg is the average RMSSD of all windows in one day, σ RMSSD is the fluctuation of RMSSD in different periods; LF w and HF w are the sum of the power of the low-frequency component and the high-frequency component in the w-th acquisition window, LFHF w To measure the balance between low-frequency components and high-frequency components, LF avg HF avg and LFHF avg They are the average value of low-frequency components, the average value of high-frequency components, and the average value of the ratio of low-frequency components to high-frequency components; Among them: SDT d SDT is the total time that the elderly person is in deep sleep on the dth day out of n days. avg and σ SDT NW is the mean and standard deviation of deep sleep time of the elderly; d NW is the number of times the elderly wake up from sleep on day d. avg and σ NW is the mean and standard deviation of the number of awakenings at night; N is the total sleep cycle, SC D is the length of the D sleep cycle of the elderly, SC avg and σ SC are the mean duration and standard deviation of sleep cycles, respectively.

3. The intelligent reminder method for assisting the elderly according to claim 2, characterized in that: After the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including: Physical activity characteristics: Among them: SS i is the step length of the i-th step, SL i is the pace of the ith step, T i is the time period of the i-th step, N′ is the total number of steps, SL avg and σ SL are the mean and standard deviation of the step length, SS avg and σ SS are the mean and standard deviation of the pace, GC i is the gait cycle of the i-th step, GC avg and σ GC are the mean and standard deviation of the gait cycle, The time difference between the left and right feet landing. and are the mean and standard deviation of the difference in landing time between the left and right feet, respectively.

4. The intelligent reminder method for assisting the elderly according to claim 2, characterized in that: After the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including: Cognitive and interactive features: Among them: OPE count is the total number of operational errors within T′, OPE avg is the average number of operational errors per day during T′, FMR count The number of times the elderly forget to check or perform the set reminder tasks, FMR rate To remind the task forgetting rate, NT total is the total number of reminder tasks set within T′, AD ratio is the ratio of the number of different applications used by the elderly within T′ to the total number of applications on the device, X is the total number of applications installed on the device, and x is the number of applications suitable for the elderly within T′.

5. The intelligent reminder method for assisting the elderly according to claim 2, characterized in that: After the collected data information is preprocessed, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly are extracted, including: Lifestyle characteristics: in: is the breakfast time on the dth day within n days, and are the mean and standard deviation of breakfast time, is the lunch time on the dth day within n days, and are the mean and standard deviation of lunch time, is the dinner time on the dth day within n days, and are the mean and standard deviation of dinner time, respectively; and D The food intake for breakfast, lunch and dinner are respectively is the total daily food intake, T and are the mean and standard deviation of total food intake within m days; is the time when the elderly take the jth drug on day d, and are the mean and standard deviation of the duration of taking the jth drug, is the actual dosage of the jth drug, The dose prescribed for the jth drug, DDR j,d is the dosage deviation rate of the jth drug.

6. The intelligent reminder method for assisting the elderly according to any one of claims 1 to 5, characterized in that: The heart rate variability of the elderly is fused with life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fused feature vector, including: Use the time series algorithm to select the sliding window size and obtain the feature vector in each window; In each window, the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics and life habit characteristics of the elderly are spliced ​​into a comprehensive feature vector in a preset order; The concatenated comprehensive feature vectors are weighted fused to obtain the target fused feature vector.

7. The intelligent reminder method for assisting the elderly according to claim 1, characterized in that: The target fusion feature vector is input into a trained global Actor network, and the trained global Actor network outputs a probability distribution of each possible reminder behavior, including: Build an environmental model that simulates the elderly's life scenes and intelligent reminder interactions; Initialize hyperparameters, including learning rate, discount factor, and exploration rate, and construct the network structures of the Actor network and the Critic network; Create multiple working threads and conduct intra-thread interaction and learning; After each worker thread independently performs a preset number of interactions and parameter updates, it synchronizes the updated parameters of the Actor and Critic networks to the global network parameters; Continuously monitor the training process and set stop conditions to determine when the training is finished.

8. The intelligent reminder method for assisting the elderly according to claim 7, characterized in that: The creation of multiple working threads and interaction and learning within the threads include: Obtain the target fusion feature vector at the current moment in each thread; Input the comprehensive feature vector of the current state into the Actor network, and obtain the probability distribution corresponding to each possible intelligent reminder behavior through the network's forward propagation; Execute the selected intelligent reminder action in the environment. The environment updates the state of the elderly according to the preset rules, and feeds back a reward signal and transfers to the next state. Use the Critic network to evaluate the value of the current state and the next state; The parameters of the Actor network and the Critic network are updated using the gradient ascent and gradient descent algorithms based on the calculated advantage estimates and the probability of selecting the action.

9. An intelligent reminder system for assisting the elderly, characterized in that: include: A data information collection unit, used to collect health status data and life behavior data information of the elderly in multiple dimensions based on wearable devices and smart home devices; The feature extraction unit for the elderly is used to extract the heart rate variability and life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics of the elderly after data preprocessing the collected data information; A target fusion feature vector determination unit is used to fuse the heart rate variability of the elderly with the life rhythm characteristics, physical activity characteristics, cognitive and interactive characteristics, and life habit characteristics to obtain a target fusion feature vector; A global Actor network learning and training unit, used for inputting the target fusion feature vector into a trained global Actor network, wherein the trained global Actor network outputs a probability distribution of each possible reminder behavior; The target intelligent reminder decision determining unit is used to determine the optimal reminder behavior as the target intelligent reminder decision according to the probability distribution.

10. A computer device, characterized in that: include: memories, transceivers, processors, and bus systems; Wherein, the memory is used to store programs; The processor is used to execute the program in the memory, including executing the intelligent reminder method for assisting the elderly according to any one of claims 1 to 8; The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other.

Images